From 01ab5d1fd5e2986f6c482647136a69a4ed7b5c36 Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Tue, 14 Jul 2026 13:28:51 -0300
Subject: [PATCH 01/14] =?UTF-8?q?chore(ci):=20add=20.mergify.yml=20to=20ma?=
 =?UTF-8?q?in=20=E2=80=94=20Mergify=20only=20reads=20config=20from=20the?=
 =?UTF-8?q?=20default=20branch=20(#7168)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .mergify.yml | 55 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 55 insertions(+)
 create mode 100644 .mergify.yml

diff --git a/.mergify.yml b/.mergify.yml
new file mode 100644
index 00000000000..131c6d71a99
--- /dev/null
+++ b/.mergify.yml
@@ -0,0 +1,55 @@
+# Mergify merge queue — WS3.4/D5 of the v3.8.49 quality/velocity master plan.
+#
+# WHY: ~85-100 active PR authors/month and 300+ PRs/week peaks, all merged by ONE
+# identity. The manual merge-train validated batches by hand; this queue automates
+# it with batching + automatic batch bisection (a red batch of N costs ~log2(N)
+# revalidations instead of N). Mergify Open Source plan: free, unlimited, public repo.
+#
+# GOVERNANCE (non-negotiable, mirrors CLAUDE.md Hard Rules #21/#22 + the owner's
+# pre-merge ⭐ gate):
+#   • A PR enters the queue ONLY via the `queue` label — applied by the owner (or a
+#     session acting for the owner) AFTER the pre-merge ⭐ report/decision. The label
+#     IS the merge approval; Mergify only executes it.
+#   • During a release-freeze (open issue labeled `release-freeze`), do NOT label PRs
+#     targeting the frozen branch — the freeze is a human-honored coordination signal
+#     the queue cannot see. Retarget to the active release/vX+1 first (Hard Rule #21).
+#   • Never label a PR another session is actively working (Hard Rule #22b).
+#   • Fallback path if Mergify misbehaves or the OSS plan changes: the manual
+#     merge-train runbook (docs/ops/MERGE_TRAIN.md) — remove labels, proceed by hand.
+
+queue_rules:
+  - name: release
+    # Any current or future release branch — the reason GitHub's native queue was
+    # rejected (no wildcard support on personal-account repos).
+    queue_conditions:
+      - base~=^release/v\d+\.\d+\.\d+$
+      - label=queue
+      - -draft
+      - -conflict
+    # "Everything that ran is green, nothing still running, AND the always-on
+    # anchor check succeeded" — robust to the path-filtered fast-gates (docs-only
+    # PRs skip code jobs; matrix shard names vary) while never fail-open: a PR with
+    # zero checks cannot vacuously merge, because `Merge integrity` runs on EVERY
+    # non-draft PR (quality.yml) and must be an affirmative success. Review approval
+    # is intentionally NOT a condition here: the owner-applied `queue` label IS the
+    # approval in this repo's single-maintainer model (see governance header).
+    merge_conditions:
+      - "#check-failure=0"
+      - "#check-pending=0"
+      - "#check-success>=1"
+      - check-success=Merge integrity (changelog + generated skills)
+    # Batching: validate up to 10 queued PRs together (the manual train's sweet spot);
+    # don't hold a lone PR hostage waiting for siblings.
+    batch_size: 10
+    batch_max_wait_time: 5 min
+    # Squash keeps the one-commit-per-PR history the CHANGELOG reconciliation expects.
+    merge_method: squash
+
+pull_request_rules:
+  - name: clean up the queue label after merge
+    conditions:
+      - merged
+    actions:
+      label:
+        remove:
+          - queue

From fb24740b73443c33e5e6f7ee8b781fa92e5e4cc1 Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Tue, 14 Jul 2026 19:11:35 -0300
Subject: [PATCH 02/14] fix(ci): add the auto-enqueue pull_request_rule to the
 Mergify config (queue_conditions alone are eligibility-only) (#7179)

---
 .mergify.yml | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/.mergify.yml b/.mergify.yml
index 131c6d71a99..88bc431369e 100644
--- a/.mergify.yml
+++ b/.mergify.yml
@@ -46,6 +46,18 @@ queue_rules:
     merge_method: squash
 
 pull_request_rules:
+  # AUTO-ENQUEUE trigger — queue_conditions alone only define ELIGIBILITY in current
+  # Mergify semantics (without this rule the PR sits at "use @Mergifyio queue",
+  # observed live on PR #7175). The conditions mirror queue_conditions on purpose.
+  - name: queue on owner-applied label
+    conditions:
+      - base~=^release/v\d+\.\d+\.\d+$
+      - label=queue
+      - -draft
+      - -conflict
+    actions:
+      queue:
+        name: release
   - name: clean up the queue label after merge
     conditions:
       - merged

From fb25ebdeb0dbd924d6c7f95fd284759e3f87d7c2 Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Tue, 14 Jul 2026 23:49:49 -0300
Subject: [PATCH 03/14] fix(ci): migrate Mergify auto-enqueue to
 merge_protections_settings.auto_merge_conditions (rules-based path is EOL
 2026-07-16) (#7216)

---
 .mergify.yml | 19 +++++++------------
 1 file changed, 7 insertions(+), 12 deletions(-)

diff --git a/.mergify.yml b/.mergify.yml
index 88bc431369e..d7eb728c0fd 100644
--- a/.mergify.yml
+++ b/.mergify.yml
@@ -17,6 +17,13 @@
 #   • Fallback path if Mergify misbehaves or the OSS plan changes: the manual
 #     merge-train runbook (docs/ops/MERGE_TRAIN.md) — remove labels, proceed by hand.
 
+# Auto-enqueue (current Mergify model, 2026): auto_merge_conditions in
+# merge_protections_settings — the rules-based queue action / autoqueue path is
+# deprecated (EOL 2026-07-16). The owner-applied `queue` label IS the approval.
+merge_protections_settings:
+  auto_merge_conditions:
+    - label = queue
+
 queue_rules:
   - name: release
     # Any current or future release branch — the reason GitHub's native queue was
@@ -46,18 +53,6 @@ queue_rules:
     merge_method: squash
 
 pull_request_rules:
-  # AUTO-ENQUEUE trigger — queue_conditions alone only define ELIGIBILITY in current
-  # Mergify semantics (without this rule the PR sits at "use @Mergifyio queue",
-  # observed live on PR #7175). The conditions mirror queue_conditions on purpose.
-  - name: queue on owner-applied label
-    conditions:
-      - base~=^release/v\d+\.\d+\.\d+$
-      - label=queue
-      - -draft
-      - -conflict
-    actions:
-      queue:
-        name: release
   - name: clean up the queue label after merge
     conditions:
       - merged

From a7956945355d8b4dbdf9df531f5f53fb4d4d4274 Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Wed, 15 Jul 2026 00:16:02 -0300
Subject: [PATCH 04/14] fix(ci): drop Mergify batch settings (batching is a
 paid-tier feature; free plan queue is serial) (#7220)

---
 .mergify.yml | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/.mergify.yml b/.mergify.yml
index d7eb728c0fd..f7cbe19d7ea 100644
--- a/.mergify.yml
+++ b/.mergify.yml
@@ -45,10 +45,9 @@ queue_rules:
       - "#check-pending=0"
       - "#check-success>=1"
       - check-success=Merge integrity (changelog + generated skills)
-    # Batching: validate up to 10 queued PRs together (the manual train's sweet spot);
-    # don't hold a lone PR hostage waiting for siblings.
-    batch_size: 10
-    batch_max_wait_time: 5 min
+    # NO batching: 'Merge Queue Batch' requires a paid Mergify tier (live finding
+    # 2026-07-15 — the queue command fails with "Cannot use Merge Queue batch" on
+    # the free plan). Serial queue (1 PR at a time) still automates the train.
     # Squash keeps the one-commit-per-PR history the CHANGELOG reconciliation expects.
     merge_method: squash
 

From 9875ccf4e62fb5a26e54f712e192647293bf646b Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Wed, 15 Jul 2026 00:48:07 -0300
Subject: [PATCH 05/14] fix(ci): merge queue tolerates the advisory dast-smoke
 failure (its GH-hosted build hang dequeued every attempt) (#7225)

---
 .mergify.yml | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/.mergify.yml b/.mergify.yml
index f7cbe19d7ea..7cc4882127e 100644
--- a/.mergify.yml
+++ b/.mergify.yml
@@ -41,7 +41,14 @@ queue_rules:
     # is intentionally NOT a condition here: the owner-applied `queue` label IS the
     # approval in this repo's single-maintainer model (see governance header).
     merge_conditions:
-      - "#check-failure=0"
+      # "Zero failures" — EXCEPT the advisory dast-smoke: it is continue-on-error by
+      # design and its GH-hosted CLI build hangs recurrently (drafts #7184/#7221 were
+      # dequeued solely by it). Any OTHER failure still blocks (anti-fail-open kept).
+      - or:
+          - "#check-failure=0"
+          - and:
+              - "#check-failure=1"
+              - check-failure=dast-smoke
       - "#check-pending=0"
       - "#check-success>=1"
       - check-success=Merge integrity (changelog + generated skills)

From 0065f1b7f03d0cfa27a42491e92526e73d9b44fd Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Thu, 16 Jul 2026 00:12:29 -0300
Subject: [PATCH 06/14] =?UTF-8?q?test(ci):=20make=20the=20#6634=20selfref?=
 =?UTF-8?q?=20guard=20hermetic=20=E2=80=94=20main's=20copy=20hard-fails=20?=
 =?UTF-8?q?every=20PR=20(#7341)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

main's copy of this test still does git I/O inside a unit test:

    const baseSrc = git(['show', 'origin/main:' + FILE]);

Runners check out a shallow single ref, so origin/main does not resolve and the
test dies with 'fatal: invalid object name origin/main'. Every PR into main
fails Unit Tests (7/8) on it — today that is #7313, #7315, #7316, #7334, #7336
and #7337, six PRs red on a defect none of them introduced. #7313 has no other
red at all.

release/v3.8.49 already carries a fix (2e42b8efc, #7174: try/catch, fetch
origin/main on demand, t.skip() when unreachable), but it only reaches main at
release time — so main stays broken for the whole cycle. Cherry-picking it would
also import a new problem: PR Test Policy classifies t.skip() as a silenced
assertion, which we watched it correctly catch on #7300 today.

This is the hermetic version instead (ported from #7327, which does the same for
the release branch): read the file straight off disk, compare against an empty
base so baseTaut/baseExtTaut are 0 — the strictest possible comparison point —
and call evaluateMasking() directly. No git ref, no fetch, no skip, nothing the
runner's checkout depth can break.

The #6634 regression stays covered: the guard's logic lives in
SELF_TEST_FIXTURE_RE (check-test-masking.mjs:337), not in the test. Proven both
ways on main before committing — neutralise SELF_TEST_FIXTURE_RE to /$^/ and
the test FAILS; restore it and it passes 2/2, with check-test-masking.mjs left
byte-identical.

Co-authored-by: growab <nekron@icloud.com>
---
 .../check-test-masking-selfref-6634.test.ts   | 23 +++++++++++--------
 1 file changed, 13 insertions(+), 10 deletions(-)

diff --git a/tests/unit/check-test-masking-selfref-6634.test.ts b/tests/unit/check-test-masking-selfref-6634.test.ts
index 6d91a2e7c21..b233a008f22 100644
--- a/tests/unit/check-test-masking-selfref-6634.test.ts
+++ b/tests/unit/check-test-masking-selfref-6634.test.ts
@@ -14,12 +14,14 @@
  * (`if (file.endsWith("check-test-masking.test.ts")) continue;` in
  * scripts/check/check-test-masking.mjs) for precisely this reason — this test
  * asserts evaluateMasking() now applies the same exclusion for its diff-based
- * tautology counters, using the real base(origin/main)/head(HEAD) diff of
+ * tautology counters, against the REAL current source of
  * tests/unit/check-test-masking.test.ts.
  */
 import test from "node:test";
 import assert from "node:assert/strict";
-import { execFileSync } from "node:child_process";
+import fs from "node:fs";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
 
 import {
   countTautologies,
@@ -28,16 +30,17 @@ import {
 } from "../../scripts/check/check-test-masking.mjs";
 
 const FILE = "tests/unit/check-test-masking.test.ts";
-
-function git(args: string[]): string {
-  return execFileSync("git", args, { encoding: "utf8" });
-}
+const REPO_ROOT = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../..");
 
 test("#6634: check-test-masking.test.ts's own tautology fixtures must not self-flag as weakening", () => {
-  // origin/main predates the #6404 fixtures (countBareTautologies/scanBareTautologies
-  // tests) that legitimately embed tautology-pattern literals as string fixtures.
-  const baseSrc = git(["show", "origin/main:" + FILE]);
-  const headSrc = git(["show", "HEAD:" + FILE]);
+  // Read the REAL current source from disk rather than a git ref: the Unit Tests
+  // job checks out a shallow/single-ref tree with no origin/main, so `git show
+  // origin/main:<file>` failed the shard before it ever exercised the masking
+  // behavior under test. An empty base models the file's pre-#6404 state (no
+  // fixtures), which maximizes headTaut - baseTaut — the strictest input for the
+  // exclusion this test asserts.
+  const baseSrc = "";
+  const headSrc = fs.readFileSync(path.join(REPO_ROOT, FILE), "utf8");
 
   const perFile = [
     {

From ddd6d09cd10d7a2d6af793a0bfe20d96c05db25c Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Thu, 16 Jul 2026 00:47:07 -0300
Subject: [PATCH 07/14] chore(quality): tighten main's coverage baseline to the
 CI's real numbers (#7347)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

main's ratchet had been failing --require-tighten on every PR: 11 metrics
improved but the baseline was never tightened. Same class as the #6634
selfref guard — an infra fix that lands only on the release branch leaves
main red for the whole cycle, and every PR into main pays for it.

Values are the merged-coverage numbers from a run on main itself (a local
run measures ~68% vs CI's ~80%; the baseline's own note warns about that
gap). Only the 11 coverage values change — gitleaks and semgrepFindings
keep main's own state.

No changelog fragment: #7326 carries it on release/v3.8.49, and a second
one here would double the entry at release time.
---
 config/quality/quality-baseline.json | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

diff --git a/config/quality/quality-baseline.json b/config/quality/quality-baseline.json
index f660fc33d30..943b0fa4b88 100644
--- a/config/quality/quality-baseline.json
+++ b/config/quality/quality-baseline.json
@@ -27,22 +27,22 @@
       "eps": 0
     },
     "coverage.statements": {
-      "value": 76.5,
+      "value": 80.8,
       "direction": "up",
       "tightenSlack": 5
     },
     "coverage.lines": {
-      "value": 76.5,
+      "value": 80.8,
       "direction": "up",
       "tightenSlack": 5
     },
     "coverage.functions": {
-      "value": 82,
+      "value": 86.44,
       "direction": "up",
       "tightenSlack": 5
     },
     "coverage.branches": {
-      "value": 73,
+      "value": 78.1,
       "direction": "up",
       "eps": 1.5,
       "tightenSlack": 5
@@ -54,43 +54,43 @@
       "tightenSlack": 10
     },
     "coverage.combo.lines": {
-      "value": 80,
+      "value": 85.42,
       "direction": "up",
       "eps": 1.5,
       "tightenSlack": 10
     },
     "coverage.accountFallback.lines": {
-      "value": 88,
+      "value": 96.78,
       "direction": "up",
       "eps": 1.5,
       "tightenSlack": 10
     },
     "coverage.auth.lines": {
-      "value": 90,
+      "value": 92.55,
       "direction": "up",
       "eps": 1.5,
       "tightenSlack": 10
     },
     "coverage.routeGuard.lines": {
-      "value": 94,
+      "value": 98.73,
       "direction": "up",
       "eps": 1.5,
       "tightenSlack": 10
     },
     "coverage.error.lines": {
-      "value": 88,
+      "value": 92.13,
       "direction": "up",
       "eps": 1.5,
       "tightenSlack": 10
     },
     "coverage.publicCreds.lines": {
-      "value": 92,
+      "value": 99.07,
       "direction": "up",
       "eps": 1.5,
       "tightenSlack": 10
     },
     "coverage.circuitBreaker.lines": {
-      "value": 92,
+      "value": 95.09,
       "direction": "up",
       "eps": 1.5,
       "tightenSlack": 10

From 37cad3eea50f7860e282b54cff8d8f030b45b75e Mon Sep 17 00:00:00 2001
From: Minxi Hou <houminxi@gmail.com>
Date: Thu, 16 Jul 2026 00:48:24 -0400
Subject: [PATCH 08/14] fix(antigravity): remove hardcoded 120s SSE collect
 timeout

The SSE collection in collectStreamToResponse had a hardcoded 120 s
timeout.  Reasoning-heavy models like gemini-3.1-pro-high on large
prompts (>30 KB) regularly exceed 120 s of generation time, causing
the executor to return a synthetic 504 before the model finishes.

Replace the hardcoded value with FETCH_TIMEOUT_MS (default 600 s,
overridable via FETCH_TIMEOUT_MS env var), which is the standard
upstream-request budget across all OmniRoute providers.

Signed-off-by: Minxi Hou <houminxi@gmail.com>
---
 open-sse/executors/antigravity.ts | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/open-sse/executors/antigravity.ts b/open-sse/executors/antigravity.ts
index 494d750836e..83b014002dc 100644
--- a/open-sse/executors/antigravity.ts
+++ b/open-sse/executors/antigravity.ts
@@ -13,6 +13,7 @@ import {
   PROVIDERS,
   OAUTH_ENDPOINTS,
   HTTP_STATUS,
+  FETCH_TIMEOUT_MS,
   STREAM_READINESS_TIMEOUT_MS,
   ANTIGRAVITY_PRE_RESPONSE_TIMEOUT_CODE,
 } from "../config/constants.ts";
@@ -948,7 +949,11 @@ export class AntigravityExecutor extends BaseExecutor {
     const decoder = new TextDecoder();
     const logger = log || undefined;
 
-    const SSE_COLLECT_TIMEOUT_MS = 120_000;
+    // Guard against indefinite hangs when the upstream sends headers but
+    // stalls on the body.  Inherit the global FETCH_TIMEOUT_MS (default 600 s,
+    // overridable via env) so reasoning-heavy models (gemini-3.1-pro-high on
+    // large prompts) are not killed by a hardcoded 120 s ceiling.
+    const SSE_COLLECT_TIMEOUT_MS = FETCH_TIMEOUT_MS;
 
     const collect = async () => {
       const collected: AntigravityCollectedStream = {

From e33937eef3583eeab8c8e417ae871ca5546a158f Mon Sep 17 00:00:00 2001
From: Minxi Hou <houminxi@gmail.com>
Date: Sat, 18 Jul 2026 07:16:17 -0400
Subject: [PATCH 09/14] fix(antigravity): streaming passthrough for
 non-streaming clients

When a client sends stream: false to the Antigravity executor
(Gemini models), OmniRoute buffered the entire SSE stream before
responding. Long-thinking models exceeded the 120s timeout.

Remove hardcoded SSE_COLLECT_TIMEOUT_MS. Extract shared
createCreditsExtractionTransform with 16KB buffer cap and abort
handling for client disconnect. Add parseSSEToGeminiResponse for
the non-streaming drain path. Fix hasGeminiTerminalFinishReason
to check top-level candidates (no response wrapper). Add signal
null guards for credits retry path. Return 499 on early abort
instead of piping cancelled body.

Also remove duplicate SKILLS_SANDBOX_RUNTIME from .env.example
and clarify .artifacts/ vs _artifacts/ in .gitignore.

Signed-off-by: Minxi Hou <houminxi@gmail.com>
---
 .env.example                                  |   2 +-
 .gitignore                                    |   4 +-
 config/quality/eslint-suppressions.json       |   5 -
 open-sse/executors/antigravity.ts             | 302 ++++++++++--------
 open-sse/handlers/chatCore/nonStreamingSse.ts |  43 ++-
 open-sse/handlers/sseParser.ts                | 153 +++++++++
 tests/unit/executor-antigravity.test.ts       | 120 ++++++-
 tests/unit/sse-parser.test.ts                 | 129 +++++++-
 8 files changed, 599 insertions(+), 159 deletions(-)

diff --git a/.env.example b/.env.example
index 60b5074e6e7..bfc6942f2dc 100644
--- a/.env.example
+++ b/.env.example
@@ -222,7 +222,7 @@ CONTAINER_HOST=docker
 # - orbstack: OrbStack (high-perf Linux VM + docker shim on macOS)
 # - podman:  Podman (rootless, daemonless)
 # - docker:  Docker (default fallback)
-SKILLS_SANDBOX_RUNTIME=auto
+# (defined under SKILLS & SANDBOXING section below)
 
 # ═══════════════════════════════════════════════════════════════════════════════
 # 4. SECURITY & AUTHENTICATION
diff --git a/.gitignore b/.gitignore
index 67d69b3d5d6..e3064a2700e 100644
--- a/.gitignore
+++ b/.gitignore
@@ -232,7 +232,7 @@ omniroute.md
 
 # mise configuration
 mise.toml
-_artifacts/
+_artifacts/          # release-green artifacts
 .claude-flow/
 
 # ESLint file cache (npm run lint --cache / complexity ratchets)
@@ -240,5 +240,5 @@ _artifacts/
 .eslintcache-complexity
 
 
-# CI/local quality artifacts (eslint-results.json, etc.)
+# CI/local quality artifacts (eslint-results.json, quality-ratchet.md, etc.)
 .artifacts/
diff --git a/config/quality/eslint-suppressions.json b/config/quality/eslint-suppressions.json
index dcea92f2a98..5cf9b231b26 100644
--- a/config/quality/eslint-suppressions.json
+++ b/config/quality/eslint-suppressions.json
@@ -2057,11 +2057,6 @@
       "count": 13
     }
   },
-  "tests/unit/sse-parser.test.ts": {
-    "@typescript-eslint/no-explicit-any": {
-      "count": 15
-    }
-  },
   "tests/unit/startup-stale-cooldown-recovery.test.ts": {
     "@typescript-eslint/no-explicit-any": {
       "count": 1
diff --git a/open-sse/executors/antigravity.ts b/open-sse/executors/antigravity.ts
index 83b014002dc..3647e82df07 100644
--- a/open-sse/executors/antigravity.ts
+++ b/open-sse/executors/antigravity.ts
@@ -272,6 +272,91 @@ export function updateAntigravityRemainingCredits(accountId: string, balance: nu
   } catch {}
 }
 
+/**
+ * Create a pass-through TransformStream that extracts `remainingCredits`
+ * from SSE data without consuming the stream.  The downstream client
+ * receives the unmodified bytes.
+ *
+ * @param accountId  Provider account ID for credit-balance persistence.
+ * @param bufferSize  Optional sliding-window buffer cap in bytes.
+ *                   Pass 0 or omit for unlimited (non-streaming callers
+ *                   where the full body is already buffered upstream).
+ *                   The streaming path uses 16384 (16 KB) to prevent OOM
+ *                   on long-lived SSE connections.  Credit-balance data
+ *                   appears near the end of the SSE stream (after
+ *                   content), so the sliding window captures it even at
+ *                   16 KB -- only truly massive responses (>16 KB of
+ *                   consecutive non-newline content) would lose credits.
+ * @internal Exported for unit testing only.
+ */
+export function createCreditsExtractionTransform(
+  accountId: string,
+  bufferSize = 0
+): TransformStream<Uint8Array, Uint8Array> {
+  let buffer = "";
+  const decoder = new TextDecoder();
+
+  return new TransformStream(
+    {
+      transform(chunk, controller) {
+        controller.enqueue(chunk);
+        try {
+          buffer += decoder.decode(chunk, { stream: true });
+          // Sliding-window cap: truncate after the last complete newline
+          // in the discard region so SSE lines are never split mid-payload.
+          if (bufferSize > 0 && buffer.length > bufferSize) {
+            const lastNewline = buffer.lastIndexOf("\n", buffer.length - bufferSize);
+            if (lastNewline !== -1) {
+              buffer = buffer.slice(lastNewline + 1);
+            } else {
+              // No newline in the discard region -- incomplete line, discard entirely.
+              buffer = "";
+            }
+          }
+        } catch {
+          /* decoding best-effort */
+        }
+      },
+      flush() {
+        try {
+          buffer += decoder.decode();
+        } catch {
+          /* decoding best-effort */
+        }
+        try {
+          const lines = buffer.split("\n");
+          for (const line of lines) {
+            const trimmed = line.trim();
+            if (!trimmed.startsWith("data:")) continue;
+            const payload = trimmed.slice(5).trim();
+            if (!payload || payload === "[DONE]") continue;
+            try {
+              const parsed = JSON.parse(payload);
+              if (Array.isArray(parsed?.remainingCredits)) {
+                const googleCredit = parsed.remainingCredits.find((c: unknown) => {
+                  const credit = asRecord(c);
+                  return credit?.creditType === "GOOGLE_ONE_AI";
+                }) as AntigravityCreditEntry | undefined;
+                if (googleCredit) {
+                  const balance = parseInt(String(googleCredit.creditAmount ?? ""), 10);
+                  if (!isNaN(balance)) updateAntigravityRemainingCredits(accountId, balance);
+                }
+              }
+            } catch {
+              /* skip malformed lines */
+            }
+          }
+        } catch {
+          /* credits extraction is best-effort */
+        }
+        buffer = "";
+      },
+    },
+    { highWaterMark: 16384 },
+    { highWaterMark: 16384 }
+  );
+}
+
 function isCreditsExhausted(accountId: string): boolean {
   const until = creditsExhaustedUntil.get(accountId);
   if (!until) return false;
@@ -931,6 +1016,10 @@ export class AntigravityExecutor extends BaseExecutor {
    * Collect an SSE streaming response into a single non-streaming JSON response.
    * Parses Gemini-format SSE chunks and assembles text content + usage into one
    * OpenAI-format chat.completion payload.
+   *
+   * @deprecated Use the non-streaming SSE path in chatCore instead, which calls
+   * parseSSEToGeminiResponse() from sseParser.ts.  This method is retained only
+   * for backward compatibility and may be removed in a future release.
    */
   collectStreamToResponse(
     response: Response,
@@ -1375,33 +1464,40 @@ export class AntigravityExecutor extends BaseExecutor {
                   });
                   if (creditsResp.ok || creditsResp.status !== HTTP_STATUS.RATE_LIMITED) {
                     log?.info?.("AG_CREDITS", `Credits retry succeeded: ${creditsResp.status}`);
-                    if (!stream) {
-                      const collected = await this.collectStreamToResponse(
-                        creditsResp,
-                        model,
-                        url,
-                        finalCreditsHeaders,
-                        creditsBody,
-                        log,
-                        signal
+                    if (!stream && creditsResp.body) {
+                      // Client already disconnected — skip pipe
+                      if (signal?.aborted) {
+                        creditsResp.body.cancel().catch(() => {});
+                        return {
+                          response: new Response(null, { status: 499 }),
+                          url,
+                          headers: finalCreditsHeaders,
+                          transformedBody: null,
+                        };
+                      }
+                      // Return raw SSE with credits extraction
+                      const crTransform = createCreditsExtractionTransform(
+                        accountId,
+                        16 * 1024 // 16KB cap
                       );
-                      // Parse _remainingCredits from the synthetic response and cache
-                      try {
-                        const syntheticJson = await collected.response.clone().json();
-                        const rc = syntheticJson?._remainingCredits;
-                        if (Array.isArray(rc)) {
-                          const googleCredit = rc.find((c) => c.creditType === "GOOGLE_ONE_AI");
-                          if (googleCredit) {
-                            const balance = parseInt(googleCredit.creditAmount, 10);
-                            if (!isNaN(balance))
-                              updateAntigravityRemainingCredits(accountId, balance);
-                          }
-                        }
-                      } catch {
-                        /**/
+                      const crTappedBody = creditsResp.body.pipeThrough(crTransform);
+                      if (signal) {
+                        signal.addEventListener(
+                          "abort",
+                          () => {
+                            creditsResp.body.cancel().catch(() => {});
+                          },
+                          { once: true }
+                        );
                       }
                       return {
-                        ...collected,
+                        response: new Response(crTappedBody, {
+                          status: creditsResp.status,
+                          statusText: creditsResp.statusText,
+                          headers: creditsResp.headers,
+                        }),
+                        url,
+                        headers: finalCreditsHeaders,
                         transformedBody: attachToolNameMap(creditsBody, requestToolNameMap),
                       };
                     }
@@ -1537,12 +1633,16 @@ export class AntigravityExecutor extends BaseExecutor {
           }
         }
 
-        // For non-streaming clients, collect the SSE stream and return a synthetic
-        // non-streaming Response so chatCore doesn't need to handle SSE conversion.
+        // For non-streaming clients, return the raw SSE stream with a
+        // credits-extraction TransformStream.  chatCore's non-streaming path
+        // (readNonStreamingResponseBody + parseNonStreamingSSEPayload with
+        // Gemini format support) handles draining and conversion to JSON.
+        // This replaces the previous collectStreamToResponse() approach which
+        // had an artificial timeout (now the standard FETCH_BODY_TIMEOUT_MS
+        // of 10 min applies).
         if (!stream) {
           // #3229: surface a real upstream error instead of masking a 4xx/5xx as an
-          // empty `chat.completion` envelope (collectStreamToResponse synthesizes a
-          // success-shaped body when the upstream returned no SSE data).
+          // empty `chat.completion` envelope.
           if (!response.ok) {
             const rawBody = await response
               .clone()
@@ -1563,35 +1663,53 @@ export class AntigravityExecutor extends BaseExecutor {
               transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
             };
           }
-          const collected = await this.collectStreamToResponse(
-            response,
-            model,
-            url,
-            finalHeaders,
-            transformedBody,
-            log,
-            signal
-          );
-          // When credits were injected (credits-first or credits-retry), the
-          // synthetic body contains _remainingCredits — mirror it into the
-          // balance cache so the dashboard stays fresh.
-          try {
-            const syntheticJson = await collected.response.clone().json();
-            const rc = syntheticJson?._remainingCredits;
-            if (Array.isArray(rc)) {
-              const googleCredit = rc.find(
-                (c: { creditType?: string }) => c?.creditType === "GOOGLE_ONE_AI"
+
+          if (response.body) {
+            // Client already disconnected — skip pipe
+            if (signal?.aborted) {
+              response.body.cancel().catch(() => {});
+              return {
+                response: new Response(null, { status: 499 }),
+                url,
+                headers: finalHeaders,
+                transformedBody: null,
+              };
+            }
+            // Cancel upstream body on client disconnect
+            if (signal) {
+              signal.addEventListener(
+                "abort",
+                () => {
+                  response.body?.cancel().catch(() => {});
+                },
+                { once: true }
               );
-              if (googleCredit) {
-                const balance = parseInt(googleCredit.creditAmount, 10);
-                if (!isNaN(balance)) updateAntigravityRemainingCredits(accountId, balance);
-              }
             }
-          } catch {
-            /* balance cache is best-effort */
+
+            // Tap the stream to extract remainingCredits while passing
+            // data through unmodified.  chatCore drains the full body.
+            const nsPassThrough = createCreditsExtractionTransform(
+              accountId,
+              16 * 1024 // 16KB sliding-window cap to prevent OOM
+            );
+            const tappedBody = response.body.pipeThrough(nsPassThrough);
+            return {
+              response: new Response(tappedBody, {
+                status: response.status,
+                statusText: response.statusText,
+                headers: response.headers,
+              }),
+              url,
+              headers: finalHeaders,
+              transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
+            };
           }
+
+          // No body -- return as-is
           return {
-            ...collected,
+            response,
+            url,
+            headers: finalHeaders,
             transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
           };
         }
@@ -1615,81 +1733,9 @@ export class AntigravityExecutor extends BaseExecutor {
             }
           }
 
-          let sseBuffer = "";
-          const decoder = new TextDecoder(); // Singleton for correct streaming decode
-          const MAX_BUFFER_SIZE = 16 * 1024; // Limit to prevent OOM on large streams
-
-          const passThrough = new TransformStream(
-            {
-              transform(chunk, controller) {
-                controller.enqueue(chunk);
-                // Accumulate text to scan for remainingCredits
-                try {
-                  const text = decoder.decode(chunk, { stream: true });
-                  sseBuffer += text;
-                  // Limit buffer size to prevent unbounded growth
-                  // Truncate only after a complete newline to avoid splitting SSE lines mid-payload
-                  if (sseBuffer.length > MAX_BUFFER_SIZE) {
-                    const lastNewline = sseBuffer.lastIndexOf(
-                      "\n",
-                      sseBuffer.length - MAX_BUFFER_SIZE
-                    );
-                    if (lastNewline !== -1) {
-                      sseBuffer = sseBuffer.slice(lastNewline + 1);
-                    } else {
-                      // No newline found in discard region — buffer contains an incomplete SSE line.
-                      // Discard it entirely to avoid returning malformed data; the remainingCredits
-                      // parser won't find valid data in a truncated line anyway.
-                      sseBuffer = "";
-                    }
-                  }
-                } catch {
-                  /* decoding best-effort */
-                }
-              },
-              flush() {
-                // Final decode for any remaining bytes
-                try {
-                  const text = decoder.decode(); // Flush pending bytes
-                  sseBuffer += text;
-                } catch {
-                  /* decoding best-effort */
-                }
-
-                // Parse the accumulated SSE data for remainingCredits
-                try {
-                  const lines = sseBuffer.split("\n");
-                  for (const line of lines) {
-                    const trimmed = line.trim();
-                    if (!trimmed.startsWith("data:")) continue;
-                    const payload = trimmed.slice(5).trim();
-                    if (!payload || payload === "[DONE]") continue;
-                    try {
-                      const parsed = JSON.parse(payload);
-                      if (Array.isArray(parsed?.remainingCredits)) {
-                        const googleCredit = parsed.remainingCredits.find((c: unknown) => {
-                          const credit = asRecord(c);
-                          return credit?.creditType === "GOOGLE_ONE_AI";
-                        }) as AntigravityCreditEntry | undefined;
-                        if (googleCredit) {
-                          const balance = parseInt(String(googleCredit.creditAmount ?? ""), 10);
-                          if (!isNaN(balance)) {
-                            updateAntigravityRemainingCredits(accountId, balance);
-                          }
-                        }
-                      }
-                    } catch {
-                      /* skip malformed lines */
-                    }
-                  }
-                } catch {
-                  /* credits extraction is best-effort */
-                }
-                sseBuffer = "";
-              },
-            },
-            { highWaterMark: 16384 },
-            { highWaterMark: 16384 }
+          const passThrough = createCreditsExtractionTransform(
+            accountId,
+            16 * 1024 // 16KB sliding-window cap to prevent OOM
           );
           const tappedBody = response.body.pipeThrough(passThrough);
           const tappedResponse = new Response(tappedBody, {
diff --git a/open-sse/handlers/chatCore/nonStreamingSse.ts b/open-sse/handlers/chatCore/nonStreamingSse.ts
index 5d7e642fc62..8c2b748b978 100644
--- a/open-sse/handlers/chatCore/nonStreamingSse.ts
+++ b/open-sse/handlers/chatCore/nonStreamingSse.ts
@@ -3,6 +3,7 @@ import {
   parseSSEToResponsesOutput,
   parseSSEToClaudeResponse,
   parseSSEToOpenAIResponse,
+  parseSSEToGeminiResponse,
 } from "../sseParser.ts";
 import { getHeaderValueCaseInsensitive } from "./headers.ts";
 
@@ -20,6 +21,7 @@ export function parseNonStreamingSSEPayload(
   };
 
   queueFormat(preferredFormat);
+  queueFormat(FORMATS.GEMINI);
   queueFormat(FORMATS.OPENAI_RESPONSES);
   queueFormat(FORMATS.CLAUDE);
   queueFormat(FORMATS.OPENAI);
@@ -30,7 +32,9 @@ export function parseNonStreamingSSEPayload(
         ? parseSSEToResponsesOutput(rawBody, fallbackModel)
         : format === FORMATS.CLAUDE
           ? parseSSEToClaudeResponse(rawBody, fallbackModel)
-          : parseSSEToOpenAIResponse(rawBody, fallbackModel);
+          : format === FORMATS.GEMINI || format === FORMATS.ANTIGRAVITY
+            ? parseSSEToGeminiResponse(rawBody, fallbackModel)
+            : parseSSEToOpenAIResponse(rawBody, fallbackModel);
     if (parsed && typeof parsed === "object") {
       return {
         body: parsed as Record<string, unknown>,
@@ -107,6 +111,21 @@ function hasClaudeTerminalMessageDelta(parsed: unknown, eventType: string): bool
   return typeof stopReason === "string" ? stopReason.length > 0 : stopReason != null;
 }
 
+// Non-empty finishReason is terminal.  Gemini SSE payloads from
+// streamGenerateContent have candidates at the top level (no
+// "response" wrapper).  Any non-empty string signals stream end.
+function hasGeminiTerminalFinishReason(parsed: unknown): boolean {
+  if (!parsed || typeof parsed !== "object") return false;
+  // Top-level candidates (streamGenerateContent?alt=sse)
+  const obj = parsed as Record<string, unknown>;
+  const candidates = obj.candidates as unknown[] | undefined;
+  if (!Array.isArray(candidates) || candidates.length === 0) return false;
+  const candidate = candidates[0] as Record<string, unknown> | undefined;
+  if (!candidate || typeof candidate !== "object") return false;
+  const finishReason = candidate.finishReason;
+  return typeof finishReason === "string" && finishReason.length > 0;
+}
+
 function processNonStreamingSseTerminalLine(
   state: NonStreamingSseTerminalState,
   rawLine: string
@@ -130,12 +149,22 @@ function processNonStreamingSseTerminalLine(
 
   // Hot-path optimization: the terminal SSE events we look for (message_stop,
   // response.completed, …) all carry a top-level "type" field, OR are signalled by a
-  // preceding `event:` line (Claude). OpenAI chat.completion chunks carry neither and
-  // terminate with `[DONE]` (handled above), so parsing every one of them here is pure
-  // waste that compounds into the CPU-runaway on large buffered responses. Skip the
-  // JSON.parse unless the line could actually be a typed terminal.
+  // preceding `event:` line (Claude). Gemini signals completion via
+  // "finishReason" inside response.candidates[0]. OpenAI chat.completion chunks
+  // carry none of these and terminate with `[DONE]` (handled above), so parsing
+  // every one of them here is pure waste that compounds into the CPU-runaway on
+  // large buffered responses. Skip the JSON.parse unless the line could actually
+  // be a typed terminal.
   if (
     !data.includes('"type"') &&
+    // NOTE: "finishReason" is a superset match -- it triggers JSON.parse on
+    // every Gemini chunk that happens to contain the string (e.g. partial
+    // candidate payloads), not just the terminal one.  This is intentional:
+    // the extra parses are cheap compared to the CPU-runaway we'd get from
+    // parsing ALL chunks unconditionally on large buffered responses, and
+    // the superset is safe (false positives just parse a non-terminal chunk
+    // and fall through to `return false`).
+    !data.includes('"finishReason"') &&
     !(state.currentEvent === "message_delta" && data.includes("stop_reason"))
   ) {
     return isNonStreamingSseTerminalType(state.currentEvent);
@@ -148,7 +177,9 @@ function processNonStreamingSseTerminalLine(
         ? parsed.type
         : state.currentEvent;
     return (
-      isNonStreamingSseTerminalType(eventType) || hasClaudeTerminalMessageDelta(parsed, eventType)
+      isNonStreamingSseTerminalType(eventType) ||
+      hasClaudeTerminalMessageDelta(parsed, eventType) ||
+      hasGeminiTerminalFinishReason(parsed)
     );
   } catch {
     // Keep reading malformed data so the parser can report a useful upstream error.
diff --git a/open-sse/handlers/sseParser.ts b/open-sse/handlers/sseParser.ts
index d2e634e12ae..b9dc848059e 100644
--- a/open-sse/handlers/sseParser.ts
+++ b/open-sse/handlers/sseParser.ts
@@ -1,5 +1,6 @@
 import { appendToolCallArgumentDelta } from "../utils/toolCallArguments.ts";
 import { sanitizeErrorMessage } from "../utils/error.ts";
+import { normalizeOpenAICompatibleFinishReasonString } from "../utils/finishReason.ts";
 
 /**
  * Extract a provider error message from a buffered SSE stream that carries an
@@ -823,3 +824,155 @@ export function parseSSEToResponsesOutput(rawSSE, fallbackModel) {
     metadata: picked.metadata || {},
   };
 }
+
+/**
+ * Convert Gemini/Antigravity SSE chunks into a single non-streaming OpenAI
+ * chat.completion JSON response.  Gemini SSE carries payloads like:
+ *
+ *   data: {"markdown":"...chunk..."}
+ *   data: {"response":{"candidates":[{"content":{"parts":[{"text":"..."}]},"finishReason":"STOP"}],"usageMetadata":{...}}}
+ *   data: {"remainingCredits":[...]}
+ *
+ * Reuses the same parsing logic as processAntigravitySSEPayload() in sseCollect.ts
+ * so that format conversion is functionally equivalent to the previous
+ * collectStreamToResponse() approach.  Intentional differences:
+ *   - remainingCredits is NOT embedded into the result (handled separately
+ *     by the credits-extraction TransformStream in antigravity.ts).
+ *   - The synthetic `id` uses `chatcmpl-${Date.now()}` (no UUID suffix)
+ *     because this path runs once per response, not per chunk.
+ */
+export function parseSSEToGeminiResponse(
+  rawSSE: string,
+  fallbackModel: string
+): Record<string, unknown> | null {
+  const lines = String(rawSSE || "").split("\n");
+  let textContent = "";
+  let finishReason = "stop";
+  let usage: Record<string, unknown> | null = null;
+  let sawContent = false;
+
+  type AccumulatedToolCall = {
+    id: string;
+    index: number;
+    type: "function";
+    function: { name: string; arguments: string };
+  };
+  const toolCalls: AccumulatedToolCall[] = [];
+
+  const stripZeroWidth = (value: unknown): unknown => {
+    if (typeof value === "string") return value.replace(/[\u200B-\u200D\uFEFF]/g, "");
+    return value;
+  };
+
+  const tryParseTextualToolCall = (text: string): { name: string; args: unknown } | null => {
+    const normalized = text.replace(/[\u200B-\u200D\uFEFF]/g, "");
+    const match = normalized.match(
+      /^[\s\S]*?\[Tool call:\s*([^\]\n]+)\]\s*\nArguments:\s*([\s\S]+?)\s*$/
+    );
+    if (!match) return null;
+    const name = match[1]?.trim();
+    const rawArgs = match[2]?.trim();
+    if (!name || !rawArgs) return null;
+    try {
+      return { name, args: stripZeroWidth(JSON.parse(rawArgs)) };
+    } catch {
+      return null;
+    }
+  };
+
+  for (const line of lines) {
+    const trimmed = line.trim();
+    if (!trimmed.startsWith("data:")) continue;
+    const payload = trimmed.slice(5).trim();
+    if (!payload || payload === "[DONE]") continue;
+
+    try {
+      const parsed = JSON.parse(payload);
+
+      // Markdown shortcut (some Gemini variants)
+      const markdown =
+        typeof parsed?.markdown === "string"
+          ? parsed.markdown
+          : typeof parsed?.response?.markdown === "string"
+            ? parsed.response.markdown
+            : null;
+      if (markdown) {
+        textContent += markdown;
+        sawContent = true;
+      }
+
+      // Candidate content parts
+      const candidate = parsed?.response?.candidates?.[0];
+      if (candidate?.content?.parts) {
+        for (const part of candidate.content.parts) {
+          if (typeof part.text === "string" && !part.thought && !part.thoughtSignature) {
+            const textualToolCall = tryParseTextualToolCall(part.text);
+            if (textualToolCall) {
+              toolCalls.push({
+                id: `${textualToolCall.name}-${Date.now()}-${toolCalls.length}`,
+                index: toolCalls.length,
+                type: "function",
+                function: {
+                  name: textualToolCall.name,
+                  arguments: JSON.stringify(textualToolCall.args || {}),
+                },
+              });
+            } else {
+              textContent += part.text;
+            }
+            sawContent = true;
+          }
+        }
+      }
+
+      if (candidate?.finishReason) {
+        finishReason = normalizeOpenAICompatibleFinishReasonString(
+          String(candidate.finishReason).toLowerCase()
+        );
+      }
+
+      if (parsed?.response?.usageMetadata) {
+        const um = parsed.response.usageMetadata;
+        usage = {
+          prompt_tokens: um.promptTokenCount || 0,
+          completion_tokens: um.candidatesTokenCount || 0,
+          total_tokens: um.totalTokenCount || 0,
+        };
+      }
+    } catch {
+      // Ignore malformed lines
+    }
+  }
+
+  if (!sawContent && toolCalls.length === 0) return null;
+
+  const message: Record<string, unknown> = {
+    role: "assistant",
+    content: textContent || null,
+  };
+
+  if (toolCalls.length > 0) {
+    message.tool_calls = toolCalls;
+    finishReason = "tool_calls";
+  }
+
+  const result: Record<string, unknown> = {
+    id: `chatcmpl-${Date.now()}`,
+    object: "chat.completion",
+    created: Math.floor(Date.now() / 1000),
+    model: fallbackModel || "unknown",
+    choices: [
+      {
+        index: 0,
+        message,
+        finish_reason: finishReason,
+      },
+    ],
+  };
+
+  if (usage) {
+    result.usage = usage;
+  }
+
+  return result;
+}
diff --git a/tests/unit/executor-antigravity.test.ts b/tests/unit/executor-antigravity.test.ts
index 99cf20b606f..03b0063af56 100644
--- a/tests/unit/executor-antigravity.test.ts
+++ b/tests/unit/executor-antigravity.test.ts
@@ -1,10 +1,14 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
 
-import { AntigravityExecutor } from "../../open-sse/executors/antigravity.ts";
+import {
+  AntigravityExecutor,
+  createCreditsExtractionTransform,
+} from "../../open-sse/executors/antigravity.ts";
 import { setCliCompatProviders } from "../../open-sse/config/cliFingerprints.ts";
 import { scrubProxyAndFingerprintHeaders } from "../../open-sse/services/antigravityHeaderScrub.ts";
 import { antigravityUserAgent } from "../../open-sse/services/antigravityHeaders.ts";
+import { parseSSEToGeminiResponse } from "../../open-sse/handlers/sseParser.ts";
 import {
   clearAntigravityVersionCache,
   seedAntigravityVersionCache,
@@ -672,7 +676,11 @@ test("AntigravityExecutor.execute auto-retries short 429 responses and collects
       credentials: { accessToken: "token", projectId: "project-1" },
       log: { debug() {}, warn() {} },
     });
-    const payload = (await result.response.json()) as ChatCompletionPayload;
+    // Non-streaming now returns raw SSE; parse it the way chatCore would.
+    const rawSSE = await result.response.text();
+    const parsed = parseSSEToGeminiResponse(rawSSE, "antigravity/gemini-2.5-flash");
+    assert.ok(parsed, "parseSSEToGeminiResponse should parse the SSE");
+    const payload = parsed as ChatCompletionPayload;
 
     assert.equal(calls.length, 2);
     assert.equal(result.response.status, 200);
@@ -939,3 +947,111 @@ test("AntigravityExecutor.transformRequest maps Claude models through Gemini con
   assert.equal(result.request.temperature, undefined);
   assert.equal(result.request.toolConfig, undefined);
 });
+
+// ---------------------------------------------------------------------------
+// createCreditsExtractionTransform -- credits extraction with buffer cap
+// ---------------------------------------------------------------------------
+
+test("createCreditsExtractionTransform extracts remainingCredits from SSE data", async () => {
+  const encoder = new TextEncoder();
+  const sseData = [
+    'data: {"response":{"candidates":[{"content":{"parts":[{"text":"hello"}]},"finishReason":"STOP"}]}}\n\n',
+    'data: {"remainingCredits":[{"creditType":"GOOGLE_ONE_AI","creditAmount":"42"}]}\n\n',
+  ].join("");
+
+  const transform = createCreditsExtractionTransform("test-account");
+  const readable = new ReadableStream({
+    start(controller) {
+      controller.enqueue(encoder.encode(sseData));
+      controller.close();
+    },
+  });
+
+  // Consume the stream through the transform
+  const output = readable.pipeThrough(transform);
+  const reader = output.getReader();
+  const chunks: Uint8Array[] = [];
+  while (true) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    chunks.push(value);
+  }
+
+  // Data must pass through unmodified
+  const collected = new TextDecoder().decode(
+    new Uint8Array(
+      chunks.reduce((acc, c) => acc + c.length, 0) > 0 ? Buffer.concat(chunks) : new Uint8Array(0)
+    )
+  );
+  assert.ok(collected.includes("hello"));
+  assert.ok(collected.includes("remainingCredits"));
+});
+
+test("createCreditsExtractionTransform with buffer cap truncates large buffers", async () => {
+  const encoder = new TextEncoder();
+  // Build a payload larger than 1KB
+  const largeText = "x".repeat(2000);
+  const ssePayload = JSON.stringify({
+    response: {
+      candidates: [{ content: { parts: [{ text: largeText }] } }],
+    },
+  });
+  const sseLine = `data: ${ssePayload}\n\n`;
+  // Append a credits line at the end
+  const creditsLine =
+    'data: {"remainingCredits":[{"creditType":"GOOGLE_ONE_AI","creditAmount":"99"}]}\n\n';
+  const fullData = sseLine + creditsLine;
+
+  // Use a 512-byte buffer cap -- the large text line should be discarded
+  const transform = createCreditsExtractionTransform("test-account", 512);
+  const readable = new ReadableStream({
+    start(controller) {
+      // Send in small chunks to exercise the sliding-window logic
+      const encoded = encoder.encode(fullData);
+      const chunkSize = 256;
+      for (let i = 0; i < encoded.length; i += chunkSize) {
+        controller.enqueue(encoded.slice(i, i + chunkSize));
+      }
+      controller.close();
+    },
+  });
+
+  const output = readable.pipeThrough(transform);
+  const reader = output.getReader();
+  while (true) {
+    const { done } = await reader.read();
+    if (done) break;
+  }
+
+  // The transform should not throw -- buffer cap just limits what the
+  // flush handler can see.  If the credits line was within the last 512
+  // bytes it will be found; otherwise it's a graceful no-op.
+  // Either way, no crash or OOM.
+  assert.ok(true);
+});
+
+test("createCreditsExtractionTransform handles malformed SSE gracefully", async () => {
+  const encoder = new TextEncoder();
+  const badData = "not valid sse\ndata: {broken json\n\ndata: [DONE]\n\n";
+
+  const transform = createCreditsExtractionTransform("test-account");
+  const readable = new ReadableStream({
+    start(controller) {
+      controller.enqueue(encoder.encode(badData));
+      controller.close();
+    },
+  });
+
+  const output = readable.pipeThrough(transform);
+  const reader = output.getReader();
+  const chunks: Uint8Array[] = [];
+  while (true) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    chunks.push(value);
+  }
+
+  // Data passes through unmodified, no crash on malformed input
+  const collected = new TextDecoder().decode(Buffer.concat(chunks));
+  assert.ok(collected.includes("not valid sse"));
+});
diff --git a/tests/unit/sse-parser.test.ts b/tests/unit/sse-parser.test.ts
index 7511834df04..dc7568af5c8 100644
--- a/tests/unit/sse-parser.test.ts
+++ b/tests/unit/sse-parser.test.ts
@@ -1,8 +1,12 @@
 import test from "node:test";
 import assert from "node:assert/strict";
 
-const { parseSSEToOpenAIResponse, parseSSEToClaudeResponse, parseSSEToResponsesOutput } =
-  await import("../../open-sse/handlers/sseParser.ts");
+const {
+  parseSSEToOpenAIResponse,
+  parseSSEToClaudeResponse,
+  parseSSEToResponsesOutput,
+  parseSSEToGeminiResponse,
+} = await import("../../open-sse/handlers/sseParser.ts");
 
 test("parseSSEToOpenAIResponse parses a single SSE event with a done marker", () => {
   const rawSSE = [
@@ -103,12 +107,12 @@ test("parseSSEToClaudeResponse parses text, thinking, tool_use, and usage events
 
   assert.equal(parsed.id, "msg_1");
   assert.equal(parsed.model, "claude-3-5-sonnet");
-  assert.equal((parsed.content[0] as any).type, "thinking");
-  assert.equal((parsed as any).content[0].thinking, "step 1");
-  assert.equal((parsed as any).content[0].signature, "sig-1");
-  (assert as any).equal((parsed.content[1] as any).text, "Hello");
-  assert.equal((parsed.content[2] as any).type, "tool_use");
-  (assert as any).deepEqual((parsed.content[2] as any).input, { q: "docs" });
+  assert.equal((parsed.content[0] as { type: string }).type, "thinking");
+  assert.equal((parsed.content[0] as { thinking: string }).thinking, "step 1");
+  assert.equal((parsed.content[0] as { signature: string }).signature, "sig-1");
+  assert.equal((parsed.content[1] as { text: string }).text, "Hello");
+  assert.equal((parsed.content[2] as { type: string }).type, "tool_use");
+  assert.deepEqual((parsed.content[2] as { input: unknown }).input, { q: "docs" });
   assert.equal(parsed.stop_reason, "tool_use");
   assert.equal(parsed.stop_sequence, "END");
   assert.deepEqual(parsed.usage, { input_tokens: 10, output_tokens: 4 });
@@ -130,7 +134,7 @@ test("parseSSEToClaudeResponse tolerates event-only types and missing blank sepa
 
   assert.equal(parsed.id, "msg_event_fallback");
   assert.equal(parsed.model, "claude-sonnet-4-6");
-  assert.equal((parsed.content[0] as any).text, "event fallback ok");
+  assert.equal((parsed.content[0] as { text: string }).text, "event fallback ok");
   assert.deepEqual(parsed.usage, { input_tokens: 3, output_tokens: 2 });
 });
 
@@ -152,9 +156,9 @@ test("parseSSEToClaudeResponse merges signature_delta into an existing thinking
 
   const parsed = parseSSEToClaudeResponse(rawSSE, "fallback-model");
 
-  assert.equal((parsed.content[0] as any).type, "thinking");
-  assert.equal((parsed.content[0] as any).thinking, "first second");
-  assert.equal((parsed.content[0] as any).signature, "sig-1");
+  assert.equal((parsed.content[0] as { type: string }).type, "thinking");
+  assert.equal((parsed.content[0] as { thinking: string }).thinking, "first second");
+  assert.equal((parsed.content[0] as { signature: string }).signature, "sig-1");
 });
 
 test("parseSSEToClaudeResponse preserves signature_delta when it arrives before thinking_delta", () => {
@@ -172,9 +176,9 @@ test("parseSSEToClaudeResponse preserves signature_delta when it arrives before
 
   const parsed = parseSSEToClaudeResponse(rawSSE, "fallback-model");
 
-  assert.equal((parsed.content[0] as any).type, "thinking");
-  assert.equal((parsed.content[0] as any).thinking, "later thinking");
-  assert.equal((parsed.content[0] as any).signature, "sig-before");
+  assert.equal((parsed.content[0] as { type: string }).type, "thinking");
+  assert.equal((parsed.content[0] as { thinking: string }).thinking, "later thinking");
+  assert.equal((parsed.content[0] as { signature: string }).signature, "sig-before");
 });
 
 test("parseSSEToClaudeResponse ignores malformed payloads and returns null when nothing valid remains", () => {
@@ -334,3 +338,98 @@ test("parseSSEToOpenAIResponse deduplicates repeated tool call snapshots", () =>
   assert.equal(toolCall.function.arguments, args);
   assert.equal(JSON.parse(toolCall.function.arguments).command, "find /tmp -name test.txt");
 });
+
+// ---------------------------------------------------------------------------
+// parseSSEToGeminiResponse
+// ---------------------------------------------------------------------------
+
+test("parseSSEToGeminiResponse extracts text content from candidate parts", () => {
+  const rawSSE = [
+    'data: {"response":{"candidates":[{"content":{"parts":[{"text":"Hello "}]}}]}}',
+    'data: {"response":{"candidates":[{"content":{"parts":[{"text":"world"}]},"finishReason":"STOP"}],"usageMetadata":{"promptTokenCount":5,"candidatesTokenCount":3,"totalTokenCount":8}}}',
+  ].join("\n");
+
+  const parsed = parseSSEToGeminiResponse(rawSSE, "gemini-2.5-flash");
+
+  assert.ok(parsed);
+  assert.equal(parsed.object, "chat.completion");
+  assert.equal(parsed.choices[0].message.content, "Hello world");
+  assert.equal(parsed.choices[0].finish_reason, "stop");
+  assert.deepEqual(parsed.usage, {
+    prompt_tokens: 5,
+    completion_tokens: 3,
+    total_tokens: 8,
+  });
+});
+
+test("parseSSEToGeminiResponse handles markdown shortcut format", () => {
+  const rawSSE = [
+    'data: {"markdown":"Hello "}',
+    'data: {"markdown":"world"}',
+    'data: {"response":{"candidates":[{"finishReason":"STOP"}]}}',
+  ].join("\n");
+
+  const parsed = parseSSEToGeminiResponse(rawSSE, "gemini-2.5-flash");
+
+  assert.ok(parsed);
+  assert.equal(parsed.choices[0].message.content, "Hello world");
+  assert.equal(parsed.choices[0].finish_reason, "stop");
+});
+
+test("parseSSEToGeminiResponse extracts tool calls from textual format", () => {
+  const rawSSE = [
+    `data: ${JSON.stringify({
+      response: {
+        candidates: [
+          {
+            content: {
+              parts: [
+                {
+                  text: '[Tool call: search_files]\nArguments: {"path":"/tmp"}',
+                },
+              ],
+            },
+            finishReason: "STOP",
+          },
+        ],
+      },
+    })}`,
+  ].join("\n");
+
+  const parsed = parseSSEToGeminiResponse(rawSSE, "gemini-3.5-flash-low");
+
+  assert.ok(parsed);
+  assert.equal(parsed.choices[0].finish_reason, "tool_calls");
+  const toolCalls = parsed.choices[0].message.tool_calls;
+  assert.equal(toolCalls.length, 1);
+  assert.equal(toolCalls[0].function.name, "search_files");
+  assert.deepEqual(JSON.parse(toolCalls[0].function.arguments), { path: "/tmp" });
+});
+
+test("parseSSEToGeminiResponse returns null for empty or non-content SSE", () => {
+  assert.equal(parseSSEToGeminiResponse("", "model"), null);
+  assert.equal(parseSSEToGeminiResponse("data: [DONE]\n", "model"), null);
+  assert.equal(parseSSEToGeminiResponse("event: ping\n", "model"), null);
+});
+
+test("parseSSEToGeminiResponse ignores thought/thoughtSignature parts", () => {
+  const rawSSE = [
+    `data: ${JSON.stringify({
+      response: {
+        candidates: [
+          {
+            content: {
+              parts: [{ text: "internal reasoning", thought: true }, { text: "visible answer" }],
+            },
+            finishReason: "STOP",
+          },
+        ],
+      },
+    })}`,
+  ].join("\n");
+
+  const parsed = parseSSEToGeminiResponse(rawSSE, "model");
+
+  assert.ok(parsed);
+  assert.equal(parsed.choices[0].message.content, "visible answer");
+});

From 846aaf9de82881d0cce4b1e1f18844c9d5f3763e Mon Sep 17 00:00:00 2001
From: HouMinXi <19586012+HouMinXi@users.noreply.github.com>
Date: Sat, 18 Jul 2026 12:06:17 -0300
Subject: [PATCH 10/14] refactor(antigravity): extract streaming passthrough to
 module (file-size cap)

#7408 added the non-streaming SSE pass-through (createCreditsExtractionTransform
plus its two call sites: the credits-retry path and the main non-streaming
path) inline in antigravity.ts, growing it to 1806 lines. Combined with two
other authorized PRs touching the same file (#6979 +11, #7290 +30), the
projected total exceeds the frozen file-size gate (1813).

Extract the new streaming-passthrough logic verbatim into
open-sse/executors/antigravity/streamingPassthrough.ts
(createCreditsExtractionTransform + a new buildSsePassthroughResult that
deduplicates the two near-identical call sites), following the existing
sseCollect.ts submodule pattern -- pure, no host state, no fetch/auth.
antigravity.ts keeps a thin wrapper for createCreditsExtractionTransform
(same public signature the existing unit tests import) that injects
updateAntigravityRemainingCredits so the two modules don't import each
other.

No behavior change: same abort handling, same 499-on-early-disconnect,
same 16KB credits sliding-window cap. antigravity.ts: 1806 -> 1693 lines
(under the 1755 pre-PR baseline, with margin). New module: 176 lines
(cap 800).

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---
 open-sse/executors/antigravity.ts             | 185 ++++--------------
 .../antigravity/streamingPassthrough.ts       | 176 +++++++++++++++++
 2 files changed, 212 insertions(+), 149 deletions(-)
 create mode 100644 open-sse/executors/antigravity/streamingPassthrough.ts

diff --git a/open-sse/executors/antigravity.ts b/open-sse/executors/antigravity.ts
index dc82426a33c..bd87c2f7dd0 100644
--- a/open-sse/executors/antigravity.ts
+++ b/open-sse/executors/antigravity.ts
@@ -54,6 +54,10 @@ import {
 } from "./antigravity/sseCollect.ts";
 // processAntigravitySSEPayload re-exported for external importers (tests).
 export { processAntigravitySSEPayload } from "./antigravity/sseCollect.ts";
+import {
+  createCreditsExtractionTransform as createCreditsExtractionTransformImpl,
+  buildSsePassthroughResult,
+} from "./antigravity/streamingPassthrough.ts";
 import {
   applyAntigravityClientProfileHeaders,
   removeHeaderCaseInsensitive,
@@ -120,11 +124,6 @@ type AntigravityChunkContent = Record<string, unknown> & {
   >;
 };
 
-type AntigravityCreditEntry = {
-  creditType?: string;
-  creditAmount?: string;
-};
-
 function getChunkedOrFixedBody(bodyStr: string, stream: boolean): BodyInit {
   if (stream) {
     return new ReadableStream(
@@ -273,87 +272,22 @@ export function updateAntigravityRemainingCredits(accountId: string, balance: nu
 }
 
 /**
- * Create a pass-through TransformStream that extracts `remainingCredits`
- * from SSE data without consuming the stream.  The downstream client
- * receives the unmodified bytes.
- *
- * @param accountId  Provider account ID for credit-balance persistence.
- * @param bufferSize  Optional sliding-window buffer cap in bytes.
- *                   Pass 0 or omit for unlimited (non-streaming callers
- *                   where the full body is already buffered upstream).
- *                   The streaming path uses 16384 (16 KB) to prevent OOM
- *                   on long-lived SSE connections.  Credit-balance data
- *                   appears near the end of the SSE stream (after
- *                   content), so the sliding window captures it even at
- *                   16 KB -- only truly massive responses (>16 KB of
- *                   consecutive non-newline content) would lose credits.
+ * Pass-through TransformStream that extracts `remainingCredits` from SSE
+ * data without consuming the stream (the downstream client receives the
+ * unmodified bytes). Thin wrapper around the pure implementation in
+ * streamingPassthrough.ts, injecting this executor's credit-balance cache
+ * writer so the two modules don't import each other. See that module's
+ * doc comment for the full parameter behavior.
  * @internal Exported for unit testing only.
  */
 export function createCreditsExtractionTransform(
   accountId: string,
   bufferSize = 0
 ): TransformStream<Uint8Array, Uint8Array> {
-  let buffer = "";
-  const decoder = new TextDecoder();
-
-  return new TransformStream(
-    {
-      transform(chunk, controller) {
-        controller.enqueue(chunk);
-        try {
-          buffer += decoder.decode(chunk, { stream: true });
-          // Sliding-window cap: truncate after the last complete newline
-          // in the discard region so SSE lines are never split mid-payload.
-          if (bufferSize > 0 && buffer.length > bufferSize) {
-            const lastNewline = buffer.lastIndexOf("\n", buffer.length - bufferSize);
-            if (lastNewline !== -1) {
-              buffer = buffer.slice(lastNewline + 1);
-            } else {
-              // No newline in the discard region -- incomplete line, discard entirely.
-              buffer = "";
-            }
-          }
-        } catch {
-          /* decoding best-effort */
-        }
-      },
-      flush() {
-        try {
-          buffer += decoder.decode();
-        } catch {
-          /* decoding best-effort */
-        }
-        try {
-          const lines = buffer.split("\n");
-          for (const line of lines) {
-            const trimmed = line.trim();
-            if (!trimmed.startsWith("data:")) continue;
-            const payload = trimmed.slice(5).trim();
-            if (!payload || payload === "[DONE]") continue;
-            try {
-              const parsed = JSON.parse(payload);
-              if (Array.isArray(parsed?.remainingCredits)) {
-                const googleCredit = parsed.remainingCredits.find((c: unknown) => {
-                  const credit = asRecord(c);
-                  return credit?.creditType === "GOOGLE_ONE_AI";
-                }) as AntigravityCreditEntry | undefined;
-                if (googleCredit) {
-                  const balance = parseInt(String(googleCredit.creditAmount ?? ""), 10);
-                  if (!isNaN(balance)) updateAntigravityRemainingCredits(accountId, balance);
-                }
-              }
-            } catch {
-              /* skip malformed lines */
-            }
-          }
-        } catch {
-          /* credits extraction is best-effort */
-        }
-        buffer = "";
-      },
-    },
-    { highWaterMark: 16384 },
-    { highWaterMark: 16384 }
+  return createCreditsExtractionTransformImpl(
+    accountId,
+    updateAntigravityRemainingCredits,
+    bufferSize
   );
 }
 
@@ -1465,41 +1399,19 @@ export class AntigravityExecutor extends BaseExecutor {
                   if (creditsResp.ok || creditsResp.status !== HTTP_STATUS.RATE_LIMITED) {
                     log?.info?.("AG_CREDITS", `Credits retry succeeded: ${creditsResp.status}`);
                     if (!stream && creditsResp.body) {
-                      // Client already disconnected — skip pipe
-                      if (signal?.aborted) {
-                        creditsResp.body.cancel().catch(() => {});
-                        return {
-                          response: new Response(null, { status: 499 }),
-                          url,
-                          headers: finalCreditsHeaders,
-                          transformedBody: null,
-                        };
-                      }
-                      // Return raw SSE with credits extraction
-                      const crTransform = createCreditsExtractionTransform(
+                      // Raw SSE pass-through + credits extraction (see
+                      // streamingPassthrough.ts); 499s early if the client
+                      // already disconnected instead of piping a cancelled body.
+                      return buildSsePassthroughResult(
+                        creditsResp.body,
+                        creditsResp,
                         accountId,
-                        16 * 1024 // 16KB cap
-                      );
-                      const crTappedBody = creditsResp.body.pipeThrough(crTransform);
-                      if (signal) {
-                        signal.addEventListener(
-                          "abort",
-                          () => {
-                            creditsResp.body.cancel().catch(() => {});
-                          },
-                          { once: true }
-                        );
-                      }
-                      return {
-                        response: new Response(crTappedBody, {
-                          status: creditsResp.status,
-                          statusText: creditsResp.statusText,
-                          headers: creditsResp.headers,
-                        }),
+                        updateAntigravityRemainingCredits,
                         url,
-                        headers: finalCreditsHeaders,
-                        transformedBody: attachToolNameMap(creditsBody, requestToolNameMap),
-                      };
+                        finalCreditsHeaders,
+                        attachToolNameMap(creditsBody, requestToolNameMap),
+                        signal
+                      );
                     }
                     return {
                       response: creditsResp,
@@ -1665,44 +1577,19 @@ export class AntigravityExecutor extends BaseExecutor {
           }
 
           if (response.body) {
-            // Client already disconnected — skip pipe
-            if (signal?.aborted) {
-              response.body.cancel().catch(() => {});
-              return {
-                response: new Response(null, { status: 499 }),
-                url,
-                headers: finalHeaders,
-                transformedBody: null,
-              };
-            }
-            // Cancel upstream body on client disconnect
-            if (signal) {
-              signal.addEventListener(
-                "abort",
-                () => {
-                  response.body?.cancel().catch(() => {});
-                },
-                { once: true }
-              );
-            }
-
-            // Tap the stream to extract remainingCredits while passing
-            // data through unmodified.  chatCore drains the full body.
-            const nsPassThrough = createCreditsExtractionTransform(
+            // Raw SSE pass-through + credits extraction (see
+            // streamingPassthrough.ts); 499s early if the client already
+            // disconnected instead of piping a cancelled body.
+            return buildSsePassthroughResult(
+              response.body,
+              response,
               accountId,
-              16 * 1024 // 16KB sliding-window cap to prevent OOM
-            );
-            const tappedBody = response.body.pipeThrough(nsPassThrough);
-            return {
-              response: new Response(tappedBody, {
-                status: response.status,
-                statusText: response.statusText,
-                headers: response.headers,
-              }),
+              updateAntigravityRemainingCredits,
               url,
-              headers: finalHeaders,
-              transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
-            };
+              finalHeaders,
+              attachToolNameMap(transformedBody, requestToolNameMap),
+              signal
+            );
           }
 
           // No body -- return as-is
diff --git a/open-sse/executors/antigravity/streamingPassthrough.ts b/open-sse/executors/antigravity/streamingPassthrough.ts
new file mode 100644
index 00000000000..69282e030c3
--- /dev/null
+++ b/open-sse/executors/antigravity/streamingPassthrough.ts
@@ -0,0 +1,176 @@
+// Pure streaming pass-through helpers for the Antigravity executor (#7408):
+// tap an upstream Gemini SSE Response through a credits-extraction
+// TransformStream instead of buffering the whole body in the executor, so
+// long-thinking models aren't killed by an artificial collection timeout.
+// Extracted from antigravity.ts (no host state, no fetch/auth) -- the
+// credit-balance cache itself stays in antigravity.ts; callers inject the
+// update function below so the two modules don't import each other.
+
+/** Shape of one entry in a Gemini `remainingCredits` SSE payload array. */
+export type AntigravityCreditEntry = {
+  creditType?: string;
+  creditAmount?: string;
+};
+
+function asCreditRecord(value: unknown): Record<string, unknown> | null {
+  return value && typeof value === "object" && !Array.isArray(value)
+    ? (value as Record<string, unknown>)
+    : null;
+}
+
+/**
+ * Create a pass-through TransformStream that extracts `remainingCredits`
+ * from SSE data without consuming the stream.  The downstream client
+ * receives the unmodified bytes.
+ *
+ * @param accountId  Provider account ID for credit-balance persistence.
+ * @param onCreditsUpdate  Invoked with the parsed GOOGLE_ONE_AI balance.
+ *                   Injected by the caller (antigravity.ts's
+ *                   updateAntigravityRemainingCredits) to avoid this module
+ *                   importing back the executor's credit-balance cache.
+ * @param bufferSize  Optional sliding-window buffer cap in bytes.
+ *                   Pass 0 or omit for unlimited (non-streaming callers
+ *                   where the full body is already buffered upstream).
+ *                   The streaming path uses 16384 (16 KB) to prevent OOM
+ *                   on long-lived SSE connections.  Credit-balance data
+ *                   appears near the end of the SSE stream (after
+ *                   content), so the sliding window captures it even at
+ *                   16 KB -- only truly massive responses (>16 KB of
+ *                   consecutive non-newline content) would lose credits.
+ */
+export function createCreditsExtractionTransform(
+  accountId: string,
+  onCreditsUpdate: (accountId: string, balance: number) => void,
+  bufferSize = 0
+): TransformStream<Uint8Array, Uint8Array> {
+  let buffer = "";
+  const decoder = new TextDecoder();
+
+  return new TransformStream(
+    {
+      transform(chunk, controller) {
+        controller.enqueue(chunk);
+        try {
+          buffer += decoder.decode(chunk, { stream: true });
+          // Sliding-window cap: truncate after the last complete newline
+          // in the discard region so SSE lines are never split mid-payload.
+          if (bufferSize > 0 && buffer.length > bufferSize) {
+            const lastNewline = buffer.lastIndexOf("\n", buffer.length - bufferSize);
+            if (lastNewline !== -1) {
+              buffer = buffer.slice(lastNewline + 1);
+            } else {
+              // No newline in the discard region -- incomplete line, discard entirely.
+              buffer = "";
+            }
+          }
+        } catch {
+          /* decoding best-effort */
+        }
+      },
+      flush() {
+        try {
+          buffer += decoder.decode();
+        } catch {
+          /* decoding best-effort */
+        }
+        try {
+          const lines = buffer.split("\n");
+          for (const line of lines) {
+            const trimmed = line.trim();
+            if (!trimmed.startsWith("data:")) continue;
+            const payload = trimmed.slice(5).trim();
+            if (!payload || payload === "[DONE]") continue;
+            try {
+              const parsed = JSON.parse(payload);
+              if (Array.isArray(parsed?.remainingCredits)) {
+                const googleCredit = parsed.remainingCredits.find((c: unknown) => {
+                  const credit = asCreditRecord(c);
+                  return credit?.creditType === "GOOGLE_ONE_AI";
+                }) as AntigravityCreditEntry | undefined;
+                if (googleCredit) {
+                  const balance = parseInt(String(googleCredit.creditAmount ?? ""), 10);
+                  if (!isNaN(balance)) onCreditsUpdate(accountId, balance);
+                }
+              }
+            } catch {
+              /* skip malformed lines */
+            }
+          }
+        } catch {
+          /* credits extraction is best-effort */
+        }
+        buffer = "";
+      },
+    },
+    { highWaterMark: 16384 },
+    { highWaterMark: 16384 }
+  );
+}
+
+/** Result shape returned to callers of AntigravityExecutor.execute(). */
+export type SsePassthroughResult = {
+  response: Response;
+  url: string;
+  headers: Record<string, string>;
+  transformedBody: unknown;
+};
+
+/** Cancel `body` when `signal` aborts, releasing the upstream connection. */
+function cancelBodyOnAbort(body: ReadableStream<Uint8Array>, signal: AbortSignal): void {
+  signal.addEventListener(
+    "abort",
+    () => {
+      body.cancel().catch(() => {});
+    },
+    { once: true }
+  );
+}
+
+/**
+ * Build the non-streaming pass-through result: tap `body` through
+ * createCreditsExtractionTransform and wrap it in a same-status Response so
+ * chatCore's non-streaming path (readNonStreamingResponseBody +
+ * parseNonStreamingSSEPayload) can drain and parse the Gemini SSE without
+ * this executor buffering the whole stream itself.
+ *
+ * If the client already disconnected (`signal.aborted`), cancels the
+ * upstream body immediately and returns a bare 499 instead of piping a
+ * cancelled body through.
+ */
+export function buildSsePassthroughResult(
+  body: ReadableStream<Uint8Array>,
+  upstream: { status: number; statusText: string; headers: Headers },
+  accountId: string,
+  onCreditsUpdate: (accountId: string, balance: number) => void,
+  url: string,
+  outHeaders: Record<string, string>,
+  transformedBody: unknown,
+  signal: AbortSignal | null | undefined
+): SsePassthroughResult {
+  // Client already disconnected — skip pipe
+  if (signal?.aborted) {
+    body.cancel().catch(() => {});
+    return {
+      response: new Response(null, { status: 499 }),
+      url,
+      headers: outHeaders,
+      transformedBody: null,
+    };
+  }
+  // Cancel upstream body on client disconnect
+  if (signal) cancelBodyOnAbort(body, signal);
+
+  const tapped = body.pipeThrough(
+    createCreditsExtractionTransform(accountId, onCreditsUpdate, 16 * 1024)
+  );
+  return {
+    response: new Response(tapped, {
+      status: upstream.status,
+      statusText: upstream.statusText,
+      headers: upstream.headers,
+    }),
+    url,
+    headers: outHeaders,
+    transformedBody,
+  };
+}

From 3180d89b460df148d0176f0b24b821bb9ad8e575 Mon Sep 17 00:00:00 2001
From: HouMinXi <19586012+HouMinXi@users.noreply.github.com>
Date: Sat, 18 Jul 2026 12:19:18 -0300
Subject: [PATCH 11/14] refactor(antigravity): split incremental parser + move
 new tests to own file (file-size caps)

Two remaining frozen file-size violations from #7408, resolved by
extraction/move with zero behavior or assert changes:

- open-sse/handlers/sseParser.ts (979 > frozen 830): the PR appended
  parseSSEToGeminiResponse (+153, the Gemini buffered-SSE ->
  chat.completion parser). Moved verbatim to
  open-sse/handlers/sseParser/geminiResponse.ts, following the handlers
  submodule pattern (chatCore/, responseSanitizer/). sseParser.ts is now
  byte-identical to its pre-PR content (825 lines; PR delta 0). Importers
  (chatCore/nonStreamingSse.ts, tests) point at the new module.

- tests/unit/executor-antigravity.test.ts (1058 > testFrozen 942): the
  PR's new streaming-passthrough tests moved verbatim (same tests, same
  asserts) to tests/unit/antigravity-streaming-passthrough.test.ts:
  the 3 createCreditsExtractionTransform tests plus the non-streaming
  passthrough drain test ("auto-retries short 429 ... collects SSE for
  non-stream clients"), which the PR rewired onto the new raw-SSE path.
  The frozen file drops to 888 lines (below its pre-PR 941).

New files: geminiResponse.ts 156 lines, passthrough test 202 lines (caps
800). Also fixes the stale sseParser.ts path in collectStreamToResponse's
deprecation note.

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---
 open-sse/executors/antigravity.ts             |   4 +-
 open-sse/handlers/chatCore/nonStreamingSse.ts |   2 +-
 open-sse/handlers/sseParser.ts                | 153 -------------
 open-sse/handlers/sseParser/geminiResponse.ts | 156 ++++++++++++++
 .../antigravity-streaming-passthrough.test.ts | 202 ++++++++++++++++++
 tests/unit/executor-antigravity.test.ts       | 177 +--------------
 tests/unit/sse-parser.test.ts                 |  10 +-
 7 files changed, 369 insertions(+), 335 deletions(-)
 create mode 100644 open-sse/handlers/sseParser/geminiResponse.ts
 create mode 100644 tests/unit/antigravity-streaming-passthrough.test.ts

diff --git a/open-sse/executors/antigravity.ts b/open-sse/executors/antigravity.ts
index bd87c2f7dd0..283cd6689c0 100644
--- a/open-sse/executors/antigravity.ts
+++ b/open-sse/executors/antigravity.ts
@@ -952,8 +952,8 @@ export class AntigravityExecutor extends BaseExecutor {
    * OpenAI-format chat.completion payload.
    *
    * @deprecated Use the non-streaming SSE path in chatCore instead, which calls
-   * parseSSEToGeminiResponse() from sseParser.ts.  This method is retained only
-   * for backward compatibility and may be removed in a future release.
+   * parseSSEToGeminiResponse() from sseParser/geminiResponse.ts.  This method is
+   * retained only for backward compatibility and may be removed in a future release.
    */
   collectStreamToResponse(
     response: Response,
diff --git a/open-sse/handlers/chatCore/nonStreamingSse.ts b/open-sse/handlers/chatCore/nonStreamingSse.ts
index 8c2b748b978..743ba81a918 100644
--- a/open-sse/handlers/chatCore/nonStreamingSse.ts
+++ b/open-sse/handlers/chatCore/nonStreamingSse.ts
@@ -3,8 +3,8 @@ import {
   parseSSEToResponsesOutput,
   parseSSEToClaudeResponse,
   parseSSEToOpenAIResponse,
-  parseSSEToGeminiResponse,
 } from "../sseParser.ts";
+import { parseSSEToGeminiResponse } from "../sseParser/geminiResponse.ts";
 import { getHeaderValueCaseInsensitive } from "./headers.ts";
 
 export function parseNonStreamingSSEPayload(
diff --git a/open-sse/handlers/sseParser.ts b/open-sse/handlers/sseParser.ts
index b9dc848059e..d2e634e12ae 100644
--- a/open-sse/handlers/sseParser.ts
+++ b/open-sse/handlers/sseParser.ts
@@ -1,6 +1,5 @@
 import { appendToolCallArgumentDelta } from "../utils/toolCallArguments.ts";
 import { sanitizeErrorMessage } from "../utils/error.ts";
-import { normalizeOpenAICompatibleFinishReasonString } from "../utils/finishReason.ts";
 
 /**
  * Extract a provider error message from a buffered SSE stream that carries an
@@ -824,155 +823,3 @@ export function parseSSEToResponsesOutput(rawSSE, fallbackModel) {
     metadata: picked.metadata || {},
   };
 }
-
-/**
- * Convert Gemini/Antigravity SSE chunks into a single non-streaming OpenAI
- * chat.completion JSON response.  Gemini SSE carries payloads like:
- *
- *   data: {"markdown":"...chunk..."}
- *   data: {"response":{"candidates":[{"content":{"parts":[{"text":"..."}]},"finishReason":"STOP"}],"usageMetadata":{...}}}
- *   data: {"remainingCredits":[...]}
- *
- * Reuses the same parsing logic as processAntigravitySSEPayload() in sseCollect.ts
- * so that format conversion is functionally equivalent to the previous
- * collectStreamToResponse() approach.  Intentional differences:
- *   - remainingCredits is NOT embedded into the result (handled separately
- *     by the credits-extraction TransformStream in antigravity.ts).
- *   - The synthetic `id` uses `chatcmpl-${Date.now()}` (no UUID suffix)
- *     because this path runs once per response, not per chunk.
- */
-export function parseSSEToGeminiResponse(
-  rawSSE: string,
-  fallbackModel: string
-): Record<string, unknown> | null {
-  const lines = String(rawSSE || "").split("\n");
-  let textContent = "";
-  let finishReason = "stop";
-  let usage: Record<string, unknown> | null = null;
-  let sawContent = false;
-
-  type AccumulatedToolCall = {
-    id: string;
-    index: number;
-    type: "function";
-    function: { name: string; arguments: string };
-  };
-  const toolCalls: AccumulatedToolCall[] = [];
-
-  const stripZeroWidth = (value: unknown): unknown => {
-    if (typeof value === "string") return value.replace(/[\u200B-\u200D\uFEFF]/g, "");
-    return value;
-  };
-
-  const tryParseTextualToolCall = (text: string): { name: string; args: unknown } | null => {
-    const normalized = text.replace(/[\u200B-\u200D\uFEFF]/g, "");
-    const match = normalized.match(
-      /^[\s\S]*?\[Tool call:\s*([^\]\n]+)\]\s*\nArguments:\s*([\s\S]+?)\s*$/
-    );
-    if (!match) return null;
-    const name = match[1]?.trim();
-    const rawArgs = match[2]?.trim();
-    if (!name || !rawArgs) return null;
-    try {
-      return { name, args: stripZeroWidth(JSON.parse(rawArgs)) };
-    } catch {
-      return null;
-    }
-  };
-
-  for (const line of lines) {
-    const trimmed = line.trim();
-    if (!trimmed.startsWith("data:")) continue;
-    const payload = trimmed.slice(5).trim();
-    if (!payload || payload === "[DONE]") continue;
-
-    try {
-      const parsed = JSON.parse(payload);
-
-      // Markdown shortcut (some Gemini variants)
-      const markdown =
-        typeof parsed?.markdown === "string"
-          ? parsed.markdown
-          : typeof parsed?.response?.markdown === "string"
-            ? parsed.response.markdown
-            : null;
-      if (markdown) {
-        textContent += markdown;
-        sawContent = true;
-      }
-
-      // Candidate content parts
-      const candidate = parsed?.response?.candidates?.[0];
-      if (candidate?.content?.parts) {
-        for (const part of candidate.content.parts) {
-          if (typeof part.text === "string" && !part.thought && !part.thoughtSignature) {
-            const textualToolCall = tryParseTextualToolCall(part.text);
-            if (textualToolCall) {
-              toolCalls.push({
-                id: `${textualToolCall.name}-${Date.now()}-${toolCalls.length}`,
-                index: toolCalls.length,
-                type: "function",
-                function: {
-                  name: textualToolCall.name,
-                  arguments: JSON.stringify(textualToolCall.args || {}),
-                },
-              });
-            } else {
-              textContent += part.text;
-            }
-            sawContent = true;
-          }
-        }
-      }
-
-      if (candidate?.finishReason) {
-        finishReason = normalizeOpenAICompatibleFinishReasonString(
-          String(candidate.finishReason).toLowerCase()
-        );
-      }
-
-      if (parsed?.response?.usageMetadata) {
-        const um = parsed.response.usageMetadata;
-        usage = {
-          prompt_tokens: um.promptTokenCount || 0,
-          completion_tokens: um.candidatesTokenCount || 0,
-          total_tokens: um.totalTokenCount || 0,
-        };
-      }
-    } catch {
-      // Ignore malformed lines
-    }
-  }
-
-  if (!sawContent && toolCalls.length === 0) return null;
-
-  const message: Record<string, unknown> = {
-    role: "assistant",
-    content: textContent || null,
-  };
-
-  if (toolCalls.length > 0) {
-    message.tool_calls = toolCalls;
-    finishReason = "tool_calls";
-  }
-
-  const result: Record<string, unknown> = {
-    id: `chatcmpl-${Date.now()}`,
-    object: "chat.completion",
-    created: Math.floor(Date.now() / 1000),
-    model: fallbackModel || "unknown",
-    choices: [
-      {
-        index: 0,
-        message,
-        finish_reason: finishReason,
-      },
-    ],
-  };
-
-  if (usage) {
-    result.usage = usage;
-  }
-
-  return result;
-}
diff --git a/open-sse/handlers/sseParser/geminiResponse.ts b/open-sse/handlers/sseParser/geminiResponse.ts
new file mode 100644
index 00000000000..d0801019bd2
--- /dev/null
+++ b/open-sse/handlers/sseParser/geminiResponse.ts
@@ -0,0 +1,156 @@
+// Gemini/Antigravity buffered-SSE -> chat.completion conversion (#7408).
+// Extracted verbatim from sseParser.ts (file-size cap): pure parsing, no host
+// state, following the handlers submodule pattern (chatCore/, responseSanitizer/).
+import { normalizeOpenAICompatibleFinishReasonString } from "../../utils/finishReason.ts";
+
+/**
+ * Convert Gemini/Antigravity SSE chunks into a single non-streaming OpenAI
+ * chat.completion JSON response.  Gemini SSE carries payloads like:
+ *
+ *   data: {"markdown":"...chunk..."}
+ *   data: {"response":{"candidates":[{"content":{"parts":[{"text":"..."}]},"finishReason":"STOP"}],"usageMetadata":{...}}}
+ *   data: {"remainingCredits":[...]}
+ *
+ * Reuses the same parsing logic as processAntigravitySSEPayload() in sseCollect.ts
+ * so that format conversion is functionally equivalent to the previous
+ * collectStreamToResponse() approach.  Intentional differences:
+ *   - remainingCredits is NOT embedded into the result (handled separately
+ *     by the credits-extraction TransformStream in antigravity.ts).
+ *   - The synthetic `id` uses `chatcmpl-${Date.now()}` (no UUID suffix)
+ *     because this path runs once per response, not per chunk.
+ */
+export function parseSSEToGeminiResponse(
+  rawSSE: string,
+  fallbackModel: string
+): Record<string, unknown> | null {
+  const lines = String(rawSSE || "").split("\n");
+  let textContent = "";
+  let finishReason = "stop";
+  let usage: Record<string, unknown> | null = null;
+  let sawContent = false;
+
+  type AccumulatedToolCall = {
+    id: string;
+    index: number;
+    type: "function";
+    function: { name: string; arguments: string };
+  };
+  const toolCalls: AccumulatedToolCall[] = [];
+
+  const stripZeroWidth = (value: unknown): unknown => {
+    if (typeof value === "string") return value.replace(/[\u200B-\u200D\uFEFF]/g, "");
+    return value;
+  };
+
+  const tryParseTextualToolCall = (text: string): { name: string; args: unknown } | null => {
+    const normalized = text.replace(/[\u200B-\u200D\uFEFF]/g, "");
+    const match = normalized.match(
+      /^[\s\S]*?\[Tool call:\s*([^\]\n]+)\]\s*\nArguments:\s*([\s\S]+?)\s*$/
+    );
+    if (!match) return null;
+    const name = match[1]?.trim();
+    const rawArgs = match[2]?.trim();
+    if (!name || !rawArgs) return null;
+    try {
+      return { name, args: stripZeroWidth(JSON.parse(rawArgs)) };
+    } catch {
+      return null;
+    }
+  };
+
+  for (const line of lines) {
+    const trimmed = line.trim();
+    if (!trimmed.startsWith("data:")) continue;
+    const payload = trimmed.slice(5).trim();
+    if (!payload || payload === "[DONE]") continue;
+
+    try {
+      const parsed = JSON.parse(payload);
+
+      // Markdown shortcut (some Gemini variants)
+      const markdown =
+        typeof parsed?.markdown === "string"
+          ? parsed.markdown
+          : typeof parsed?.response?.markdown === "string"
+            ? parsed.response.markdown
+            : null;
+      if (markdown) {
+        textContent += markdown;
+        sawContent = true;
+      }
+
+      // Candidate content parts
+      const candidate = parsed?.response?.candidates?.[0];
+      if (candidate?.content?.parts) {
+        for (const part of candidate.content.parts) {
+          if (typeof part.text === "string" && !part.thought && !part.thoughtSignature) {
+            const textualToolCall = tryParseTextualToolCall(part.text);
+            if (textualToolCall) {
+              toolCalls.push({
+                id: `${textualToolCall.name}-${Date.now()}-${toolCalls.length}`,
+                index: toolCalls.length,
+                type: "function",
+                function: {
+                  name: textualToolCall.name,
+                  arguments: JSON.stringify(textualToolCall.args || {}),
+                },
+              });
+            } else {
+              textContent += part.text;
+            }
+            sawContent = true;
+          }
+        }
+      }
+
+      if (candidate?.finishReason) {
+        finishReason = normalizeOpenAICompatibleFinishReasonString(
+          String(candidate.finishReason).toLowerCase()
+        );
+      }
+
+      if (parsed?.response?.usageMetadata) {
+        const um = parsed.response.usageMetadata;
+        usage = {
+          prompt_tokens: um.promptTokenCount || 0,
+          completion_tokens: um.candidatesTokenCount || 0,
+          total_tokens: um.totalTokenCount || 0,
+        };
+      }
+    } catch {
+      // Ignore malformed lines
+    }
+  }
+
+  if (!sawContent && toolCalls.length === 0) return null;
+
+  const message: Record<string, unknown> = {
+    role: "assistant",
+    content: textContent || null,
+  };
+
+  if (toolCalls.length > 0) {
+    message.tool_calls = toolCalls;
+    finishReason = "tool_calls";
+  }
+
+  const result: Record<string, unknown> = {
+    id: `chatcmpl-${Date.now()}`,
+    object: "chat.completion",
+    created: Math.floor(Date.now() / 1000),
+    model: fallbackModel || "unknown",
+    choices: [
+      {
+        index: 0,
+        message,
+        finish_reason: finishReason,
+      },
+    ],
+  };
+
+  if (usage) {
+    result.usage = usage;
+  }
+
+  return result;
+}
diff --git a/tests/unit/antigravity-streaming-passthrough.test.ts b/tests/unit/antigravity-streaming-passthrough.test.ts
new file mode 100644
index 00000000000..23715cbae30
--- /dev/null
+++ b/tests/unit/antigravity-streaming-passthrough.test.ts
@@ -0,0 +1,202 @@
+// Antigravity streaming-passthrough behavior (#7408): the non-streaming drain
+// path (raw SSE + chatCore-side parse) and the credits-extraction pass-through
+// TransformStream. Moved verbatim from tests/unit/executor-antigravity.test.ts
+// (frozen file-size cap) — same tests, same asserts.
+import { test } from "node:test";
+import assert from "node:assert/strict";
+
+import {
+  AntigravityExecutor,
+  createCreditsExtractionTransform,
+} from "../../open-sse/executors/antigravity.ts";
+import { parseSSEToGeminiResponse } from "../../open-sse/handlers/sseParser/geminiResponse.ts";
+import {
+  clearAntigravityVersionCache,
+  seedAntigravityVersionCache,
+} from "../../open-sse/services/antigravityVersion.ts";
+
+type ChatCompletionPayload = {
+  object?: string;
+  choices: Array<{
+    message: { content: string };
+    finish_reason: string;
+  }>;
+  usage?: {
+    prompt_tokens: number;
+    completion_tokens: number;
+    total_tokens: number;
+  };
+};
+
+test.afterEach(() => {
+  clearAntigravityVersionCache();
+});
+
+test("AntigravityExecutor.execute auto-retries short 429 responses and collects SSE for non-stream clients", async () => {
+  const executor = new AntigravityExecutor();
+  const originalFetch = globalThis.fetch;
+  const originalSetTimeout = globalThis.setTimeout;
+  const calls = [];
+  seedAntigravityVersionCache("2026.04.17-test");
+
+  globalThis.fetch = async (url) => {
+    calls.push(String(url));
+
+    if (calls.length === 1) {
+      return new Response(JSON.stringify({ error: { message: "rate limited" } }), {
+        status: 429,
+        headers: { "Content-Type": "application/json" },
+      });
+    }
+
+    return new Response(
+      [
+        'data: {"response":{"candidates":[{"content":{"parts":[{"text":"Hello "}]},"finishReason":"STOP"}]}}\n\n',
+        'data: {"response":{"candidates":[{"content":{"parts":[{"text":"again"}]},"finishReason":"STOP"}],"usageMetadata":{"promptTokenCount":2,"candidatesTokenCount":3,"totalTokenCount":5}}}\n\n',
+      ].join(""),
+      {
+        status: 200,
+        headers: { "Content-Type": "text/event-stream" },
+      }
+    );
+  };
+  globalThis.setTimeout = ((callback) => {
+    (callback as () => void)();
+    return 0;
+  }) as typeof setTimeout;
+
+  try {
+    const result = await executor.execute({
+      model: "antigravity/gemini-2.5-flash",
+      body: { request: { contents: [] } },
+      stream: false,
+      credentials: { accessToken: "token", projectId: "project-1" },
+      log: { debug() {}, warn() {} },
+    });
+    // Non-streaming now returns raw SSE; parse it the way chatCore would.
+    const rawSSE = await result.response.text();
+    const parsed = parseSSEToGeminiResponse(rawSSE, "antigravity/gemini-2.5-flash");
+    assert.ok(parsed, "parseSSEToGeminiResponse should parse the SSE");
+    const payload = parsed as ChatCompletionPayload;
+
+    assert.equal(calls.length, 2);
+    assert.equal(result.response.status, 200);
+    assert.equal(payload.choices[0].message.content, "Hello again");
+    assert.deepEqual(payload.usage, {
+      prompt_tokens: 2,
+      completion_tokens: 3,
+      total_tokens: 5,
+    });
+  } finally {
+    globalThis.fetch = originalFetch;
+    globalThis.setTimeout = originalSetTimeout;
+  }
+});
+
+// ---------------------------------------------------------------------------
+// createCreditsExtractionTransform -- credits extraction with buffer cap
+// ---------------------------------------------------------------------------
+
+test("createCreditsExtractionTransform extracts remainingCredits from SSE data", async () => {
+  const encoder = new TextEncoder();
+  const sseData = [
+    'data: {"response":{"candidates":[{"content":{"parts":[{"text":"hello"}]},"finishReason":"STOP"}]}}\n\n',
+    'data: {"remainingCredits":[{"creditType":"GOOGLE_ONE_AI","creditAmount":"42"}]}\n\n',
+  ].join("");
+
+  const transform = createCreditsExtractionTransform("test-account");
+  const readable = new ReadableStream({
+    start(controller) {
+      controller.enqueue(encoder.encode(sseData));
+      controller.close();
+    },
+  });
+
+  // Consume the stream through the transform
+  const output = readable.pipeThrough(transform);
+  const reader = output.getReader();
+  const chunks: Uint8Array[] = [];
+  while (true) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    chunks.push(value);
+  }
+
+  // Data must pass through unmodified
+  const collected = new TextDecoder().decode(
+    new Uint8Array(
+      chunks.reduce((acc, c) => acc + c.length, 0) > 0 ? Buffer.concat(chunks) : new Uint8Array(0)
+    )
+  );
+  assert.ok(collected.includes("hello"));
+  assert.ok(collected.includes("remainingCredits"));
+});
+
+test("createCreditsExtractionTransform with buffer cap truncates large buffers", async () => {
+  const encoder = new TextEncoder();
+  // Build a payload larger than 1KB
+  const largeText = "x".repeat(2000);
+  const ssePayload = JSON.stringify({
+    response: {
+      candidates: [{ content: { parts: [{ text: largeText }] } }],
+    },
+  });
+  const sseLine = `data: ${ssePayload}\n\n`;
+  // Append a credits line at the end
+  const creditsLine =
+    'data: {"remainingCredits":[{"creditType":"GOOGLE_ONE_AI","creditAmount":"99"}]}\n\n';
+  const fullData = sseLine + creditsLine;
+
+  // Use a 512-byte buffer cap -- the large text line should be discarded
+  const transform = createCreditsExtractionTransform("test-account", 512);
+  const readable = new ReadableStream({
+    start(controller) {
+      // Send in small chunks to exercise the sliding-window logic
+      const encoded = encoder.encode(fullData);
+      const chunkSize = 256;
+      for (let i = 0; i < encoded.length; i += chunkSize) {
+        controller.enqueue(encoded.slice(i, i + chunkSize));
+      }
+      controller.close();
+    },
+  });
+
+  const output = readable.pipeThrough(transform);
+  const reader = output.getReader();
+  while (true) {
+    const { done } = await reader.read();
+    if (done) break;
+  }
+
+  // The transform should not throw -- buffer cap just limits what the
+  // flush handler can see.  If the credits line was within the last 512
+  // bytes it will be found; otherwise it's a graceful no-op.
+  // Either way, no crash or OOM.
+  assert.ok(true);
+});
+
+test("createCreditsExtractionTransform handles malformed SSE gracefully", async () => {
+  const encoder = new TextEncoder();
+  const badData = "not valid sse\ndata: {broken json\n\ndata: [DONE]\n\n";
+
+  const transform = createCreditsExtractionTransform("test-account");
+  const readable = new ReadableStream({
+    start(controller) {
+      controller.enqueue(encoder.encode(badData));
+      controller.close();
+    },
+  });
+
+  const output = readable.pipeThrough(transform);
+  const reader = output.getReader();
+  const chunks: Uint8Array[] = [];
+  while (true) {
+    const { done, value } = await reader.read();
+    if (done) break;
+    chunks.push(value);
+  }
+
+  // Data passes through unmodified, no crash on malformed input
+  const collected = new TextDecoder().decode(Buffer.concat(chunks));
+  assert.ok(collected.includes("not valid sse"));
+});
diff --git a/tests/unit/executor-antigravity.test.ts b/tests/unit/executor-antigravity.test.ts
index 03b0063af56..d92db9fb8a0 100644
--- a/tests/unit/executor-antigravity.test.ts
+++ b/tests/unit/executor-antigravity.test.ts
@@ -1,14 +1,10 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
 
-import {
-  AntigravityExecutor,
-  createCreditsExtractionTransform,
-} from "../../open-sse/executors/antigravity.ts";
+import { AntigravityExecutor } from "../../open-sse/executors/antigravity.ts";
 import { setCliCompatProviders } from "../../open-sse/config/cliFingerprints.ts";
 import { scrubProxyAndFingerprintHeaders } from "../../open-sse/services/antigravityHeaderScrub.ts";
 import { antigravityUserAgent } from "../../open-sse/services/antigravityHeaders.ts";
-import { parseSSEToGeminiResponse } from "../../open-sse/handlers/sseParser.ts";
 import {
   clearAntigravityVersionCache,
   seedAntigravityVersionCache,
@@ -635,66 +631,9 @@ test("AntigravityExecutor.refreshCredentials refreshes Google OAuth tokens", asy
   }
 });
 
-test("AntigravityExecutor.execute auto-retries short 429 responses and collects SSE for non-stream clients", async () => {
-  const executor = new AntigravityExecutor();
-  const originalFetch = globalThis.fetch;
-  const originalSetTimeout = globalThis.setTimeout;
-  const calls = [];
-  seedAntigravityVersionCache("2026.04.17-test");
-
-  globalThis.fetch = async (url) => {
-    calls.push(String(url));
-
-    if (calls.length === 1) {
-      return new Response(JSON.stringify({ error: { message: "rate limited" } }), {
-        status: 429,
-        headers: { "Content-Type": "application/json" },
-      });
-    }
-
-    return new Response(
-      [
-        'data: {"response":{"candidates":[{"content":{"parts":[{"text":"Hello "}]},"finishReason":"STOP"}]}}\n\n',
-        'data: {"response":{"candidates":[{"content":{"parts":[{"text":"again"}]},"finishReason":"STOP"}],"usageMetadata":{"promptTokenCount":2,"candidatesTokenCount":3,"totalTokenCount":5}}}\n\n',
-      ].join(""),
-      {
-        status: 200,
-        headers: { "Content-Type": "text/event-stream" },
-      }
-    );
-  };
-  globalThis.setTimeout = ((callback) => {
-    (callback as () => void)();
-    return 0;
-  }) as typeof setTimeout;
-
-  try {
-    const result = await executor.execute({
-      model: "antigravity/gemini-2.5-flash",
-      body: { request: { contents: [] } },
-      stream: false,
-      credentials: { accessToken: "token", projectId: "project-1" },
-      log: { debug() {}, warn() {} },
-    });
-    // Non-streaming now returns raw SSE; parse it the way chatCore would.
-    const rawSSE = await result.response.text();
-    const parsed = parseSSEToGeminiResponse(rawSSE, "antigravity/gemini-2.5-flash");
-    assert.ok(parsed, "parseSSEToGeminiResponse should parse the SSE");
-    const payload = parsed as ChatCompletionPayload;
-
-    assert.equal(calls.length, 2);
-    assert.equal(result.response.status, 200);
-    assert.equal(payload.choices[0].message.content, "Hello again");
-    assert.deepEqual(payload.usage, {
-      prompt_tokens: 2,
-      completion_tokens: 3,
-      total_tokens: 5,
-    });
-  } finally {
-    globalThis.fetch = originalFetch;
-    globalThis.setTimeout = originalSetTimeout;
-  }
-});
+// The non-streaming passthrough drain test ("auto-retries short 429 responses and
+// collects SSE for non-stream clients") lives in
+// tests/unit/antigravity-streaming-passthrough.test.ts with the other passthrough tests.
 
 test("AntigravityExecutor.execute embeds retryAfterMs when the upstream asks for a long wait", async () => {
   const executor = new AntigravityExecutor();
@@ -947,111 +886,3 @@ test("AntigravityExecutor.transformRequest maps Claude models through Gemini con
   assert.equal(result.request.temperature, undefined);
   assert.equal(result.request.toolConfig, undefined);
 });
-
-// ---------------------------------------------------------------------------
-// createCreditsExtractionTransform -- credits extraction with buffer cap
-// ---------------------------------------------------------------------------
-
-test("createCreditsExtractionTransform extracts remainingCredits from SSE data", async () => {
-  const encoder = new TextEncoder();
-  const sseData = [
-    'data: {"response":{"candidates":[{"content":{"parts":[{"text":"hello"}]},"finishReason":"STOP"}]}}\n\n',
-    'data: {"remainingCredits":[{"creditType":"GOOGLE_ONE_AI","creditAmount":"42"}]}\n\n',
-  ].join("");
-
-  const transform = createCreditsExtractionTransform("test-account");
-  const readable = new ReadableStream({
-    start(controller) {
-      controller.enqueue(encoder.encode(sseData));
-      controller.close();
-    },
-  });
-
-  // Consume the stream through the transform
-  const output = readable.pipeThrough(transform);
-  const reader = output.getReader();
-  const chunks: Uint8Array[] = [];
-  while (true) {
-    const { done, value } = await reader.read();
-    if (done) break;
-    chunks.push(value);
-  }
-
-  // Data must pass through unmodified
-  const collected = new TextDecoder().decode(
-    new Uint8Array(
-      chunks.reduce((acc, c) => acc + c.length, 0) > 0 ? Buffer.concat(chunks) : new Uint8Array(0)
-    )
-  );
-  assert.ok(collected.includes("hello"));
-  assert.ok(collected.includes("remainingCredits"));
-});
-
-test("createCreditsExtractionTransform with buffer cap truncates large buffers", async () => {
-  const encoder = new TextEncoder();
-  // Build a payload larger than 1KB
-  const largeText = "x".repeat(2000);
-  const ssePayload = JSON.stringify({
-    response: {
-      candidates: [{ content: { parts: [{ text: largeText }] } }],
-    },
-  });
-  const sseLine = `data: ${ssePayload}\n\n`;
-  // Append a credits line at the end
-  const creditsLine =
-    'data: {"remainingCredits":[{"creditType":"GOOGLE_ONE_AI","creditAmount":"99"}]}\n\n';
-  const fullData = sseLine + creditsLine;
-
-  // Use a 512-byte buffer cap -- the large text line should be discarded
-  const transform = createCreditsExtractionTransform("test-account", 512);
-  const readable = new ReadableStream({
-    start(controller) {
-      // Send in small chunks to exercise the sliding-window logic
-      const encoded = encoder.encode(fullData);
-      const chunkSize = 256;
-      for (let i = 0; i < encoded.length; i += chunkSize) {
-        controller.enqueue(encoded.slice(i, i + chunkSize));
-      }
-      controller.close();
-    },
-  });
-
-  const output = readable.pipeThrough(transform);
-  const reader = output.getReader();
-  while (true) {
-    const { done } = await reader.read();
-    if (done) break;
-  }
-
-  // The transform should not throw -- buffer cap just limits what the
-  // flush handler can see.  If the credits line was within the last 512
-  // bytes it will be found; otherwise it's a graceful no-op.
-  // Either way, no crash or OOM.
-  assert.ok(true);
-});
-
-test("createCreditsExtractionTransform handles malformed SSE gracefully", async () => {
-  const encoder = new TextEncoder();
-  const badData = "not valid sse\ndata: {broken json\n\ndata: [DONE]\n\n";
-
-  const transform = createCreditsExtractionTransform("test-account");
-  const readable = new ReadableStream({
-    start(controller) {
-      controller.enqueue(encoder.encode(badData));
-      controller.close();
-    },
-  });
-
-  const output = readable.pipeThrough(transform);
-  const reader = output.getReader();
-  const chunks: Uint8Array[] = [];
-  while (true) {
-    const { done, value } = await reader.read();
-    if (done) break;
-    chunks.push(value);
-  }
-
-  // Data passes through unmodified, no crash on malformed input
-  const collected = new TextDecoder().decode(Buffer.concat(chunks));
-  assert.ok(collected.includes("not valid sse"));
-});
diff --git a/tests/unit/sse-parser.test.ts b/tests/unit/sse-parser.test.ts
index dc7568af5c8..36fcd7dedac 100644
--- a/tests/unit/sse-parser.test.ts
+++ b/tests/unit/sse-parser.test.ts
@@ -1,12 +1,10 @@
 import test from "node:test";
 import assert from "node:assert/strict";
 
-const {
-  parseSSEToOpenAIResponse,
-  parseSSEToClaudeResponse,
-  parseSSEToResponsesOutput,
-  parseSSEToGeminiResponse,
-} = await import("../../open-sse/handlers/sseParser.ts");
+const { parseSSEToOpenAIResponse, parseSSEToClaudeResponse, parseSSEToResponsesOutput } =
+  await import("../../open-sse/handlers/sseParser.ts");
+const { parseSSEToGeminiResponse } =
+  await import("../../open-sse/handlers/sseParser/geminiResponse.ts");
 
 test("parseSSEToOpenAIResponse parses a single SSE event with a done marker", () => {
   const rawSSE = [

From a9574d8e60210f345a6788dd6e30fb946354095d Mon Sep 17 00:00:00 2001
From: HouMinXi <19586012+HouMinXi@users.noreply.github.com>
Date: Sat, 18 Jul 2026 14:24:16 -0300
Subject: [PATCH 12/14] refactor(antigravity): decompose execute + gemini
 parser below complexity gate
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

executeOnce() (complexity 127, 436 lines) and parseSSEToGeminiResponse()
(complexity 39, 117 lines) were both over the check-complexity.mjs gate
(complexity>15, max-lines-per-function>80). Decomposed each into small
named helpers, no behavior change:

- geminiResponse.ts: split into pure per-concern functions (markdown
  shortcut, candidate-parts walk, finishReason, usageMetadata, final
  response assembly).
- antigravity.ts: extracted the per-url-index attempt pipeline
  (runAntigravityAttempt, handleAntigravityRateLimit,
  tryResolveRetryFromErrorBody, shouldAutoRetryTransient) and moved the
  request/result-building helpers (send, credits-retry, embed-retry,
  non-streaming/streaming result builders) into a new
  antigravity/executeAttempt.ts submodule, mirroring the existing
  streamingPassthrough.ts/sseCollect.ts pattern. Also fixes the
  antigravity.ts file-size cap (was pushed to 2084 lines > 1813 frozen
  ceiling by the decomposition itself; now 1428).

check-complexity.mjs: 2054 violations (baseline 2058) — net improvement.
execute/executeOnce/parseSSEToGeminiResponse no longer appear with
ruleId complexity or max-lines-per-function.

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---
 open-sse/executors/antigravity.ts             | 961 +++++++-----------
 .../executors/antigravity/executeAttempt.ts   | 687 +++++++++++++
 open-sse/handlers/sseParser/geminiResponse.ts | 290 +++---
 3 files changed, 1209 insertions(+), 729 deletions(-)
 create mode 100644 open-sse/executors/antigravity/executeAttempt.ts

diff --git a/open-sse/executors/antigravity.ts b/open-sse/executors/antigravity.ts
index 283cd6689c0..e3fc66b378b 100644
--- a/open-sse/executors/antigravity.ts
+++ b/open-sse/executors/antigravity.ts
@@ -1,21 +1,16 @@
 import crypto, { randomUUID } from "crypto";
 import {
   BaseExecutor,
-  mergeAbortSignals,
   mergeUpstreamExtraHeaders,
   type ExecuteInput,
   type ExecutorLog,
   type ProviderCredentials,
 } from "./base.ts";
-import { applyFingerprint, isCliCompatEnabled } from "../config/cliFingerprints.ts";
-import { buildAntigravityUpstreamError } from "./antigravityUpstreamError.ts";
 import {
   PROVIDERS,
   OAUTH_ENDPOINTS,
   HTTP_STATUS,
   FETCH_TIMEOUT_MS,
-  STREAM_READINESS_TIMEOUT_MS,
-  ANTIGRAVITY_PRE_RESPONSE_TIMEOUT_CODE,
 } from "../config/constants.ts";
 import { scrubProxyAndFingerprintHeaders } from "../services/antigravityHeaderScrub.ts";
 import {
@@ -24,7 +19,6 @@ import {
 } from "../services/antigravityHeaders.ts";
 import { classify429, decide429, type Decision } from "../services/antigravity429Engine.ts";
 import {
-  injectCreditsField,
   shouldRetryWithCredits,
   shouldUseCreditsFirst,
   getCreditsMode,
@@ -40,7 +34,6 @@ import {
   resolveAntigravityModelId,
   getAntigravityModelFallbacks,
 } from "../config/antigravityModelAliases.ts";
-import { cloakAntigravityToolPayload } from "../config/toolCloaking.ts";
 import {
   shouldStripCloudCodeThinking,
   stripCloudCodeThinkingConfig,
@@ -56,25 +49,33 @@ import {
 export { processAntigravitySSEPayload } from "./antigravity/sseCollect.ts";
 import {
   createCreditsExtractionTransform as createCreditsExtractionTransformImpl,
-  buildSsePassthroughResult,
+  type SsePassthroughResult,
 } from "./antigravity/streamingPassthrough.ts";
 import {
-  applyAntigravityClientProfileHeaders,
-  removeHeaderCaseInsensitive,
-} from "../services/antigravityClientProfile.ts";
+  toSafeAntigravityLog,
+  finalizeAntigravityRequestBody,
+  sendAntigravityRequest,
+  tryCreditsRetry,
+  tryEmbedLongRetryAfter,
+  buildFinalAntigravityResult,
+  buildAntigravity429ErrorMessage,
+  markCreditsExhausted,
+  type SafeAntigravityLog,
+} from "./antigravity/executeAttempt.ts";
 import {
   generateAntigravityRequestId,
   getAntigravityEnvelopeUserAgent,
   getAntigravitySessionId,
 } from "../services/antigravityIdentity.ts";
-import * as prl from "../utils/providerRequestLogging.ts";
 
 const MAX_RETRY_AFTER_MS = 60_000;
 const LONG_RETRY_THRESHOLD_MS = 60_000;
-const CREDITS_EXHAUSTED_TTL_MS = 5 * 60 * 60 * 1000; // 5 hours
 // Cap for transient 5xx backoff — shorter than the 429 cap to avoid long stalls on
 // infra hiccups ("Agent execution terminated", "high traffic", capacity errors).
 const ANTIGRAVITY_TRANSIENT_RETRY_MAX_MS = 15_000;
+// Bounded per-URL auto-retry count for both the Retry-After-driven short retry and
+// the no-Retry-After transient/429 backoff loop in executeOnce().
+const MAX_AUTO_RETRIES = 3;
 
 const ANTIGRAVITY_TRANSIENT_ERROR_PATTERNS: RegExp[] = [
   /high\s+traffic/i,
@@ -106,7 +107,7 @@ interface AntigravityContent {
   [key: string]: unknown;
 }
 
-type AntigravityCredentials = ProviderCredentials & {
+export type AntigravityCredentials = ProviderCredentials & {
   projectId?: string | null;
   expiresIn?: number;
 };
@@ -124,46 +125,6 @@ type AntigravityChunkContent = Record<string, unknown> & {
   >;
 };
 
-function getChunkedOrFixedBody(bodyStr: string, stream: boolean): BodyInit {
-  if (stream) {
-    return new ReadableStream(
-      {
-        async start(controller) {
-          controller.enqueue(new TextEncoder().encode(bodyStr));
-          controller.close();
-        },
-      },
-      { highWaterMark: 16384 }
-    );
-  }
-  return bodyStr;
-}
-
-function cloneAntigravityRequestBody(body: unknown): unknown {
-  if (!body || typeof body !== "object") {
-    return body;
-  }
-
-  try {
-    return structuredClone(body);
-  } catch {
-    return JSON.parse(JSON.stringify(body));
-  }
-}
-
-function serializeAntigravityRequest(
-  provider: string,
-  headers: Record<string, string>,
-  body: unknown
-): { headers: Record<string, string>; bodyString: string } {
-  const serializedBody = cloneAntigravityRequestBody(body);
-
-  if (!isCliCompatEnabled(provider)) {
-    return { headers, bodyString: JSON.stringify(serializedBody) };
-  }
-  return applyFingerprint(provider, { ...headers }, serializedBody);
-}
-
 type AntigravityRequestEnvelope = Record<string, unknown> & {
   project: string;
   model?: string;
@@ -174,44 +135,6 @@ type AntigravityRequestEnvelope = Record<string, unknown> & {
   enabledCreditTypes?: string[];
 };
 
-class AntigravityPreResponseTimeoutError extends Error {
-  code = ANTIGRAVITY_PRE_RESPONSE_TIMEOUT_CODE;
-  status = HTTP_STATUS.GATEWAY_TIMEOUT;
-
-  constructor(timeoutMs: number, url: string) {
-    super(`Antigravity upstream did not return response headers within ${timeoutMs}ms: ${url}`);
-    this.name = "TimeoutError";
-  }
-}
-
-function getAbortErrorCode(error: unknown): string | null {
-  if (!error || typeof error !== "object") return null;
-  const value = (error as { code?: unknown }).code;
-  return typeof value === "string" ? value : null;
-}
-
-function isAntigravityPreResponseTimeout(error: unknown): boolean {
-  return getAbortErrorCode(error) === ANTIGRAVITY_PRE_RESPONSE_TIMEOUT_CODE;
-}
-
-/**
- * Per-account GOOGLE_ONE_AI credits-exhausted tracker.
- * Key: accountId (OAuth subject / email). Value: expiry timestamp.
- * When credits hit 0 we skip the credit retry for CREDITS_EXHAUSTED_TTL_MS.
- */
-const MAX_CREDITS_EXHAUSTED_ENTRIES = 50;
-const creditsExhaustedUntil = new Map<string, number>();
-
-const _creditsExhaustedSweep = setInterval(() => {
-  const now = Date.now();
-  for (const [key, until] of creditsExhaustedUntil) {
-    if (now >= until) creditsExhaustedUntil.delete(key);
-  }
-}, 60_000);
-if (typeof _creditsExhaustedSweep === "object" && "unref" in _creditsExhaustedSweep) {
-  (_creditsExhaustedSweep as { unref?: () => void }).unref?.();
-}
-
 const MAX_CREDIT_BALANCE_ENTRIES = 50;
 const CREDIT_BALANCE_TTL_MS = 5 * 60 * 1000;
 const creditBalanceCache = new Map<string, { balance: number; updatedAt: number }>();
@@ -291,35 +214,6 @@ export function createCreditsExtractionTransform(
   );
 }
 
-function isCreditsExhausted(accountId: string): boolean {
-  const until = creditsExhaustedUntil.get(accountId);
-  if (!until) return false;
-  if (Date.now() >= until) {
-    creditsExhaustedUntil.delete(accountId);
-    return false;
-  }
-  return true;
-}
-
-function markCreditsExhausted(accountId: string): void {
-  if (
-    creditsExhaustedUntil.size >= MAX_CREDITS_EXHAUSTED_ENTRIES &&
-    !creditsExhaustedUntil.has(accountId)
-  ) {
-    const now = Date.now();
-    for (const [key, until] of creditsExhaustedUntil) {
-      if (now >= until) {
-        creditsExhaustedUntil.delete(key);
-      }
-    }
-    if (creditsExhaustedUntil.size >= MAX_CREDITS_EXHAUSTED_ENTRIES) {
-      const oldestKey = creditsExhaustedUntil.keys().next().value;
-      if (oldestKey !== undefined) creditsExhaustedUntil.delete(oldestKey);
-    }
-  }
-  creditsExhaustedUntil.set(accountId, Date.now() + CREDITS_EXHAUSTED_TTL_MS);
-}
-
 /**
  * Persist a quota-exhausted cooldown to the DB for `connectionId` so that
  * cross-request and post-restart routing skips this connection until the
@@ -382,26 +276,6 @@ async function cleanModelName(model: string, modelIdOverride?: string): Promise<
   return clean;
 }
 
-function attachToolNameMap<T>(payload: T, toolNameMap: Map<string, string> | null): T {
-  if (!toolNameMap?.size || !payload || typeof payload !== "object") {
-    return payload;
-  }
-
-  const copy = Array.isArray(payload) ? ([...payload] as T) : ({ ...(payload as object) } as T);
-  Object.defineProperty(copy, "_toolNameMap", {
-    value: toolNameMap,
-    enumerable: false,
-    configurable: true,
-    writable: true,
-  });
-  return copy;
-}
-
-function getRequestTargetModel(body: Record<string, unknown>): string {
-  const target = body.model;
-  return typeof target === "string" && target.length > 0 ? target : "unknown";
-}
-
 /**
  * Hard ceiling on `generationConfig.maxOutputTokens` for Antigravity Cloud Code.
  *
@@ -551,6 +425,49 @@ function stripTrailingAntigravityAssistantTurn(
 // Test-only export so the unit suite can exercise the strip logic directly.
 export const __test_stripTrailingAntigravityAssistantTurn = stripTrailingAntigravityAssistantTurn;
 
+/** Base per-url-index attempt context, before the request has been sent. */
+type AntigravityAttemptContext = {
+  url: string;
+  model: string;
+  /** Pre-serialization headers (built by buildHeaders + mergeUpstreamExtraHeaders) — the
+   * credits-retry re-serializes from these, NOT from `finalHeaders` (already fingerprinted). */
+  headers: Record<string, string>;
+  transformedBody: Record<string, unknown>;
+  requestToolNameMap: Map<string, string> | null;
+  credentials: AntigravityCredentials;
+  stream: boolean;
+  signal: AbortSignal | null | undefined;
+  log: SafeAntigravityLog;
+  accountId: string;
+  creditsMode: ReturnType<typeof getCreditsMode>;
+  urlIndex: number;
+  retryAttemptsByUrl: Record<number, number>;
+  fallbackCount: number;
+};
+
+/** Context threaded through the 429/503 handling helpers — adds the sent response. */
+type AntigravityRateLimitContext = AntigravityAttemptContext & {
+  response: Response;
+  finalHeaders: Record<string, string>;
+};
+
+/**
+ * Outcome of handling a 429/503 response — tells executeOnce()'s loop what to do next.
+ * `lastStatus` mirrors the original inline code, which only updated the outer
+ * `lastStatus` variable when NOT retrying the same url (i.e. on retryNextUrl/fallthrough,
+ * never on the bounded-short-retry or transient-auto-retry same-url paths).
+ */
+type AntigravityRateLimitOutcome =
+  | { action: "return"; result: SsePassthroughResult }
+  | { action: "retrySameUrl" }
+  | { action: "retryNextUrl"; lastStatus: number }
+  | { action: "fallthrough"; retryMs: number | null; lastStatus: number };
+
+/** Outcome of one full per-url attempt in executeOnce() — return a result, or retry. */
+type AntigravityAttemptOutcome =
+  | { action: "return"; result: SsePassthroughResult }
+  | { action: "retry"; sameUrl: boolean; lastStatus?: number };
+
 export class AntigravityExecutor extends BaseExecutor {
   constructor() {
     super("antigravity", PROVIDERS.antigravity);
@@ -1142,9 +1059,9 @@ export class AntigravityExecutor extends BaseExecutor {
   ) {
     await resolveAntigravityVersion();
     const fallbackCount = this.getFallbackCount();
+    const l = toSafeAntigravityLog(log);
     let lastError = null;
     let lastStatus = 0;
-    const MAX_AUTO_RETRIES = 3;
     const retryAttemptsByUrl: Record<number, number> = {}; // Track retry attempts per URL
 
     // Always stream upstream — buildUrl always returns the streaming endpoint.
@@ -1164,44 +1081,6 @@ export class AntigravityExecutor extends BaseExecutor {
     const creditsMode = getCreditsMode();
     const useCreditsFirst = shouldUseCreditsFirst(credentials?.accessToken || "", creditsMode);
 
-    const fetchWithReadinessTimeout = async (
-      url: string,
-      init: RequestInit,
-      timeoutMs = STREAM_READINESS_TIMEOUT_MS
-    ): Promise<Response> => {
-      const boundedTimeoutMs = Math.max(0, Math.floor(timeoutMs));
-      if (boundedTimeoutMs <= 0) {
-        return fetch(url, init);
-      }
-
-      const timeoutController = new AbortController();
-      let timeoutId: ReturnType<typeof setTimeout> | null = setTimeout(() => {
-        timeoutController.abort(new AntigravityPreResponseTimeoutError(boundedTimeoutMs, url));
-      }, boundedTimeoutMs);
-
-      const existingSignal = init.signal instanceof AbortSignal ? init.signal : null;
-      const combinedSignal = existingSignal
-        ? mergeAbortSignals(existingSignal, timeoutController.signal)
-        : timeoutController.signal;
-
-      try {
-        return await fetch(url, { ...init, signal: combinedSignal });
-      } catch (error) {
-        if (
-          timeoutController.signal.aborted &&
-          isAntigravityPreResponseTimeout(timeoutController.signal.reason)
-        ) {
-          throw timeoutController.signal.reason;
-        }
-        throw error;
-      } finally {
-        if (timeoutId) {
-          clearTimeout(timeoutId);
-          timeoutId = null;
-        }
-      }
-    };
-
     for (let urlIndex = 0; urlIndex < fallbackCount; urlIndex++) {
       const url = this.buildUrl(model, upstreamStream, urlIndex);
       const headers = this.buildHeaders(credentials, upstreamStream);
@@ -1213,27 +1092,16 @@ export class AntigravityExecutor extends BaseExecutor {
         credentials,
         modelIdOverride
       );
-      let requestToolNameMap: Map<string, string> | null = null;
 
       if (transformed instanceof Response) {
         return { response: transformed, url, headers, transformedBody: body };
       }
 
-      let transformedBody: Record<string, unknown> = transformed;
-
-      if (transformedBody && typeof transformedBody === "object") {
-        const cloaked = cloakAntigravityToolPayload(transformedBody);
-        transformedBody = cloaked.body;
-        requestToolNameMap = cloaked.toolNameMap;
-      }
-
-      // Credits-first: inject GOOGLE_ONE_AI upfront so we never try the normal
-      // quota path. If credits are exhausted / disabled shouldUseCreditsFirst()
-      // returns false and we fall back to the legacy retry-on-429 flow.
-      if (useCreditsFirst) {
-        transformedBody = injectCreditsField(transformedBody);
-        log?.debug?.("AG_CREDITS", "Credits-first enabled (ANTIGRAVITY_CREDITS=always)");
-      }
+      const { transformedBody, requestToolNameMap } = finalizeAntigravityRequestBody(
+        transformed,
+        useCreditsFirst,
+        l
+      );
 
       // Initialize retry counter for this URL
       if (!retryAttemptsByUrl[urlIndex]) {
@@ -1241,452 +1109,319 @@ export class AntigravityExecutor extends BaseExecutor {
       }
 
       try {
-        const serializedRequest = serializeAntigravityRequest(
-          this.provider,
+        const outcome = await this.runAntigravityAttempt({
+          url,
+          model,
           headers,
-          transformedBody
-        );
-        let finalHeaders = serializedRequest.headers;
-        const capture = (h: Record<string, string>, s: string) =>
-          prl.captureCurrentProviderBody(url, h, s, log);
-        const clientProfile = applyAntigravityClientProfileHeaders(
-          finalHeaders,
+          transformedBody,
+          requestToolNameMap,
           credentials,
-          transformedBody
-        );
+          stream,
+          signal,
+          log: l,
+          accountId,
+          creditsMode,
+          urlIndex,
+          retryAttemptsByUrl,
+          fallbackCount,
+        });
 
-        log?.debug?.(
+        if (outcome.action === "return") return outcome.result;
+        if (outcome.lastStatus !== undefined) lastStatus = outcome.lastStatus;
+        if (outcome.sameUrl) urlIndex--;
+        continue;
+      } catch (error) {
+        lastError = error;
+        l.error(
           "TELEMETRY",
-          `[Antigravity] Execute - URL: ${url}, Model: ${model}, Target: ${getRequestTargetModel(transformedBody)}, RetryAttempt: ${retryAttemptsByUrl[urlIndex]}`
+          `[Antigravity] Network/Fetch Error - URL: ${url}, Model: ${model}, Error: ${error instanceof Error ? error.message : String(error)}`
         );
-
-        // Dump outgoing headers (mask Authorization) and envelope shape for debugging
-        if (log?.debug) {
-          const safeHeaders = { ...finalHeaders };
-          if (safeHeaders["Authorization"]) safeHeaders["Authorization"] = "Bearer ***";
-          log.debug("AG_REQUEST_HEADERS", JSON.stringify(safeHeaders));
-
-          const envelope = transformedBody as Record<string, unknown>;
-          const requestInner = envelope.request as Record<string, unknown> | undefined;
-          log.debug(
-            "AG_REQUEST_ENVELOPE",
-            JSON.stringify({
-              fieldOrder: Object.keys(envelope),
-              project: envelope.project,
-              requestId: envelope.requestId,
-              model: envelope.model,
-              userAgent: envelope.userAgent,
-              requestType: envelope.requestType,
-              enabledCreditTypes: envelope.enabledCreditTypes,
-              clientProfile,
-              sessionId: requestInner?.sessionId,
-              generationConfig: requestInner?.generationConfig,
-            })
-          );
+        if (urlIndex + 1 < fallbackCount) {
+          l.debug("RETRY", `Error on ${url}, trying fallback ${urlIndex + 1}`);
+          continue;
         }
+        throw error;
+      }
+    }
 
-        await capture(finalHeaders, serializedRequest.bodyString);
-        let response = await fetchWithReadinessTimeout(url, {
-          method: "POST",
-          headers: finalHeaders,
-          body: getChunkedOrFixedBody(serializedRequest.bodyString, stream),
-          ...(stream ? { duplex: "half" } : {}),
-          signal,
-        });
+    throw lastError || new Error(`All ${fallbackCount} URLs failed with status ${lastStatus}`);
+  }
 
-        if (response.status === HTTP_STATUS.FORBIDDEN && finalHeaders["x-goog-user-project"]) {
-          const retryHeaders = { ...finalHeaders };
-          removeHeaderCaseInsensitive(retryHeaders, "x-goog-user-project");
-          log?.debug?.("RETRY", "403 with x-goog-user-project, retrying once without it");
-          await capture(retryHeaders, serializedRequest.bodyString);
-          response = await fetchWithReadinessTimeout(url, {
-            method: "POST",
-            headers: retryHeaders,
-            body: getChunkedOrFixedBody(serializedRequest.bodyString, stream),
-            ...(stream ? { duplex: "half" } : {}),
-            signal,
-          });
-          finalHeaders = retryHeaders;
-        }
+  /**
+   * Run one full per-url-index attempt: send the request, handle a 429/503 (retry
+   * same/next url, or a Google One AI credits retry), fall back on other retryable
+   * statuses, optionally embed a long Retry-After, then build the final non-streaming
+   * or streaming result. Returns a result to hand back from execute(), or a retry
+   * instruction for executeOnce()'s loop to act on (continue, optionally urlIndex--).
+   */
+  private async runAntigravityAttempt(
+    ctx: AntigravityAttemptContext
+  ): Promise<AntigravityAttemptOutcome> {
+    const {
+      url,
+      model,
+      headers,
+      transformedBody,
+      requestToolNameMap,
+      credentials,
+      stream,
+      signal,
+      log,
+      accountId,
+      urlIndex,
+      retryAttemptsByUrl,
+      fallbackCount,
+    } = ctx;
+
+    const { response, finalHeaders } = await sendAntigravityRequest(
+      this.provider,
+      url,
+      model,
+      headers,
+      transformedBody,
+      credentials,
+      stream,
+      signal,
+      log,
+      retryAttemptsByUrl[urlIndex]
+    );
 
-        if (!response.ok) {
-          log?.warn?.(
-            "TELEMETRY",
-            `[Antigravity] Error Response - URL: ${url}, Status: ${response.status}, Model: ${model}`
-          );
-        }
+    let retryMs: number | null = null;
 
-        // Parse retry time for 429/503 responses
-        let retryMs: number | null = null;
-
-        if (
-          response.status === HTTP_STATUS.RATE_LIMITED ||
-          response.status === HTTP_STATUS.SERVICE_UNAVAILABLE
-        ) {
-          // Try to get retry time from headers first
-          retryMs = this.parseRetryHeaders(response.headers);
-
-          // If no retry time in headers, try to parse from error message body
-          if (!retryMs) {
-            try {
-              const errorBody = await response.clone().text();
-              const errorJson = JSON.parse(errorBody);
-              let errorMessage = errorJson?.error?.message || errorJson?.message || "";
-              if (errorJson?.error?.details && Array.isArray(errorJson.error.details)) {
-                for (const detail of errorJson.error.details) {
-                  if (detail?.reason) {
-                    errorMessage += ` ${detail.reason}`;
-                  }
-                }
-              }
-
-              // 1. Try to parse explicit retry time from message
-              const parsedRetryMs = this.parseRetryFromErrorMessage(errorMessage);
-
-              // 2. Classify 429 (pass header-parsed retry hint as fallback
-              //    signal — multi-hour Retry-After upgrades rate_limited to
-              //    quota_exhausted so the GOOGLE_ONE_AI credits retry fires).
-              const effectiveRetryHintMs = retryMs ?? parsedRetryMs ?? null;
-              const category = classify429(errorMessage);
-
-              // 3. Decide final retry time BEFORE the credits retry so that
-              //    full_quota_exhausted can skip the credits attempt entirely
-              //    (avoids ~41s hold on an already-exhausted account) and
-              //    persist the cooldown to DB for post-restart routing.
-              const decision: Decision = decide429(category, parsedRetryMs);
-              retryMs = decision.retryAfterMs;
-              log?.debug?.(
-                "AG_429",
-                `Category: ${category}, Decision: ${decision.kind} — ${decision.reason}`
-              );
-
-              if (decision.kind === "full_quota_exhausted" && retryMs) {
-                markConnectionQuotaExhausted(accountId, retryMs);
-              }
-
-              const creditsAlreadyInjected =
-                (transformedBody as { enabledCreditTypes?: unknown }).enabledCreditTypes != null;
-
-              if (category === "quota_exhausted" && creditsAlreadyInjected) {
-                handleCreditsFailure(credentials?.accessToken || "");
-                log?.warn?.("AG_CREDITS", "Credits-first request 429'd — credits likely exhausted");
-                markCreditsExhausted(accountId);
-              }
-
-              if (
-                category === "quota_exhausted" &&
-                decision.kind !== "full_quota_exhausted" &&
-                !creditsAlreadyInjected &&
-                shouldRetryWithCredits(credentials?.accessToken || "", creditsMode !== "off")
-              ) {
-                log?.info?.("AG_CREDITS", "Retrying with Google One AI credits");
-                const creditsBody = injectCreditsField(transformedBody);
-                const serializedCreditsRequest = serializeAntigravityRequest(
-                  this.provider,
-                  headers,
-                  creditsBody
-                );
-                const finalCreditsHeaders = serializedCreditsRequest.headers;
-                try {
-                  await capture(finalCreditsHeaders, serializedCreditsRequest.bodyString);
-                  const creditsResp = await fetchWithReadinessTimeout(url, {
-                    method: "POST",
-                    headers: finalCreditsHeaders,
-                    body: getChunkedOrFixedBody(serializedCreditsRequest.bodyString, stream),
-                    ...(stream ? { duplex: "half" } : {}),
-                    signal,
-                  });
-                  if (creditsResp.ok || creditsResp.status !== HTTP_STATUS.RATE_LIMITED) {
-                    log?.info?.("AG_CREDITS", `Credits retry succeeded: ${creditsResp.status}`);
-                    if (!stream && creditsResp.body) {
-                      // Raw SSE pass-through + credits extraction (see
-                      // streamingPassthrough.ts); 499s early if the client
-                      // already disconnected instead of piping a cancelled body.
-                      return buildSsePassthroughResult(
-                        creditsResp.body,
-                        creditsResp,
-                        accountId,
-                        updateAntigravityRemainingCredits,
-                        url,
-                        finalCreditsHeaders,
-                        attachToolNameMap(creditsBody, requestToolNameMap),
-                        signal
-                      );
-                    }
-                    return {
-                      response: creditsResp,
-                      url,
-                      headers: finalCreditsHeaders,
-                      transformedBody: attachToolNameMap(creditsBody, requestToolNameMap),
-                    };
-                  }
+    if (
+      response.status === HTTP_STATUS.RATE_LIMITED ||
+      response.status === HTTP_STATUS.SERVICE_UNAVAILABLE
+    ) {
+      const rateLimitOutcome = await this.handleAntigravityRateLimit({
+        ...ctx,
+        response,
+        finalHeaders,
+      });
 
-                  // Credit retry also 429'd
-                  handleCreditsFailure(credentials?.accessToken || "");
-                  log?.warn?.("AG_CREDITS", "Credits retry also 429'd");
-
-                  // Also mark in our legacy exhaustion map to avoid retrying other routes
-                  markCreditsExhausted(accountId);
-                } catch (creditsErr) {
-                  handleCreditsFailure(credentials?.accessToken || "");
-                  log?.warn?.("AG_CREDITS", `Credits retry failed: ${creditsErr}`);
-                }
-              }
-            } catch (e) {
-              // Ignore parse errors, will fall back to exponential backoff
-            }
-          }
+      if (rateLimitOutcome.action === "return") {
+        return { action: "return", result: rateLimitOutcome.result };
+      }
+      if (rateLimitOutcome.action === "retrySameUrl") return { action: "retry", sameUrl: true };
+      if (rateLimitOutcome.action === "retryNextUrl") {
+        return { action: "retry", sameUrl: false, lastStatus: rateLimitOutcome.lastStatus };
+      }
+      // Only "fallthrough" remains: last url, no more retries — proceed below with
+      // the resolved retryMs so a long Retry-After can still be embedded in the body.
+      retryMs = rateLimitOutcome.retryMs;
+    }
 
-          // Bounded short-retry: a non-null retryAfterMs ≤ 60s covers nearly every
-          // 429 (decide429 returns 2s/5s/60s defaults), so this branch MUST share the
-          // per-URL attempt counter. Without the bound a persistent 429 loops forever
-          // on the same endpoint/account (urlIndex-- cancels the loop's urlIndex++) and
-          // never returns the 429 to the account-fallback layer in chat.ts.
-          if (
-            retryMs &&
-            retryMs <= LONG_RETRY_THRESHOLD_MS &&
-            retryAttemptsByUrl[urlIndex] < MAX_AUTO_RETRIES
-          ) {
-            retryAttemptsByUrl[urlIndex]++;
-            const effectiveRetryMs = Math.min(retryMs, MAX_RETRY_AFTER_MS);
-            log?.debug?.(
-              "RETRY",
-              `${response.status} retry ${retryAttemptsByUrl[urlIndex]}/${MAX_AUTO_RETRIES} with Retry-After: ${Math.ceil(effectiveRetryMs / 1000)}s, waiting...`
-            );
-            await new Promise((resolve) => setTimeout(resolve, effectiveRetryMs));
-            urlIndex--;
-            continue;
-          }
+    if (this.shouldRetry(response.status, urlIndex)) {
+      log.debug("RETRY", `${response.status} on ${url}, trying fallback ${urlIndex + 1}`);
+      return { action: "retry", sameUrl: false, lastStatus: response.status };
+    }
 
-          // Auto retry for 429 (no Retry-After) or transient 5xx errors.
-          // For 5xx we read the body to detect known transient patterns
-          // ("Agent execution terminated due to error", "high traffic", "capacity").
-          if ((!retryMs || retryMs === 0) && retryAttemptsByUrl[urlIndex] < MAX_AUTO_RETRIES) {
-            let shouldAutoRetry = response.status === HTTP_STATUS.RATE_LIMITED;
-            if (!shouldAutoRetry && ANTIGRAVITY_TRANSIENT_STATUSES.has(response.status)) {
-              try {
-                const errBody = await response.clone().text();
-                let errJson: unknown = null;
-                try {
-                  errJson = errBody ? JSON.parse(errBody) : null;
-                } catch {
-                  // non-JSON body — fall through to pattern match against raw text
-                }
-                const errMsg = this.extractErrorMessage(errJson, errBody);
-                shouldAutoRetry = this.isTransientAntigravityError(response.status, errMsg);
-              } catch {
-                // ignore body read errors
-              }
-            }
-            if (shouldAutoRetry) {
-              retryAttemptsByUrl[urlIndex]++;
-              // Exponential backoff: 2s, 4s, 8s… capped per-status
-              const cap =
-                response.status === HTTP_STATUS.RATE_LIMITED
-                  ? MAX_RETRY_AFTER_MS
-                  : ANTIGRAVITY_TRANSIENT_RETRY_MAX_MS;
-              const backoffMs = Math.min(1000 * 2 ** retryAttemptsByUrl[urlIndex], cap);
-              log?.debug?.(
-                "RETRY",
-                `${response.status} transient auto retry ${retryAttemptsByUrl[urlIndex]}/${MAX_AUTO_RETRIES} after ${backoffMs / 1000}s`
-              );
-              await new Promise((resolve) => setTimeout(resolve, backoffMs));
-              urlIndex--;
-              continue;
-            }
-          }
+    // If we have a 429 with long retry time, embed it in response body
+    const embedded = await tryEmbedLongRetryAfter(
+      response,
+      retryMs,
+      url,
+      finalHeaders,
+      transformedBody,
+      requestToolNameMap,
+      log
+    );
+    if (embedded) return { action: "return", result: embedded };
+
+    const result = await buildFinalAntigravityResult(
+      stream,
+      response,
+      url,
+      finalHeaders,
+      transformedBody,
+      requestToolNameMap,
+      accountId,
+      signal,
+      updateAntigravityRemainingCredits
+    );
+    return { action: "return", result };
+  }
 
-          log?.debug?.(
-            "RETRY",
-            `${response.status}, Retry-After ${retryMs ? `too long (${Math.ceil(retryMs / 1000)}s)` : "missing"}, trying fallback`
-          );
-          lastStatus = response.status;
+  /**
+   * Handle a 429/503 response for one URL-index attempt: resolve the retry-after
+   * time (headers, then error-body classification + Google-One-AI credits retry),
+   * then decide whether to retry the SAME url, fall back to the NEXT url, or (on
+   * the last url with no more retries left) fall through with the resolved retryMs
+   * so the caller can still embed a long Retry-After in the final response body.
+   */
+  private async handleAntigravityRateLimit(
+    ctx: AntigravityRateLimitContext
+  ): Promise<AntigravityRateLimitOutcome> {
+    const { response, log, urlIndex, retryAttemptsByUrl, fallbackCount } = ctx;
+
+    // Try to get retry time from headers first
+    let retryMs: number | null = this.parseRetryHeaders(response.headers);
+
+    // If no retry time in headers, try to parse from error message body
+    if (!retryMs) {
+      const resolved = await this.tryResolveRetryFromErrorBody(ctx);
+      if (resolved.kind === "return") return { action: "return", result: resolved.result };
+      retryMs = resolved.retryMs;
+    }
 
-          if (urlIndex + 1 < fallbackCount) {
-            continue;
-          }
-        }
+    // Bounded short-retry: a non-null retryAfterMs ≤ 60s covers nearly every
+    // 429 (decide429 returns 2s/5s/60s defaults), so this branch MUST share the
+    // per-URL attempt counter. Without the bound a persistent 429 loops forever
+    // on the same endpoint/account (urlIndex-- cancels the loop's urlIndex++) and
+    // never returns the 429 to the account-fallback layer in chat.ts.
+    if (
+      retryMs &&
+      retryMs <= LONG_RETRY_THRESHOLD_MS &&
+      retryAttemptsByUrl[urlIndex] < MAX_AUTO_RETRIES
+    ) {
+      retryAttemptsByUrl[urlIndex]++;
+      const effectiveRetryMs = Math.min(retryMs, MAX_RETRY_AFTER_MS);
+      log.debug(
+        "RETRY",
+        `${response.status} retry ${retryAttemptsByUrl[urlIndex]}/${MAX_AUTO_RETRIES} with Retry-After: ${Math.ceil(effectiveRetryMs / 1000)}s, waiting...`
+      );
+      await new Promise((resolve) => setTimeout(resolve, effectiveRetryMs));
+      return { action: "retrySameUrl" };
+    }
 
-        if (this.shouldRetry(response.status, urlIndex)) {
-          log?.debug?.("RETRY", `${response.status} on ${url}, trying fallback ${urlIndex + 1}`);
-          lastStatus = response.status;
-          continue;
-        }
+    // Auto retry for 429 (no Retry-After) or transient 5xx errors.
+    // For 5xx we read the body to detect known transient patterns
+    // ("Agent execution terminated due to error", "high traffic", "capacity").
+    if ((!retryMs || retryMs === 0) && retryAttemptsByUrl[urlIndex] < MAX_AUTO_RETRIES) {
+      const shouldAutoRetry = await this.shouldAutoRetryTransient(response);
+      if (shouldAutoRetry) {
+        retryAttemptsByUrl[urlIndex]++;
+        // Exponential backoff: 2s, 4s, 8s… capped per-status
+        const cap =
+          response.status === HTTP_STATUS.RATE_LIMITED
+            ? MAX_RETRY_AFTER_MS
+            : ANTIGRAVITY_TRANSIENT_RETRY_MAX_MS;
+        const backoffMs = Math.min(1000 * 2 ** retryAttemptsByUrl[urlIndex], cap);
+        log.debug(
+          "RETRY",
+          `${response.status} transient auto retry ${retryAttemptsByUrl[urlIndex]}/${MAX_AUTO_RETRIES} after ${backoffMs / 1000}s`
+        );
+        await new Promise((resolve) => setTimeout(resolve, backoffMs));
+        return { action: "retrySameUrl" };
+      }
+    }
 
-        // If we have a 429 with long retry time, embed it in response body
-        if (
-          response.status === HTTP_STATUS.RATE_LIMITED &&
-          retryMs &&
-          retryMs > LONG_RETRY_THRESHOLD_MS
-        ) {
-          try {
-            const respBody = await response.clone().text();
-            let obj;
-            try {
-              obj = JSON.parse(respBody);
-            } catch {
-              obj = {};
-            }
-            obj.retryAfterMs = retryMs;
-            const modifiedBody = JSON.stringify(obj);
-            const modifiedResponse = new Response(modifiedBody, {
-              status: response.status,
-              headers: response.headers,
-            });
-            return {
-              response: modifiedResponse,
-              url,
-              headers: finalHeaders,
-              transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
-            };
-          } catch (err) {
-            log?.warn?.("RETRY", `Failed to embed retryAfterMs: ${err}`);
-            // Fall back to original response
-          }
-        }
+    log.debug(
+      "RETRY",
+      `${response.status}, Retry-After ${retryMs ? `too long (${Math.ceil(retryMs / 1000)}s)` : "missing"}, trying fallback`
+    );
 
-        // For non-streaming clients, return the raw SSE stream with a
-        // credits-extraction TransformStream.  chatCore's non-streaming path
-        // (readNonStreamingResponseBody + parseNonStreamingSSEPayload with
-        // Gemini format support) handles draining and conversion to JSON.
-        // This replaces the previous collectStreamToResponse() approach which
-        // had an artificial timeout (now the standard FETCH_BODY_TIMEOUT_MS
-        // of 10 min applies).
-        if (!stream) {
-          // #3229: surface a real upstream error instead of masking a 4xx/5xx as an
-          // empty `chat.completion` envelope.
-          if (!response.ok) {
-            const rawBody = await response
-              .clone()
-              .text()
-              .catch(() => "");
-            const errorBody = buildAntigravityUpstreamError(
-              response.status,
-              response.statusText,
-              rawBody
-            );
-            return {
-              response: new Response(JSON.stringify(errorBody), {
-                status: response.status,
-                headers: { "Content-Type": "application/json" },
-              }),
-              url,
-              headers: finalHeaders,
-              transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
-            };
-          }
+    if (urlIndex + 1 < fallbackCount) {
+      return { action: "retryNextUrl", lastStatus: response.status };
+    }
 
-          if (response.body) {
-            // Raw SSE pass-through + credits extraction (see
-            // streamingPassthrough.ts); 499s early if the client already
-            // disconnected instead of piping a cancelled body.
-            return buildSsePassthroughResult(
-              response.body,
-              response,
-              accountId,
-              updateAntigravityRemainingCredits,
-              url,
-              finalHeaders,
-              attachToolNameMap(transformedBody, requestToolNameMap),
-              signal
-            );
-          }
+    return { action: "fallthrough", retryMs, lastStatus: response.status };
+  }
 
-          // No body -- return as-is
-          return {
-            response,
-            url,
-            headers: finalHeaders,
-            transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
-          };
-        }
+  /**
+   * Parse the 429/503 response body to classify the failure and (for
+   * quota_exhausted, non-full-exhaustion cases) attempt a Google One AI
+   * credits retry. Returns the resolved retryMs, or an early "return" result
+   * when the credits retry itself produced a response to hand back to the client.
+   */
+  private async tryResolveRetryFromErrorBody(
+    ctx: AntigravityRateLimitContext
+  ): Promise<
+    { kind: "return"; result: SsePassthroughResult } | { kind: "resolved"; retryMs: number | null }
+  > {
+    const {
+      response,
+      url,
+      headers,
+      transformedBody,
+      requestToolNameMap,
+      credentials,
+      stream,
+      signal,
+      log,
+      accountId,
+      creditsMode,
+    } = ctx;
 
-        // #2461: a non-ok upstream response (e.g. 403) must never be piped through the
-        // streaming pass-through below as if it were an SSE body. Google occasionally
-        // returns non-UTF8/binary error bodies (observed: gzip-magic-byte payloads) for
-        // 403s on this endpoint; reading/forwarding those raw bytes corrupts the
-        // client-visible error message. Mirror the non-streaming branch above and build
-        // a sanitized JSON error via buildAntigravityUpstreamError (hard rule #12)
-        // instead of streaming unknown bytes straight through.
-        if (!response.ok) {
-          const rawBody = await response
-            .clone()
-            .text()
-            .catch(() => "");
-          const errorBody = buildAntigravityUpstreamError(
-            response.status,
-            response.statusText,
-            rawBody
-          );
-          return {
-            response: new Response(JSON.stringify(errorBody), {
-              status: response.status,
-              headers: { "Content-Type": "application/json" },
-            }),
-            url,
-            headers: finalHeaders,
-            transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
-          };
-        }
+    try {
+      const errorBody = await response.clone().text();
+      const errorJson = JSON.parse(errorBody);
+      const errorMessage = buildAntigravity429ErrorMessage(errorJson);
+
+      // 1. Try to parse explicit retry time from message
+      const parsedRetryMs = this.parseRetryFromErrorMessage(errorMessage);
+
+      // 2. Classify 429, then decide the final retry time BEFORE the credits
+      //    retry so that full_quota_exhausted can skip the credits attempt
+      //    entirely (avoids ~41s hold on an already-exhausted account) and
+      //    persist the cooldown to DB for post-restart routing.
+      const category = classify429(errorMessage);
+      const decision: Decision = decide429(category, parsedRetryMs);
+      const retryMs = decision.retryAfterMs;
+      log.debug("AG_429", `Category: ${category}, Decision: ${decision.kind} — ${decision.reason}`);
+
+      if (decision.kind === "full_quota_exhausted" && retryMs) {
+        markConnectionQuotaExhausted(accountId, retryMs);
+      }
 
-        // Streaming path: wrap the response body in a pass-through TransformStream
-        // that extracts remainingCredits from the final SSE chunk(s) without
-        // consuming the stream. The client receives the unmodified SSE data.
-        if (response.body) {
-          // If the downstream client aborts, cancel the upstream fetch body immediately
-          // to release the socket back to the Undici agent pool and prevent memory leaks.
-          if (signal) {
-            const abortHandler = () => {
-              try {
-                response.body?.cancel().catch(() => {});
-              } catch (_) {}
-            };
-            if (signal.aborted) {
-              abortHandler();
-            } else {
-              signal.addEventListener("abort", abortHandler, { once: true });
-            }
-          }
+      const creditsAlreadyInjected =
+        (transformedBody as { enabledCreditTypes?: unknown }).enabledCreditTypes != null;
 
-          const passThrough = createCreditsExtractionTransform(
-            accountId,
-            16 * 1024 // 16KB sliding-window cap to prevent OOM
-          );
-          const tappedBody = response.body.pipeThrough(passThrough);
-          const tappedResponse = new Response(tappedBody, {
-            status: response.status,
-            statusText: response.statusText,
-            headers: response.headers,
-          });
-          return {
-            response: tappedResponse,
-            url,
-            headers: finalHeaders,
-            transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
-          };
-        }
+      if (category === "quota_exhausted" && creditsAlreadyInjected) {
+        handleCreditsFailure(credentials?.accessToken || "");
+        log.warn("AG_CREDITS", "Credits-first request 429'd — credits likely exhausted");
+        markCreditsExhausted(accountId);
+      }
 
-        return {
-          response,
+      if (
+        category === "quota_exhausted" &&
+        decision.kind !== "full_quota_exhausted" &&
+        !creditsAlreadyInjected &&
+        shouldRetryWithCredits(credentials?.accessToken || "", creditsMode !== "off")
+      ) {
+        const creditsResult = await tryCreditsRetry(
+          this.provider,
           url,
-          headers: finalHeaders,
-          transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
-        };
-      } catch (error) {
-        lastError = error;
-        log?.error?.(
-          "TELEMETRY",
-          `[Antigravity] Network/Fetch Error - URL: ${url}, Model: ${model}, Error: ${error instanceof Error ? error.message : String(error)}`
+          headers,
+          transformedBody,
+          requestToolNameMap,
+          credentials,
+          stream,
+          signal,
+          log,
+          accountId,
+          updateAntigravityRemainingCredits
         );
-        if (urlIndex + 1 < fallbackCount) {
-          log?.debug?.("RETRY", `Error on ${url}, trying fallback ${urlIndex + 1}`);
-          continue;
-        }
-        throw error;
+        if (creditsResult) return { kind: "return", result: creditsResult };
       }
+
+      return { kind: "resolved", retryMs };
+    } catch {
+      // Ignore parse errors, will fall back to exponential backoff
+      return { kind: "resolved", retryMs: null };
     }
+  }
 
-    throw lastError || new Error(`All ${fallbackCount} URLs failed with status ${lastStatus}`);
+  /**
+   * True for 429 always; for transient 5xx (500/502/503/504) only when the body
+   * matches a known capacity/traffic/agent-terminated pattern.
+   */
+  private async shouldAutoRetryTransient(response: Response): Promise<boolean> {
+    if (response.status === HTTP_STATUS.RATE_LIMITED) return true;
+    if (!ANTIGRAVITY_TRANSIENT_STATUSES.has(response.status)) return false;
+    try {
+      const errBody = await response.clone().text();
+      let errJson: unknown = null;
+      try {
+        errJson = errBody ? JSON.parse(errBody) : null;
+      } catch {
+        // non-JSON body — fall through to pattern match against raw text
+      }
+      const errMsg = this.extractErrorMessage(errJson, errBody);
+      return this.isTransientAntigravityError(response.status, errMsg);
+    } catch {
+      // ignore body read errors
+      return false;
+    }
   }
 }
 
diff --git a/open-sse/executors/antigravity/executeAttempt.ts b/open-sse/executors/antigravity/executeAttempt.ts
new file mode 100644
index 00000000000..566e6517132
--- /dev/null
+++ b/open-sse/executors/antigravity/executeAttempt.ts
@@ -0,0 +1,687 @@
+// Pure-ish per-attempt request/result helpers for the Antigravity executor (#7408
+// complexity-gate decomposition): building + sending one upstream request, and
+// building the final non-streaming/streaming result. No host state of their own —
+// callers inject `provider` and `onCreditsUpdate` so this module doesn't need to
+// import the executor's credit-balance cache. Extracted from antigravity.ts
+// (file-size cap), mirroring the existing antigravity/streamingPassthrough.ts and
+// antigravity/sseCollect.ts submodule pattern.
+import { mergeAbortSignals, type ExecutorLog } from "../base.ts";
+import { applyFingerprint, isCliCompatEnabled } from "../../config/cliFingerprints.ts";
+import { buildAntigravityUpstreamError } from "../antigravityUpstreamError.ts";
+import {
+  HTTP_STATUS,
+  STREAM_READINESS_TIMEOUT_MS,
+  ANTIGRAVITY_PRE_RESPONSE_TIMEOUT_CODE,
+} from "../../config/constants.ts";
+import { injectCreditsField, handleCreditsFailure } from "../../services/antigravityCredits.ts";
+import { cloakAntigravityToolPayload } from "../../config/toolCloaking.ts";
+import {
+  applyAntigravityClientProfileHeaders,
+  removeHeaderCaseInsensitive,
+} from "../../services/antigravityClientProfile.ts";
+import * as prl from "../../utils/providerRequestLogging.ts";
+import {
+  createCreditsExtractionTransform as createCreditsExtractionTransformImpl,
+  buildSsePassthroughResult,
+  type SsePassthroughResult,
+} from "./streamingPassthrough.ts";
+import type { AntigravityCredentials } from "../antigravity.ts";
+
+const LONG_RETRY_THRESHOLD_MS = 60_000;
+const CREDITS_EXHAUSTED_TTL_MS = 5 * 60 * 60 * 1000; // 5 hours
+
+/** Invoked with a fresh GOOGLE_ONE_AI credit balance to persist in the caller's cache. */
+export type OnAntigravityCreditsUpdate = (accountId: string, balance: number) => void;
+
+/**
+ * Per-account GOOGLE_ONE_AI credits-exhausted tracker.
+ * Key: accountId (OAuth subject / email). Value: expiry timestamp.
+ * When credits hit 0 we skip the credit retry for CREDITS_EXHAUSTED_TTL_MS.
+ * Lives here (not antigravity.ts) so both this module's tryCreditsRetry and
+ * antigravity.ts's tryResolveRetryFromErrorBody can share it via a single import
+ * direction (antigravity.ts -> executeAttempt.ts), avoiding a circular import.
+ */
+const MAX_CREDITS_EXHAUSTED_ENTRIES = 50;
+const creditsExhaustedUntil = new Map<string, number>();
+
+const _creditsExhaustedSweep = setInterval(() => {
+  const now = Date.now();
+  for (const [key, until] of creditsExhaustedUntil) {
+    if (now >= until) creditsExhaustedUntil.delete(key);
+  }
+}, 60_000);
+if (typeof _creditsExhaustedSweep === "object" && "unref" in _creditsExhaustedSweep) {
+  (_creditsExhaustedSweep as { unref?: () => void }).unref?.();
+}
+
+/** True while `accountId`'s Google One AI credits are marked exhausted. @internal */
+export function isCreditsExhausted(accountId: string): boolean {
+  const until = creditsExhaustedUntil.get(accountId);
+  if (!until) return false;
+  if (Date.now() >= until) {
+    creditsExhaustedUntil.delete(accountId);
+    return false;
+  }
+  return true;
+}
+
+/** Mark an account's Google One AI credits as exhausted for CREDITS_EXHAUSTED_TTL_MS. */
+export function markCreditsExhausted(accountId: string): void {
+  if (
+    creditsExhaustedUntil.size >= MAX_CREDITS_EXHAUSTED_ENTRIES &&
+    !creditsExhaustedUntil.has(accountId)
+  ) {
+    const now = Date.now();
+    for (const [key, until] of creditsExhaustedUntil) {
+      if (now >= until) {
+        creditsExhaustedUntil.delete(key);
+      }
+    }
+    if (creditsExhaustedUntil.size >= MAX_CREDITS_EXHAUSTED_ENTRIES) {
+      const oldestKey = creditsExhaustedUntil.keys().next().value;
+      if (oldestKey !== undefined) creditsExhaustedUntil.delete(oldestKey);
+    }
+  }
+  creditsExhaustedUntil.set(accountId, Date.now() + CREDITS_EXHAUSTED_TTL_MS);
+}
+
+class AntigravityPreResponseTimeoutError extends Error {
+  code = ANTIGRAVITY_PRE_RESPONSE_TIMEOUT_CODE;
+  status = HTTP_STATUS.GATEWAY_TIMEOUT;
+
+  constructor(timeoutMs: number, url: string) {
+    super(`Antigravity upstream did not return response headers within ${timeoutMs}ms: ${url}`);
+    this.name = "TimeoutError";
+  }
+}
+
+function getAbortErrorCode(error: unknown): string | null {
+  if (!error || typeof error !== "object") return null;
+  const value = (error as { code?: unknown }).code;
+  return typeof value === "string" ? value : null;
+}
+
+function isAntigravityPreResponseTimeout(error: unknown): boolean {
+  return getAbortErrorCode(error) === ANTIGRAVITY_PRE_RESPONSE_TIMEOUT_CODE;
+}
+
+/**
+ * `fetch()` wrapper that aborts if the upstream never returns response headers
+ * within `timeoutMs` (default STREAM_READINESS_TIMEOUT_MS) — distinct from the
+ * overall FETCH_TIMEOUT_MS, which bounds the whole request including body streaming.
+ * Shared by every fetch attempt in executeOnce() (initial, 403-retry, credits-retry).
+ */
+export async function fetchAntigravityWithReadinessTimeout(
+  url: string,
+  init: RequestInit,
+  timeoutMs = STREAM_READINESS_TIMEOUT_MS
+): Promise<Response> {
+  const boundedTimeoutMs = Math.max(0, Math.floor(timeoutMs));
+  if (boundedTimeoutMs <= 0) {
+    return fetch(url, init);
+  }
+
+  const timeoutController = new AbortController();
+  let timeoutId: ReturnType<typeof setTimeout> | null = setTimeout(() => {
+    timeoutController.abort(new AntigravityPreResponseTimeoutError(boundedTimeoutMs, url));
+  }, boundedTimeoutMs);
+
+  const existingSignal = init.signal instanceof AbortSignal ? init.signal : null;
+  const combinedSignal = existingSignal
+    ? mergeAbortSignals(existingSignal, timeoutController.signal)
+    : timeoutController.signal;
+
+  try {
+    return await fetch(url, { ...init, signal: combinedSignal });
+  } catch (error) {
+    if (
+      timeoutController.signal.aborted &&
+      isAntigravityPreResponseTimeout(timeoutController.signal.reason)
+    ) {
+      throw timeoutController.signal.reason;
+    }
+    throw error;
+  } finally {
+    if (timeoutId) {
+      clearTimeout(timeoutId);
+      timeoutId = null;
+    }
+  }
+}
+
+/** ExecutorLog with every method always callable — see toSafeAntigravityLog(). */
+export type SafeAntigravityLog = Required<ExecutorLog>;
+
+function noopLogFn(): void {}
+
+/**
+ * Normalize a possibly-null/undefined ExecutorLog into an object with all four
+ * methods always callable, so the request/retry helpers below can call
+ * `l.debug(...)` directly instead of repeating `log?.debug?.(...)` at every call
+ * site. This isn't just style: the complexity linter (eslint `complexity` rule)
+ * weighs each `?.` link in a chain as its own branch — a doubly-chained
+ * `log?.debug?.(...)` costs +2 — so a logging-heavy helper can rack up a large
+ * complexity score with zero real decision points. Resolving once here keeps
+ * the actual branch count legible in the functions that matter.
+ */
+export function toSafeAntigravityLog(log: ExecutorLog | null | undefined): SafeAntigravityLog {
+  return {
+    debug: log?.debug ? log.debug.bind(log) : noopLogFn,
+    info: log?.info ? log.info.bind(log) : noopLogFn,
+    warn: log?.warn ? log.warn.bind(log) : noopLogFn,
+    error: log?.error ? log.error.bind(log) : noopLogFn,
+  };
+}
+
+/** Flatten a 429/503 error JSON body (message + `error.details[].reason`) into one string. */
+export function buildAntigravity429ErrorMessage(errorJson: unknown): string {
+  const obj = errorJson as
+    | { error?: { message?: unknown; details?: unknown }; message?: unknown }
+    | null
+    | undefined;
+  let errorMessage = String(obj?.error?.message || obj?.message || "");
+  const details = obj?.error?.details;
+  if (Array.isArray(details)) {
+    for (const detail of details) {
+      const reason = (detail as { reason?: unknown } | null)?.reason;
+      if (reason) errorMessage += ` ${reason}`;
+    }
+  }
+  return errorMessage;
+}
+
+function getChunkedOrFixedBody(bodyStr: string, stream: boolean): BodyInit {
+  if (stream) {
+    return new ReadableStream(
+      {
+        async start(controller) {
+          controller.enqueue(new TextEncoder().encode(bodyStr));
+          controller.close();
+        },
+      },
+      { highWaterMark: 16384 }
+    );
+  }
+  return bodyStr;
+}
+
+function cloneAntigravityRequestBody(body: unknown): unknown {
+  if (!body || typeof body !== "object") {
+    return body;
+  }
+
+  try {
+    return structuredClone(body);
+  } catch {
+    return JSON.parse(JSON.stringify(body));
+  }
+}
+
+function serializeAntigravityRequest(
+  provider: string,
+  headers: Record<string, string>,
+  body: unknown
+): { headers: Record<string, string>; bodyString: string } {
+  const serializedBody = cloneAntigravityRequestBody(body);
+
+  if (!isCliCompatEnabled(provider)) {
+    return { headers, bodyString: JSON.stringify(serializedBody) };
+  }
+  return applyFingerprint(provider, { ...headers }, serializedBody);
+}
+
+function getRequestTargetModel(body: Record<string, unknown>): string {
+  const target = body.model;
+  return typeof target === "string" && target.length > 0 ? target : "unknown";
+}
+
+function attachToolNameMap<T>(payload: T, toolNameMap: Map<string, string> | null): T {
+  if (!toolNameMap?.size || !payload || typeof payload !== "object") {
+    return payload;
+  }
+
+  const copy = Array.isArray(payload) ? ([...payload] as T) : ({ ...(payload as object) } as T);
+  Object.defineProperty(copy, "_toolNameMap", {
+    value: toolNameMap,
+    enumerable: false,
+    configurable: true,
+    writable: true,
+  });
+  return copy;
+}
+
+/** Cloak the tool-name payload, then apply credits-first injection, for one attempt. */
+export function finalizeAntigravityRequestBody(
+  transformed: Record<string, unknown>,
+  useCreditsFirst: boolean,
+  log: SafeAntigravityLog
+): {
+  transformedBody: Record<string, unknown>;
+  requestToolNameMap: Map<string, string> | null;
+} {
+  let transformedBody: Record<string, unknown> = transformed;
+  let requestToolNameMap: Map<string, string> | null = null;
+
+  if (transformedBody && typeof transformedBody === "object") {
+    const cloaked = cloakAntigravityToolPayload(transformedBody);
+    transformedBody = cloaked.body;
+    requestToolNameMap = cloaked.toolNameMap;
+  }
+
+  // Credits-first: inject GOOGLE_ONE_AI upfront so we never try the normal
+  // quota path. If credits are exhausted / disabled shouldUseCreditsFirst()
+  // returns false and we fall back to the legacy retry-on-429 flow.
+  if (useCreditsFirst) {
+    transformedBody = injectCreditsField(transformedBody);
+    log.debug("AG_CREDITS", "Credits-first enabled (ANTIGRAVITY_CREDITS=always)");
+  }
+
+  return { transformedBody, requestToolNameMap };
+}
+
+/** Debug-only dump of outgoing headers (mask Authorization) and envelope shape. */
+function dumpAntigravityRequestDebug(
+  finalHeaders: Record<string, string>,
+  transformedBody: Record<string, unknown>,
+  clientProfile: unknown,
+  log: SafeAntigravityLog
+): void {
+  const safeHeaders = { ...finalHeaders };
+  if (safeHeaders["Authorization"]) safeHeaders["Authorization"] = "Bearer ***";
+  log.debug("AG_REQUEST_HEADERS", JSON.stringify(safeHeaders));
+
+  const envelope = transformedBody as Record<string, unknown>;
+  const requestInner = envelope.request as Record<string, unknown> | undefined;
+  log.debug(
+    "AG_REQUEST_ENVELOPE",
+    JSON.stringify({
+      fieldOrder: Object.keys(envelope),
+      project: envelope.project,
+      requestId: envelope.requestId,
+      model: envelope.model,
+      userAgent: envelope.userAgent,
+      requestType: envelope.requestType,
+      enabledCreditTypes: envelope.enabledCreditTypes,
+      clientProfile,
+      sessionId: requestInner?.sessionId,
+      generationConfig: requestInner?.generationConfig,
+    })
+  );
+}
+
+/**
+ * Send one Antigravity request attempt: serialize + apply the client-profile
+ * fingerprint, debug-dump the outgoing envelope, fetch with a readiness timeout,
+ * and transparently retry once without `x-goog-user-project` on a 403 (some
+ * projects reject that header). Returns the (possibly 403-retried) response
+ * plus the headers actually used for it.
+ */
+export async function sendAntigravityRequest(
+  provider: string,
+  url: string,
+  model: string,
+  headers: Record<string, string>,
+  transformedBody: Record<string, unknown>,
+  credentials: AntigravityCredentials,
+  stream: boolean,
+  signal: AbortSignal | null | undefined,
+  log: SafeAntigravityLog,
+  retryAttempt: number
+): Promise<{ response: Response; finalHeaders: Record<string, string> }> {
+  const serializedRequest = serializeAntigravityRequest(provider, headers, transformedBody);
+  let finalHeaders = serializedRequest.headers;
+  const clientProfile = applyAntigravityClientProfileHeaders(
+    finalHeaders,
+    credentials,
+    transformedBody
+  );
+
+  log.debug(
+    "TELEMETRY",
+    `[Antigravity] Execute - URL: ${url}, Model: ${model}, Target: ${getRequestTargetModel(transformedBody)}, RetryAttempt: ${retryAttempt}`
+  );
+
+  // Dump outgoing headers (mask Authorization) and envelope shape for debugging.
+  // Gated behind an explicit typeof check (not just calling log.debug() unconditionally)
+  // so the JSON.stringify work below is skipped entirely when debug logging is off.
+  if (typeof log.debug === "function") {
+    dumpAntigravityRequestDebug(finalHeaders, transformedBody, clientProfile, log);
+  }
+
+  await prl.captureCurrentProviderBody(url, finalHeaders, serializedRequest.bodyString, log);
+  let response = await fetchAntigravityWithReadinessTimeout(url, {
+    method: "POST",
+    headers: finalHeaders,
+    body: getChunkedOrFixedBody(serializedRequest.bodyString, stream),
+    ...(stream ? { duplex: "half" } : {}),
+    signal,
+  });
+
+  if (response.status === HTTP_STATUS.FORBIDDEN && finalHeaders["x-goog-user-project"]) {
+    const retryHeaders = { ...finalHeaders };
+    removeHeaderCaseInsensitive(retryHeaders, "x-goog-user-project");
+    log.debug("RETRY", "403 with x-goog-user-project, retrying once without it");
+    await prl.captureCurrentProviderBody(url, retryHeaders, serializedRequest.bodyString, log);
+    response = await fetchAntigravityWithReadinessTimeout(url, {
+      method: "POST",
+      headers: retryHeaders,
+      body: getChunkedOrFixedBody(serializedRequest.bodyString, stream),
+      ...(stream ? { duplex: "half" } : {}),
+      signal,
+    });
+    finalHeaders = retryHeaders;
+  }
+
+  if (!response.ok) {
+    log.warn(
+      "TELEMETRY",
+      `[Antigravity] Error Response - URL: ${url}, Status: ${response.status}, Model: ${model}`
+    );
+  }
+
+  return { response, finalHeaders };
+}
+
+/**
+ * Retry the SAME url with `enabledCreditTypes: ["GOOGLE_ONE_AI"]` injected, for a
+ * quota_exhausted 429 that hasn't already tried credits. Returns the result to hand
+ * back to the caller of execute() on success (or a non-429 status), or null if the
+ * credits retry also failed/429'd (caller falls through to the normal retry logic).
+ */
+export async function tryCreditsRetry(
+  provider: string,
+  url: string,
+  headers: Record<string, string>,
+  transformedBody: Record<string, unknown>,
+  requestToolNameMap: Map<string, string> | null,
+  credentials: AntigravityCredentials,
+  stream: boolean,
+  signal: AbortSignal | null | undefined,
+  log: SafeAntigravityLog,
+  accountId: string,
+  onCreditsUpdate: OnAntigravityCreditsUpdate
+): Promise<SsePassthroughResult | null> {
+  log.info("AG_CREDITS", "Retrying with Google One AI credits");
+  const creditsBody = injectCreditsField(transformedBody);
+  const serializedCreditsRequest = serializeAntigravityRequest(provider, headers, creditsBody);
+  const finalCreditsHeaders = serializedCreditsRequest.headers;
+  try {
+    await prl.captureCurrentProviderBody(
+      url,
+      finalCreditsHeaders,
+      serializedCreditsRequest.bodyString,
+      log
+    );
+    const creditsResp = await fetchAntigravityWithReadinessTimeout(url, {
+      method: "POST",
+      headers: finalCreditsHeaders,
+      body: getChunkedOrFixedBody(serializedCreditsRequest.bodyString, stream),
+      ...(stream ? { duplex: "half" } : {}),
+      signal,
+    });
+    if (creditsResp.ok || creditsResp.status !== HTTP_STATUS.RATE_LIMITED) {
+      log.info("AG_CREDITS", `Credits retry succeeded: ${creditsResp.status}`);
+      if (!stream && creditsResp.body) {
+        // Raw SSE pass-through + credits extraction (see
+        // streamingPassthrough.ts); 499s early if the client
+        // already disconnected instead of piping a cancelled body.
+        return buildSsePassthroughResult(
+          creditsResp.body,
+          creditsResp,
+          accountId,
+          onCreditsUpdate,
+          url,
+          finalCreditsHeaders,
+          attachToolNameMap(creditsBody, requestToolNameMap),
+          signal
+        );
+      }
+      return {
+        response: creditsResp,
+        url,
+        headers: finalCreditsHeaders,
+        transformedBody: attachToolNameMap(creditsBody, requestToolNameMap),
+      };
+    }
+
+    // Credit retry also 429'd
+    handleCreditsFailure(credentials?.accessToken || "");
+    log.warn("AG_CREDITS", "Credits retry also 429'd");
+
+    // Also mark in our legacy exhaustion map to avoid retrying other routes
+    markCreditsExhausted(accountId);
+    return null;
+  } catch (creditsErr) {
+    handleCreditsFailure(credentials?.accessToken || "");
+    log.warn("AG_CREDITS", `Credits retry failed: ${creditsErr}`);
+    return null;
+  }
+}
+
+/**
+ * If we have a 429 with a long retry time (> LONG_RETRY_THRESHOLD_MS), embed
+ * `retryAfterMs` in the response body so the caller (combo/account-fallback
+ * layer) can read it back out. Returns null (fall back to the original
+ * response handling) when the status/retryMs don't qualify, or on error.
+ */
+export async function tryEmbedLongRetryAfter(
+  response: Response,
+  retryMs: number | null,
+  url: string,
+  finalHeaders: Record<string, string>,
+  transformedBody: Record<string, unknown>,
+  requestToolNameMap: Map<string, string> | null,
+  log: ExecutorLog | null | undefined
+): Promise<SsePassthroughResult | null> {
+  if (
+    response.status !== HTTP_STATUS.RATE_LIMITED ||
+    !retryMs ||
+    retryMs <= LONG_RETRY_THRESHOLD_MS
+  ) {
+    return null;
+  }
+  try {
+    const respBody = await response.clone().text();
+    let obj;
+    try {
+      obj = JSON.parse(respBody);
+    } catch {
+      obj = {};
+    }
+    obj.retryAfterMs = retryMs;
+    const modifiedBody = JSON.stringify(obj);
+    const modifiedResponse = new Response(modifiedBody, {
+      status: response.status,
+      headers: response.headers,
+    });
+    return {
+      response: modifiedResponse,
+      url,
+      headers: finalHeaders,
+      transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
+    };
+  } catch (err) {
+    log?.warn?.("RETRY", `Failed to embed retryAfterMs: ${err}`);
+    return null;
+  }
+}
+
+/** Build the sanitized JSON error result shared by the non-streaming and streaming paths. */
+async function buildUpstreamErrorResult(
+  response: Response,
+  url: string,
+  finalHeaders: Record<string, string>,
+  transformedBody: Record<string, unknown>,
+  requestToolNameMap: Map<string, string> | null
+): Promise<SsePassthroughResult> {
+  const rawBody = await response
+    .clone()
+    .text()
+    .catch(() => "");
+  const errorBody = buildAntigravityUpstreamError(response.status, response.statusText, rawBody);
+  return {
+    response: new Response(JSON.stringify(errorBody), {
+      status: response.status,
+      headers: { "Content-Type": "application/json" },
+    }),
+    url,
+    headers: finalHeaders,
+    transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
+  };
+}
+
+/**
+ * For non-streaming clients, return the raw SSE stream with a
+ * credits-extraction TransformStream.  chatCore's non-streaming path
+ * (readNonStreamingResponseBody + parseNonStreamingSSEPayload with
+ * Gemini format support) handles draining and conversion to JSON.
+ * This replaces the previous collectStreamToResponse() approach which
+ * had an artificial timeout (now the standard FETCH_BODY_TIMEOUT_MS
+ * of 10 min applies).
+ */
+async function buildNonStreamingExecuteOnceResult(
+  response: Response,
+  url: string,
+  finalHeaders: Record<string, string>,
+  transformedBody: Record<string, unknown>,
+  requestToolNameMap: Map<string, string> | null,
+  accountId: string,
+  signal: AbortSignal | null | undefined,
+  onCreditsUpdate: OnAntigravityCreditsUpdate
+): Promise<SsePassthroughResult> {
+  // #3229: surface a real upstream error instead of masking a 4xx/5xx as an
+  // empty `chat.completion` envelope.
+  if (!response.ok) {
+    return buildUpstreamErrorResult(response, url, finalHeaders, transformedBody, requestToolNameMap);
+  }
+
+  if (response.body) {
+    // Raw SSE pass-through + credits extraction (see
+    // streamingPassthrough.ts); 499s early if the client already
+    // disconnected instead of piping a cancelled body.
+    return buildSsePassthroughResult(
+      response.body,
+      response,
+      accountId,
+      onCreditsUpdate,
+      url,
+      finalHeaders,
+      attachToolNameMap(transformedBody, requestToolNameMap),
+      signal
+    );
+  }
+
+  // No body -- return as-is
+  return {
+    response,
+    url,
+    headers: finalHeaders,
+    transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
+  };
+}
+
+/**
+ * Streaming path: wrap the response body in a pass-through TransformStream
+ * that extracts remainingCredits from the final SSE chunk(s) without
+ * consuming the stream. The client receives the unmodified SSE data.
+ *
+ * #2461: a non-ok upstream response (e.g. 403) must never be piped through the
+ * streaming pass-through below as if it were an SSE body. Google occasionally
+ * returns non-UTF8/binary error bodies (observed: gzip-magic-byte payloads) for
+ * 403s on this endpoint; reading/forwarding those raw bytes corrupts the
+ * client-visible error message. Mirror the non-streaming branch above and build
+ * a sanitized JSON error via buildAntigravityUpstreamError (hard rule #12)
+ * instead of streaming unknown bytes straight through.
+ */
+async function buildStreamingExecuteOnceResult(
+  response: Response,
+  url: string,
+  finalHeaders: Record<string, string>,
+  transformedBody: Record<string, unknown>,
+  requestToolNameMap: Map<string, string> | null,
+  accountId: string,
+  signal: AbortSignal | null | undefined,
+  onCreditsUpdate: OnAntigravityCreditsUpdate
+): Promise<SsePassthroughResult> {
+  if (!response.ok) {
+    return buildUpstreamErrorResult(response, url, finalHeaders, transformedBody, requestToolNameMap);
+  }
+
+  if (response.body) {
+    // If the downstream client aborts, cancel the upstream fetch body immediately
+    // to release the socket back to the Undici agent pool and prevent memory leaks.
+    if (signal) {
+      const abortHandler = () => {
+        try {
+          response.body?.cancel().catch(() => {});
+        } catch (_) {}
+      };
+      if (signal.aborted) {
+        abortHandler();
+      } else {
+        signal.addEventListener("abort", abortHandler, { once: true });
+      }
+    }
+
+    const passThrough = createCreditsExtractionTransformImpl(
+      accountId,
+      onCreditsUpdate,
+      16 * 1024 // 16KB sliding-window cap to prevent OOM
+    );
+    const tappedBody = response.body.pipeThrough(passThrough);
+    const tappedResponse = new Response(tappedBody, {
+      status: response.status,
+      statusText: response.statusText,
+      headers: response.headers,
+    });
+    return {
+      response: tappedResponse,
+      url,
+      headers: finalHeaders,
+      transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
+    };
+  }
+
+  return {
+    response,
+    url,
+    headers: finalHeaders,
+    transformedBody: attachToolNameMap(transformedBody, requestToolNameMap),
+  };
+}
+
+/** Dispatch to the non-streaming or streaming final-result builder. */
+export async function buildFinalAntigravityResult(
+  stream: boolean,
+  response: Response,
+  url: string,
+  finalHeaders: Record<string, string>,
+  transformedBody: Record<string, unknown>,
+  requestToolNameMap: Map<string, string> | null,
+  accountId: string,
+  signal: AbortSignal | null | undefined,
+  onCreditsUpdate: OnAntigravityCreditsUpdate
+): Promise<SsePassthroughResult> {
+  if (!stream) {
+    return buildNonStreamingExecuteOnceResult(
+      response,
+      url,
+      finalHeaders,
+      transformedBody,
+      requestToolNameMap,
+      accountId,
+      signal,
+      onCreditsUpdate
+    );
+  }
+  return buildStreamingExecuteOnceResult(
+    response,
+    url,
+    finalHeaders,
+    transformedBody,
+    requestToolNameMap,
+    accountId,
+    signal,
+    onCreditsUpdate
+  );
+}
diff --git a/open-sse/handlers/sseParser/geminiResponse.ts b/open-sse/handlers/sseParser/geminiResponse.ts
index d0801019bd2..1657cf20309 100644
--- a/open-sse/handlers/sseParser/geminiResponse.ts
+++ b/open-sse/handlers/sseParser/geminiResponse.ts
@@ -3,134 +3,149 @@
 // state, following the handlers submodule pattern (chatCore/, responseSanitizer/).
 import { normalizeOpenAICompatibleFinishReasonString } from "../../utils/finishReason.ts";
 
+type AccumulatedToolCall = {
+  id: string;
+  index: number;
+  type: "function";
+  function: { name: string; arguments: string };
+};
+
+/** Mutable accumulator threaded through one SSE payload's worth of parsing. */
+type GeminiSSEAccumulator = {
+  textContent: string;
+  finishReason: string;
+  usage: Record<string, unknown> | null;
+  sawContent: boolean;
+  toolCalls: AccumulatedToolCall[];
+};
+
+function stripZeroWidth(value: unknown): unknown {
+  if (typeof value === "string") return value.replace(/[\u200B-\u200D\uFEFF]/g, "");
+  return value;
+}
+
 /**
- * Convert Gemini/Antigravity SSE chunks into a single non-streaming OpenAI
- * chat.completion JSON response.  Gemini SSE carries payloads like:
- *
- *   data: {"markdown":"...chunk..."}
- *   data: {"response":{"candidates":[{"content":{"parts":[{"text":"..."}]},"finishReason":"STOP"}],"usageMetadata":{...}}}
- *   data: {"remainingCredits":[...]}
- *
- * Reuses the same parsing logic as processAntigravitySSEPayload() in sseCollect.ts
- * so that format conversion is functionally equivalent to the previous
- * collectStreamToResponse() approach.  Intentional differences:
- *   - remainingCredits is NOT embedded into the result (handled separately
- *     by the credits-extraction TransformStream in antigravity.ts).
- *   - The synthetic `id` uses `chatcmpl-${Date.now()}` (no UUID suffix)
- *     because this path runs once per response, not per chunk.
+ * Detect the `[Tool call: name]\nArguments: {...}` textual convention some
+ * Gemini/Antigravity models emit instead of a native functionCall part.
  */
-export function parseSSEToGeminiResponse(
-  rawSSE: string,
-  fallbackModel: string
-): Record<string, unknown> | null {
-  const lines = String(rawSSE || "").split("\n");
-  let textContent = "";
-  let finishReason = "stop";
-  let usage: Record<string, unknown> | null = null;
-  let sawContent = false;
-
-  type AccumulatedToolCall = {
-    id: string;
-    index: number;
-    type: "function";
-    function: { name: string; arguments: string };
-  };
-  const toolCalls: AccumulatedToolCall[] = [];
+function tryParseTextualToolCall(text: string): { name: string; args: unknown } | null {
+  const normalized = text.replace(/[\u200B-\u200D\uFEFF]/g, "");
+  const match = normalized.match(
+    /^[\s\S]*?\[Tool call:\s*([^\]\n]+)\]\s*\nArguments:\s*([\s\S]+?)\s*$/
+  );
+  if (!match) return null;
+  const name = match[1]?.trim();
+  const rawArgs = match[2]?.trim();
+  if (!name || !rawArgs) return null;
+  try {
+    return { name, args: stripZeroWidth(JSON.parse(rawArgs)) };
+  } catch {
+    return null;
+  }
+}
 
-  const stripZeroWidth = (value: unknown): unknown => {
-    if (typeof value === "string") return value.replace(/[\u200B-\u200D\uFEFF]/g, "");
-    return value;
-  };
+/** Extract the markdown shortcut some Gemini variants send (top-level or nested). */
+function extractGeminiMarkdownShortcut(parsed: Record<string, unknown>): string | null {
+  if (typeof parsed.markdown === "string") return parsed.markdown;
+  const response = parsed.response as Record<string, unknown> | undefined;
+  return typeof response?.markdown === "string" ? response.markdown : null;
+}
 
-  const tryParseTextualToolCall = (text: string): { name: string; args: unknown } | null => {
-    const normalized = text.replace(/[\u200B-\u200D\uFEFF]/g, "");
-    const match = normalized.match(
-      /^[\s\S]*?\[Tool call:\s*([^\]\n]+)\]\s*\nArguments:\s*([\s\S]+?)\s*$/
-    );
-    if (!match) return null;
-    const name = match[1]?.trim();
-    const rawArgs = match[2]?.trim();
-    if (!name || !rawArgs) return null;
-    try {
-      return { name, args: stripZeroWidth(JSON.parse(rawArgs)) };
-    } catch {
-      return null;
-    }
+/** Append one candidate content part (text or textual tool call) onto the accumulator. */
+function applyCandidatePart(part: Record<string, unknown>, acc: GeminiSSEAccumulator): void {
+  if (typeof part.text !== "string" || part.thought || part.thoughtSignature) return;
+
+  const textualToolCall = tryParseTextualToolCall(part.text);
+  if (textualToolCall) {
+    acc.toolCalls.push({
+      id: `${textualToolCall.name}-${Date.now()}-${acc.toolCalls.length}`,
+      index: acc.toolCalls.length,
+      type: "function",
+      function: {
+        name: textualToolCall.name,
+        arguments: JSON.stringify(textualToolCall.args || {}),
+      },
+    });
+  } else {
+    acc.textContent += part.text;
+  }
+  acc.sawContent = true;
+}
+
+/** Walk the first candidate's content parts, if present, mutating the accumulator. */
+function applyCandidateContentParts(
+  candidate: Record<string, unknown> | undefined,
+  acc: GeminiSSEAccumulator
+): void {
+  const content = candidate?.content as Record<string, unknown> | undefined;
+  const parts = content?.parts;
+  if (!Array.isArray(parts)) return;
+  for (const part of parts) {
+    applyCandidatePart(part as Record<string, unknown>, acc);
+  }
+}
+
+/** Normalize and apply the candidate's finishReason, if present. */
+function applyFinishReason(
+  candidate: Record<string, unknown> | undefined,
+  acc: GeminiSSEAccumulator
+): void {
+  if (!candidate?.finishReason) return;
+  acc.finishReason = normalizeOpenAICompatibleFinishReasonString(
+    String(candidate.finishReason).toLowerCase()
+  );
+}
+
+/** Extract usageMetadata into the OpenAI-shaped usage object, if present. */
+function applyUsageMetadata(parsed: Record<string, unknown>, acc: GeminiSSEAccumulator): void {
+  const response = parsed.response as Record<string, unknown> | undefined;
+  const um = response?.usageMetadata as Record<string, unknown> | undefined;
+  if (!um) return;
+  acc.usage = {
+    prompt_tokens: um.promptTokenCount || 0,
+    completion_tokens: um.candidatesTokenCount || 0,
+    total_tokens: um.totalTokenCount || 0,
   };
+}
 
-  for (const line of lines) {
-    const trimmed = line.trim();
-    if (!trimmed.startsWith("data:")) continue;
-    const payload = trimmed.slice(5).trim();
-    if (!payload || payload === "[DONE]") continue;
+/** Parse one `data:` line's JSON payload and fold it into the accumulator (best-effort). */
+function applyGeminiSSEDataLine(payload: string, acc: GeminiSSEAccumulator): void {
+  try {
+    const parsed = JSON.parse(payload) as Record<string, unknown>;
 
-    try {
-      const parsed = JSON.parse(payload);
-
-      // Markdown shortcut (some Gemini variants)
-      const markdown =
-        typeof parsed?.markdown === "string"
-          ? parsed.markdown
-          : typeof parsed?.response?.markdown === "string"
-            ? parsed.response.markdown
-            : null;
-      if (markdown) {
-        textContent += markdown;
-        sawContent = true;
-      }
-
-      // Candidate content parts
-      const candidate = parsed?.response?.candidates?.[0];
-      if (candidate?.content?.parts) {
-        for (const part of candidate.content.parts) {
-          if (typeof part.text === "string" && !part.thought && !part.thoughtSignature) {
-            const textualToolCall = tryParseTextualToolCall(part.text);
-            if (textualToolCall) {
-              toolCalls.push({
-                id: `${textualToolCall.name}-${Date.now()}-${toolCalls.length}`,
-                index: toolCalls.length,
-                type: "function",
-                function: {
-                  name: textualToolCall.name,
-                  arguments: JSON.stringify(textualToolCall.args || {}),
-                },
-              });
-            } else {
-              textContent += part.text;
-            }
-            sawContent = true;
-          }
-        }
-      }
-
-      if (candidate?.finishReason) {
-        finishReason = normalizeOpenAICompatibleFinishReasonString(
-          String(candidate.finishReason).toLowerCase()
-        );
-      }
-
-      if (parsed?.response?.usageMetadata) {
-        const um = parsed.response.usageMetadata;
-        usage = {
-          prompt_tokens: um.promptTokenCount || 0,
-          completion_tokens: um.candidatesTokenCount || 0,
-          total_tokens: um.totalTokenCount || 0,
-        };
-      }
-    } catch {
-      // Ignore malformed lines
+    const markdown = extractGeminiMarkdownShortcut(parsed);
+    if (markdown) {
+      acc.textContent += markdown;
+      acc.sawContent = true;
     }
-  }
 
-  if (!sawContent && toolCalls.length === 0) return null;
+    const response = parsed.response as Record<string, unknown> | undefined;
+    const candidates = response?.candidates;
+    const candidate = Array.isArray(candidates)
+      ? (candidates[0] as Record<string, unknown> | undefined)
+      : undefined;
+
+    applyCandidateContentParts(candidate, acc);
+    applyFinishReason(candidate, acc);
+    applyUsageMetadata(parsed, acc);
+  } catch {
+    // Ignore malformed lines
+  }
+}
 
+/** Assemble the final non-streaming chat.completion payload from the accumulator. */
+function buildChatCompletionFromAccumulator(
+  acc: GeminiSSEAccumulator,
+  fallbackModel: string
+): Record<string, unknown> {
   const message: Record<string, unknown> = {
     role: "assistant",
-    content: textContent || null,
+    content: acc.textContent || null,
   };
 
-  if (toolCalls.length > 0) {
-    message.tool_calls = toolCalls;
+  let finishReason = acc.finishReason;
+  if (acc.toolCalls.length > 0) {
+    message.tool_calls = acc.toolCalls;
     finishReason = "tool_calls";
   }
 
@@ -148,9 +163,52 @@ export function parseSSEToGeminiResponse(
     ],
   };
 
-  if (usage) {
-    result.usage = usage;
+  if (acc.usage) {
+    result.usage = acc.usage;
   }
 
   return result;
 }
+
+/**
+ * Convert Gemini/Antigravity SSE chunks into a single non-streaming OpenAI
+ * chat.completion JSON response.  Gemini SSE carries payloads like:
+ *
+ *   data: {"markdown":"...chunk..."}
+ *   data: {"response":{"candidates":[{"content":{"parts":[{"text":"..."}]},"finishReason":"STOP"}],"usageMetadata":{...}}}
+ *   data: {"remainingCredits":[...]}
+ *
+ * Reuses the same parsing logic as processAntigravitySSEPayload() in sseCollect.ts
+ * so that format conversion is functionally equivalent to the previous
+ * collectStreamToResponse() approach.  Intentional differences:
+ *   - remainingCredits is NOT embedded into the result (handled separately
+ *     by the credits-extraction TransformStream in antigravity.ts).
+ *   - The synthetic `id` uses `chatcmpl-${Date.now()}` (no UUID suffix)
+ *     because this path runs once per response, not per chunk.
+ */
+export function parseSSEToGeminiResponse(
+  rawSSE: string,
+  fallbackModel: string
+): Record<string, unknown> | null {
+  const lines = String(rawSSE || "").split("\n");
+  const acc: GeminiSSEAccumulator = {
+    textContent: "",
+    finishReason: "stop",
+    usage: null,
+    sawContent: false,
+    toolCalls: [],
+  };
+
+  for (const line of lines) {
+    const trimmed = line.trim();
+    if (!trimmed.startsWith("data:")) continue;
+    const payload = trimmed.slice(5).trim();
+    if (!payload || payload === "[DONE]") continue;
+
+    applyGeminiSSEDataLine(payload, acc);
+  }
+
+  if (!acc.sawContent && acc.toolCalls.length === 0) return null;
+
+  return buildChatCompletionFromAccumulator(acc, fallbackModel);
+}

From b3054d19cc47638ad801de6dead20e9670bfc2a1 Mon Sep 17 00:00:00 2001
From: Minxi Hou <houminxi@gmail.com>
Date: Sat, 18 Jul 2026 16:05:32 -0300
Subject: [PATCH 13/14] Merge branch 'release/v3.8.49' into
 fix/antigravity-streaming-passthrough
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Resolves conflict in open-sse/executors/antigravity.ts between this
branch's streaming-passthrough decomposition and #7290's fallback-chain
decomposition (already merged into release/v3.8.49) — both sides added
imports from the same new antigravity/ submodule files, kept both.

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---
 .env.example                                  |    4 +
 .github/workflows/ci.yml                      |   36 +-
 .github/workflows/codeql.yml                  |    4 +-
 .github/workflows/dast-smoke.yml              |    2 +-
 .github/workflows/electron-release.yml        |    2 +-
 .github/workflows/mutation-redundancy.yml     |    2 +-
 .github/workflows/nightly-compat.yml          |    4 +-
 .github/workflows/nightly-llm-security.yml    |    4 +-
 .github/workflows/nightly-mutation.yml        |    4 +-
 .github/workflows/nightly-property.yml        |    2 +-
 .github/workflows/nightly-release-green.yml   |    4 +-
 .github/workflows/nightly-resilience.yml      |    8 +-
 .github/workflows/nightly-schemathesis.yml    |    2 +-
 .github/workflows/npm-publish.yml             |    4 +-
 .github/workflows/opencode-plugin-ci.yml      |    4 +-
 .github/workflows/opencode-provider-ci.yml    |    4 +-
 .github/workflows/quality.yml                 |   14 +-
 .github/workflows/wiki-sync.yml               |    2 +-
 README.md                                     |    6 +-
 .../features/7318-router-eval-harness.md      |    1 +
 .../provider-quota-connection-visibility.md   |    3 +
 .../fixes/7032-auggie-model-ids-v032.md       |    1 +
 changelog.d/fixes/7490-align-engines-node.md  |    1 +
 .../fixes/7547-prefer-public-endpoint-url.md  |    1 +
 .../fixes/pending-cli-service-detection.md    |    1 +
 .../fixes/pending-react-flow-dark-theme.md    |    1 +
 .../7334-incident-response-runbook.md         |    1 +
 .../7336-perf-latency-budgets-doc.md          |    1 +
 config/quality/complexity-baseline.json       |    3 +-
 config/quality/dependency-allowlist.json      |    1 +
 config/quality/file-size-baseline.json        |   14 +-
 config/quality/test-discovery-baseline.json   |    1 -
 docs/INCIDENT_RESPONSE.md                     |  194 +++
 docs/PERF_BUDGETS.md                          |  227 +++
 docs/compression/COMPRESSION_GUIDE.md         |    2 +
 docs/compression/RTK_COMPRESSION.md           |   52 +-
 docs/frameworks/MCP-SERVER.md                 |   30 +-
 docs/openapi.yaml                             |   30 +
 docs/reference/ENVIRONMENT.md                 |    1 +
 docs/reference/PROVIDER_PLUGIN_MANIFEST.md    |    8 +
 .../00_SESSION_OVERVIEW.md                    |   36 +
 .../01_RESEARCH.md                            |   28 +
 .../02_SPECIFICATIONS.md                      |   43 +
 .../03_DAG_WBS.md                             |   25 +
 .../04_IMPLEMENTATION_STRATEGY.md             |   37 +
 .../05_KNOWN_ISSUES.md                        |   21 +
 .../06_TESTING_STRATEGY.md                    |   25 +
 electron/package-lock.json                    |   12 +-
 electron/package.json                         |    2 +-
 next.config.mjs                               |    4 +-
 open-sse/config/agyModels.ts                  |    4 +-
 open-sse/config/antigravityModelAliases.ts    |    6 +-
 open-sse/config/freeModelCatalog.data.ts      |    1 +
 open-sse/config/imageRegistry.ts              |   61 +-
 .../config/nvidiaHostedModels.snapshot.json   |   18 +
 open-sse/config/providers/index.ts            |    2 +
 .../config/providers/registry/agnes/index.ts  |   26 +
 .../config/providers/registry/auggie/index.ts |   54 +-
 .../config/providers/registry/claude/index.ts |    2 +-
 .../providers/registry/gemini/imageModels.ts  |   32 +
 .../providers/registry/opencode/go/index.ts   |   14 +-
 .../registry/stability-ai/imageModels.ts      |   76 +
 open-sse/executors/antigravity.ts             |   69 +-
 .../executors/antigravity/proFallbackChain.ts |  104 ++
 open-sse/executors/antigravity/sseCollect.ts  |   16 +-
 open-sse/executors/auggie.ts                  |  147 +-
 open-sse/executors/base/reasoningEffort.ts    |  124 +-
 open-sse/executors/default.ts                 |   27 +-
 open-sse/executors/duckduckgo-web.ts          |   65 +
 open-sse/executors/opencode.ts                |   38 +-
 open-sse/executors/theoldllm.ts               |  216 ++-
 open-sse/handlers/chatCore.ts                 |  147 +-
 .../chatCore/cavemanOutputAnalytics.ts        |    8 +-
 .../chatCore/compressionAnalyticsWrite.ts     |   61 +-
 .../handlers/chatCore/outputTokenBudget.ts    |  117 ++
 .../handlers/chatCore/streamingPipeline.ts    |   21 +-
 open-sse/handlers/imageGeneration.ts          |   12 +
 .../imageGeneration/providers/googleImagen.ts |  147 ++
 open-sse/mcp-server/README.md                 |   16 +-
 .../__tests__/toolSearch.catalog.test.ts      |    9 +-
 open-sse/mcp-server/schemas/ccrTools.ts       |  202 +++
 open-sse/mcp-server/schemas/index.ts          |   20 +
 open-sse/mcp-server/schemas/tools.ts          |    3 +
 open-sse/mcp-server/server.ts                 |    1 +
 open-sse/mcp-server/toolSearch/catalog.ts     |    5 +-
 open-sse/mcp-server/tools/compressionTools.ts |  259 ++-
 open-sse/package.json                         |    3 +-
 open-sse/services/accountFallback.ts          |   15 +-
 .../accountFallback/lockoutEviction.ts        |   53 +
 open-sse/services/claudeCodeCompatible.ts     |    1 +
 open-sse/services/combo.ts                    |   81 +-
 open-sse/services/combo/autoConfig.ts         |    4 +-
 open-sse/services/combo/comboPredicates.ts    |   44 +-
 open-sse/services/combo/comboStructure.ts     |   71 +-
 .../services/combo/knownContextOverflow.ts    |   97 ++
 .../services/combo/resolveAutoStrategy.ts     |   14 +-
 open-sse/services/combo/sessionStickiness.ts  |   34 +
 open-sse/services/combo/targetExhaustion.ts   |   17 +-
 open-sse/services/combo/targetSorters.ts      |   48 +-
 open-sse/services/combo/validateQuality.ts    |   26 +-
 open-sse/services/compression/bodyAdapter.ts  |   67 +-
 .../services/compression/engines/ccr/index.ts |  347 +++-
 .../compression/engines/ionizer/sample.ts     |   19 +-
 .../compression/engines/rtk/filterLoader.ts   |  167 +-
 .../compression/engines/rtk/filterSchema.ts   |    6 +
 .../compression/engines/rtk/lineFilter.ts     |   48 +
 .../engines/rtk/tomlCompatibility.ts          |  334 ++++
 .../engines/session-dedup/fuzzy.ts            |   11 +-
 open-sse/services/compression/liveZone.ts     |  397 +++++
 open-sse/services/compression/types.ts        |   13 +
 open-sse/services/responsesInputSanitizer.ts  |   26 +-
 .../translator/request/claude-to-gemini.ts    |    4 +-
 .../request/openai-responses/toResponses.ts   |   37 +
 open-sse/utils/diagnostics.ts                 |   22 +
 open-sse/utils/sseHeartbeat.ts                |   21 +
 open-sse/utils/stream.ts                      |   18 +
 package-lock.json                             | 1513 ++++++++---------
 package.json                                  |   10 +-
 scripts/check/check-nvidia-catalog-drift.ts   |  108 ++
 scripts/check/check-router-eval-regression.ts |  287 ++++
 scripts/router-eval/compare.ts                |  135 ++
 scripts/router-eval/index.ts                  |  483 ++++++
 scripts/router-eval/patch-compare.ts          |  315 ++++
 scripts/router-eval/search.ts                 |  439 +++++
 scripts/router-eval/trends.ts                 |  198 +++
 skills/omni-context-rtk/SKILL.md              |   11 +
 .../dashboard/HomeProviderTopologySection.tsx |   18 +-
 .../analytics/CompressionAnalyticsTab.tsx     |   10 +-
 src/app/(dashboard)/dashboard/combos/page.tsx |    5 +-
 .../context/rtk/RtkContextPageClient.tsx      |    8 +-
 .../context/rtk/RtkTomlImportCard.tsx         |  255 +++
 .../context/settings/CompressionPanel.tsx     |   57 +-
 .../dashboard/endpoint/EndpointPageClient.tsx |   56 +-
 .../__tests__/ApiEndpointsTab.test.tsx        |   16 +-
 .../__tests__/EndpointPageClient.test.tsx     |   52 +
 .../(dashboard)/dashboard/onboarding/page.tsx |   11 +-
 .../[id]/ProviderDetailPageClient.tsx         |    7 +-
 .../providers/[id]/__tests__/phase1f.test.tsx |    6 +-
 .../[id]/components/ConnectionRow.tsx         |   11 +
 .../[id]/components/ConnectionsListPanel.tsx  |   22 +
 .../ProviderQuotaVisibilityToggle.tsx         |   43 +
 .../components/__tests__/phase1d.test.tsx     |   37 +-
 .../[id]/hooks/useProviderConnections.ts      |    7 +-
 .../[id]/hooks/useProviderQuotaVisibility.ts  |   45 +
 .../settings/components/ProxyTab.tsx          |   16 +-
 .../settings/components/SidebarTab.tsx        |   65 +-
 .../settings/components/proxy/FreePoolTab.tsx |   84 +-
 .../dashboard/usage/components/EvalsTab.tsx   |    6 +-
 .../ProviderLimits/CodexResetCreditsModal.tsx |  277 +++
 .../components/ProviderLimits/QuotaCard.tsx   |   19 +-
 .../ProviderLimits/QuotaCardGrid.tsx          |    9 +-
 .../usage/components/ProviderLimits/index.tsx |   17 +-
 .../parts/QuotaCardExpanded.tsx               |   52 +-
 .../useCodexResetCreditRedemption.ts          |  207 ++-
 .../(dashboard)/home/ProviderQuotaWidget.tsx  |    3 +
 src/app/api/cli-tools/all-statuses/route.ts   |   17 +-
 src/app/api/context/rtk/import/route.ts       |   81 +
 src/app/api/issue-agent/runs/route.ts         |  141 ++
 src/app/api/network/info/route.ts             |    3 +-
 .../api/oauth/[provider]/[action]/route.ts    |    7 +-
 .../models/discovery/providerModelsConfig.ts  |   24 +
 src/app/api/providers/[id]/route.ts           |    7 +-
 .../agent-bridge/agents/[id]/dns/route.ts     |    4 +-
 src/app/api/usage/codex-reset-credit/route.ts |   65 +-
 src/app/api/v1/images/generations/route.ts    |    9 +-
 src/app/api/v1/models/catalog.ts              |   56 +-
 src/app/api/v1/models/catalogHelpers.ts       |   57 +-
 .../api/v1/provider-plugin-manifest/route.ts  |   42 +-
 src/app/globals.css                           |   22 +-
 src/i18n/messages/de.json                     |   69 +-
 src/i18n/messages/en.json                     |   43 +-
 src/i18n/messages/pt-BR.json                  |   43 +-
 src/instrumentation-node.ts                   |   58 +
 src/lib/cliTools/batchStatusCache.ts          |   24 +-
 src/lib/config/runtimeSettings.ts             |    9 +-
 src/lib/credentialHealth/scheduler.ts         |    9 +-
 src/lib/db/backup.ts                          |    8 +-
 src/lib/db/caseMapping.ts                     |   15 +-
 src/lib/db/cleanup.ts                         |  104 +-
 src/lib/db/compression.ts                     |    4 +
 src/lib/db/core.ts                            |   21 +-
 src/lib/db/databaseSettings.ts                |    3 +
 src/lib/db/migrationRunner.ts                 |    6 +-
 ...5_provider_connection_quota_visibility.sql |    5 +
 src/lib/db/providers.ts                       |  168 +-
 src/lib/db/proxies.ts                         |   83 +-
 src/lib/db/proxies/guards.ts                  |   81 +
 src/lib/db/schemaColumns.ts                   |    6 +
 src/lib/initCloudSync.ts                      |    9 +-
 src/lib/issueAgent/audit.ts                   |   43 +
 src/lib/issueAgent/execution.ts               |  114 ++
 src/lib/issueAgent/githubExport.ts            |   57 +
 src/lib/issueAgent/recordedTriage.ts          |  124 ++
 src/lib/localHealthCheck.ts                   |    9 +-
 src/lib/oauth/constants/oauth.ts              |    6 +-
 src/lib/oauth/providers/codebuddy-cn.ts       |    5 +-
 src/lib/providers/validation/webProvidersB.ts |   19 +-
 src/lib/quota/connectionRecovery.ts           |    9 +-
 src/lib/resilience/modelLockoutSettings.ts    |    7 +-
 src/lib/routerEval/index.ts                   |  428 +++++
 src/lib/tokenHealthCheck.ts                   |   43 +-
 src/lib/usage/codexResetCredits.ts            |  182 +-
 src/lib/usage/comboScoringInspector.ts        |  111 +-
 src/mitm/dns/provision.ts                     |  128 +-
 src/server/authz/routeGuard.ts                |    1 +
 src/server/ws/liveServer.ts                   |    6 +-
 src/shared/components/Sidebar.tsx             |   47 +-
 src/shared/components/flow/FlowCanvas.tsx     |    1 +
 src/shared/components/flow/edgeStyles.ts      |    4 +-
 src/shared/constants/mcpScopes.ts             |    6 +
 .../constants/providers/apikey/regional.ts    |   12 +
 src/shared/constants/sidebarVisibility.ts     |    5 +-
 .../constants/sidebarVisibility/sections.ts   |    1 -
 .../constants/sidebarVisibility/types.ts      |   10 +-
 .../__tests__/useDisplayBaseUrl.test.tsx      |  108 +-
 src/shared/hooks/cli/useToolBatchStatuses.ts  |   11 +-
 src/shared/hooks/index.ts                     |    7 +-
 src/shared/hooks/useDisplayBaseUrl.ts         |  122 +-
 src/shared/services/cliRuntime.ts             |    8 +-
 src/shared/types/utilization.ts               |   12 +-
 src/shared/utils/providerQuotaVisibility.ts   |   13 +
 src/shared/utils/sidebarExpansionState.ts     |   29 +
 src/shared/utils/testProcess.ts               |   44 +
 .../validation/compressionConfigSchemas.ts    |    1 +
 src/shared/validation/schemas/provider.ts     |    1 +
 src/sse/handlers/chat.ts                      |   15 +-
 src/sse/handlers/comboFailureLogging.ts       |   20 +
 src/types/databaseSettings.ts                 |    6 +
 stryker.conf.json                             |    1 +
 tests/fixtures/router-eval/baseline.ndjson    |    2 +
 tests/fixtures/router-eval/candidate.ndjson   |    2 +
 tests/integration/all-statuses-route.test.ts  |   44 +-
 tests/theoldllm-stress.test.ts                |   72 +-
 .../account-fallback-lockout-eviction.test.ts |  129 ++
 .../unit/agent-bridge-dns-params-7271.test.ts |   37 +
 .../agentrouter-models-discovery-7016.test.ts |   74 +
 tests/unit/agnes-provider.test.ts             |   70 +
 .../unit/agy-pro-fallback-chain-3786.test.ts  |  203 +++
 tests/unit/antigravity-model-aliases.test.ts  |   10 +-
 ...antigravity-server-side-tools-6914.test.ts |   43 +
 .../compression/rtk-toml-import-route.test.ts |  125 ++
 .../v1/provider-plugin-manifest-route.test.ts |   34 +-
 .../apikey-connection-health-check.test.ts    |   74 +
 tests/unit/auggie-executor.test.ts            |  170 +-
 .../auto-combo-context-advertising.test.ts    |    8 +-
 .../autoCombo/provider-family-combos.test.ts  |    7 +-
 .../base-executor-sanitize-effort.test.ts     |   35 +
 tests/unit/batch-status-cache.test.ts         |    8 +
 ...atcore-compression-analytics-write.test.ts |   20 +-
 tests/unit/chatcore-sanitization.test.ts      |    2 +-
 .../unit/chatcore-streaming-pipeline.test.ts  |   39 +-
 ...claude-context-1m-supported-models.test.ts |  153 ++
 ...-to-gemini-budget-tokens-zero-6813.test.ts |   42 +
 tests/unit/cli-runtime-detection.test.ts      |   56 +-
 tests/unit/clinepass-thinking-budget.test.ts  |   16 +-
 tests/unit/codex-reset-credits.test.ts        |  147 ++
 tests/unit/combo-auto-config-split.test.ts    |   17 +
 tests/unit/combo-breaker-429.test.ts          |   11 +
 .../unit/combo-context-window-filter.test.ts  |  204 ++-
 tests/unit/combo-failure-log-message.test.ts  |   29 +
 tests/unit/combo-least-used-account.test.ts   |   86 +
 .../combo-resolve-auto-strategy-split.test.ts |  136 ++
 tests/unit/combo-scoring-inspector.test.ts    |  114 +-
 ...bo-stickiness-responses-input-7270.test.ts |  235 +++
 .../combo/combo-target-exhaustion.test.ts     |   19 +
 tests/unit/compression/body-adapter.test.ts   |   86 +
 .../unit/compression/ccr-cross-tenant.test.ts |   18 +-
 .../compression/ccr-mcp-integration.test.ts   |  220 +++
 .../compression-preview-api.test.ts           |    4 +
 tests/unit/compression/db.test.ts             |   11 +
 tests/unit/compression/live-zone.test.ts      |  254 +++
 .../unit/compression/rtk-line-filter.test.ts  |   16 +
 .../rtk-toml-compatibility.test.ts            |  313 ++++
 tests/unit/db-core-temp-store-pragma.test.ts  |   21 +
 tests/unit/db-providers-crud.test.ts          |   95 ++
 tests/unit/db-schema-columns-split.test.ts    |   22 +
 ...uit-breaker-null-content-6999-7000.test.ts |  248 +++
 tests/unit/diagnostics.test.ts                |   23 +-
 tests/unit/executor-agy.test.ts               |   85 +
 tests/unit/executor-antigravity.test.ts       |    2 +-
 tests/unit/free-models.test.ts                |    4 +
 tests/unit/gemini-imagen-predict.test.ts      |   86 +
 .../unit/image-text-to-image-modality.test.ts |  145 ++
 ...instrumentation-warm-catalog-cache.test.ts |  162 ++
 tests/unit/issue-agent-audit.test.ts          |   31 +
 tests/unit/issue-agent-execution.test.ts      |   70 +
 tests/unit/issue-agent-github-export.test.ts  |   48 +
 .../unit/issue-agent-route-execution.test.ts  |  135 ++
 tests/unit/issue-agent-runner.test.ts         |   63 +
 tests/unit/issue-agent-runs-route.test.ts     |  285 ++++
 .../m365-web-token-extraction-7078.test.ts    |   48 +
 tests/unit/mcp-tool-count-dedup-6854.test.ts  |   11 +-
 .../mitm-dns-graceful-degrade-6127.test.ts    |  127 +-
 .../models-catalog-combo-metadata.test.ts     |  119 ++
 tests/unit/models-catalog-route.test.ts       |    2 +-
 tests/unit/node-runtime-support.test.ts       |    9 +
 tests/unit/nvidia-catalog-drift.test.ts       |   28 +
 .../unit/oauth-device-code-endpoints.test.ts  |   32 +
 ...uth-device-code-error-transparency.test.ts |   32 +
 .../openai-responses-request-split.test.ts    |   67 +
 .../opencode-go-effort-aliases-6922.test.ts   |  164 ++
 tests/unit/output-token-budget.test.ts        |   66 +
 ...ovider-connections-quota-threshold.test.ts |   23 +
 tests/unit/provider-limits-ui.test.ts         |   14 +
 tests/unit/provider-quota-visibility.test.ts  |   18 +
 .../unit/proxy-bulk-import-dedup-7594.test.ts |  105 ++
 .../responses-input-sanitizer-name.test.ts    |   58 +-
 tests/unit/route-guard-private-lan.test.ts    |    5 +
 tests/unit/router-eval-check.test.ts          |  484 ++++++
 tests/unit/router-eval-cli.test.ts            |  270 +++
 tests/unit/router-eval-compare.test.ts        |   83 +
 tests/unit/router-eval-e2e-chain.test.ts      |  135 ++
 tests/unit/router-eval-patch-compare.test.ts  |  312 ++++
 tests/unit/router-eval-search.test.ts         |  258 +++
 tests/unit/router-eval-trends.test.ts         |  128 ++
 tests/unit/router-eval.test.ts                |   92 +
 tests/unit/sseHeartbeat.test.ts               |   75 +
 ...stream-request-body-size-mark-7045.test.ts |  122 ++
 .../unit/telemetry-auto-cleanup-6848.test.ts  |  214 +++
 tests/unit/test-process-detection.test.ts     |   52 +
 tests/unit/theoldllm-provider-proxy.test.ts   |   60 +
 tests/unit/token-health-check-sweep.test.ts   |  146 ++
 tests/unit/ui/combos-page-smoke.test.tsx      |   12 +
 tests/unit/ui/edgeStyles.test.ts              |    6 +-
 tests/unit/ui/evals-tab-smoke.test.tsx        |   12 +
 tests/unit/ui/flowCanvas.test.tsx             |   13 +-
 tests/unit/{ => ui}/free-pool-tab.test.tsx    |   37 +-
 ...me-provider-topology-section-4606.test.tsx |   14 +-
 .../ui/onboarding-public-endpoint.test.tsx    |  107 ++
 tests/unit/ui/rtkTomlImportCard.test.tsx      |  179 ++
 .../unit/ui/sidebar-proxy-expansion.test.tsx  |   46 +
 tests/unit/ui/useToolBatchStatuses.test.tsx   |   11 +-
 332 files changed, 19674 insertions(+), 2135 deletions(-)
 create mode 100644 changelog.d/features/7318-router-eval-harness.md
 create mode 100644 changelog.d/features/provider-quota-connection-visibility.md
 create mode 100644 changelog.d/fixes/7032-auggie-model-ids-v032.md
 create mode 100644 changelog.d/fixes/7490-align-engines-node.md
 create mode 100644 changelog.d/fixes/7547-prefer-public-endpoint-url.md
 create mode 100644 changelog.d/fixes/pending-cli-service-detection.md
 create mode 100644 changelog.d/fixes/pending-react-flow-dark-theme.md
 create mode 100644 changelog.d/maintenance/7334-incident-response-runbook.md
 create mode 100644 changelog.d/maintenance/7336-perf-latency-budgets-doc.md
 create mode 100644 docs/INCIDENT_RESPONSE.md
 create mode 100644 docs/PERF_BUDGETS.md
 create mode 100644 docs/sessions/20260714-issue-agent-executable-triage/00_SESSION_OVERVIEW.md
 create mode 100644 docs/sessions/20260714-issue-agent-executable-triage/01_RESEARCH.md
 create mode 100644 docs/sessions/20260714-issue-agent-executable-triage/02_SPECIFICATIONS.md
 create mode 100644 docs/sessions/20260714-issue-agent-executable-triage/03_DAG_WBS.md
 create mode 100644 docs/sessions/20260714-issue-agent-executable-triage/04_IMPLEMENTATION_STRATEGY.md
 create mode 100644 docs/sessions/20260714-issue-agent-executable-triage/05_KNOWN_ISSUES.md
 create mode 100644 docs/sessions/20260714-issue-agent-executable-triage/06_TESTING_STRATEGY.md
 create mode 100644 open-sse/config/nvidiaHostedModels.snapshot.json
 create mode 100644 open-sse/config/providers/registry/agnes/index.ts
 create mode 100644 open-sse/config/providers/registry/gemini/imageModels.ts
 create mode 100644 open-sse/config/providers/registry/stability-ai/imageModels.ts
 create mode 100644 open-sse/executors/antigravity/proFallbackChain.ts
 create mode 100644 open-sse/handlers/chatCore/outputTokenBudget.ts
 create mode 100644 open-sse/handlers/imageGeneration/providers/googleImagen.ts
 create mode 100644 open-sse/mcp-server/schemas/ccrTools.ts
 create mode 100644 open-sse/services/accountFallback/lockoutEviction.ts
 create mode 100644 open-sse/services/combo/knownContextOverflow.ts
 create mode 100644 open-sse/services/compression/engines/rtk/tomlCompatibility.ts
 create mode 100644 open-sse/services/compression/liveZone.ts
 create mode 100644 scripts/check/check-nvidia-catalog-drift.ts
 create mode 100644 scripts/check/check-router-eval-regression.ts
 create mode 100644 scripts/router-eval/compare.ts
 create mode 100644 scripts/router-eval/index.ts
 create mode 100644 scripts/router-eval/patch-compare.ts
 create mode 100644 scripts/router-eval/search.ts
 create mode 100644 scripts/router-eval/trends.ts
 create mode 100644 src/app/(dashboard)/dashboard/context/rtk/RtkTomlImportCard.tsx
 create mode 100644 src/app/(dashboard)/dashboard/providers/[id]/components/ProviderQuotaVisibilityToggle.tsx
 create mode 100644 src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderQuotaVisibility.ts
 create mode 100644 src/app/(dashboard)/dashboard/usage/components/ProviderLimits/CodexResetCreditsModal.tsx
 create mode 100644 src/app/api/context/rtk/import/route.ts
 create mode 100644 src/app/api/issue-agent/runs/route.ts
 create mode 100644 src/lib/db/migrations/125_provider_connection_quota_visibility.sql
 create mode 100644 src/lib/db/proxies/guards.ts
 create mode 100644 src/lib/issueAgent/audit.ts
 create mode 100644 src/lib/issueAgent/execution.ts
 create mode 100644 src/lib/issueAgent/githubExport.ts
 create mode 100644 src/lib/issueAgent/recordedTriage.ts
 create mode 100644 src/lib/routerEval/index.ts
 create mode 100644 src/shared/utils/providerQuotaVisibility.ts
 create mode 100644 src/shared/utils/sidebarExpansionState.ts
 create mode 100644 src/shared/utils/testProcess.ts
 create mode 100644 src/sse/handlers/comboFailureLogging.ts
 create mode 100644 tests/fixtures/router-eval/baseline.ndjson
 create mode 100644 tests/fixtures/router-eval/candidate.ndjson
 create mode 100644 tests/unit/account-fallback-lockout-eviction.test.ts
 create mode 100644 tests/unit/agent-bridge-dns-params-7271.test.ts
 create mode 100644 tests/unit/agentrouter-models-discovery-7016.test.ts
 create mode 100644 tests/unit/agnes-provider.test.ts
 create mode 100644 tests/unit/antigravity-server-side-tools-6914.test.ts
 create mode 100644 tests/unit/api/compression/rtk-toml-import-route.test.ts
 create mode 100644 tests/unit/claude-context-1m-supported-models.test.ts
 create mode 100644 tests/unit/claude-to-gemini-budget-tokens-zero-6813.test.ts
 create mode 100644 tests/unit/combo-failure-log-message.test.ts
 create mode 100644 tests/unit/combo-least-used-account.test.ts
 create mode 100644 tests/unit/combo-stickiness-responses-input-7270.test.ts
 create mode 100644 tests/unit/compression/ccr-mcp-integration.test.ts
 create mode 100644 tests/unit/compression/live-zone.test.ts
 create mode 100644 tests/unit/compression/rtk-toml-compatibility.test.ts
 create mode 100644 tests/unit/db-core-temp-store-pragma.test.ts
 create mode 100644 tests/unit/ddg-circuit-breaker-null-content-6999-7000.test.ts
 create mode 100644 tests/unit/gemini-imagen-predict.test.ts
 create mode 100644 tests/unit/image-text-to-image-modality.test.ts
 create mode 100644 tests/unit/instrumentation-warm-catalog-cache.test.ts
 create mode 100644 tests/unit/issue-agent-audit.test.ts
 create mode 100644 tests/unit/issue-agent-execution.test.ts
 create mode 100644 tests/unit/issue-agent-github-export.test.ts
 create mode 100644 tests/unit/issue-agent-route-execution.test.ts
 create mode 100644 tests/unit/issue-agent-runner.test.ts
 create mode 100644 tests/unit/issue-agent-runs-route.test.ts
 create mode 100644 tests/unit/m365-web-token-extraction-7078.test.ts
 create mode 100644 tests/unit/models-catalog-combo-metadata.test.ts
 create mode 100644 tests/unit/nvidia-catalog-drift.test.ts
 create mode 100644 tests/unit/oauth-device-code-endpoints.test.ts
 create mode 100644 tests/unit/oauth-device-code-error-transparency.test.ts
 create mode 100644 tests/unit/opencode-go-effort-aliases-6922.test.ts
 create mode 100644 tests/unit/output-token-budget.test.ts
 create mode 100644 tests/unit/provider-quota-visibility.test.ts
 create mode 100644 tests/unit/proxy-bulk-import-dedup-7594.test.ts
 create mode 100644 tests/unit/router-eval-check.test.ts
 create mode 100644 tests/unit/router-eval-cli.test.ts
 create mode 100644 tests/unit/router-eval-compare.test.ts
 create mode 100644 tests/unit/router-eval-e2e-chain.test.ts
 create mode 100644 tests/unit/router-eval-patch-compare.test.ts
 create mode 100644 tests/unit/router-eval-search.test.ts
 create mode 100644 tests/unit/router-eval-trends.test.ts
 create mode 100644 tests/unit/router-eval.test.ts
 create mode 100644 tests/unit/sseHeartbeat.test.ts
 create mode 100644 tests/unit/stream-request-body-size-mark-7045.test.ts
 create mode 100644 tests/unit/telemetry-auto-cleanup-6848.test.ts
 create mode 100644 tests/unit/test-process-detection.test.ts
 create mode 100644 tests/unit/theoldllm-provider-proxy.test.ts
 create mode 100644 tests/unit/token-health-check-sweep.test.ts
 create mode 100644 tests/unit/ui/combos-page-smoke.test.tsx
 create mode 100644 tests/unit/ui/evals-tab-smoke.test.tsx
 rename tests/unit/{ => ui}/free-pool-tab.test.tsx (90%)
 create mode 100644 tests/unit/ui/onboarding-public-endpoint.test.tsx
 create mode 100644 tests/unit/ui/rtkTomlImportCard.test.tsx
 create mode 100644 tests/unit/ui/sidebar-proxy-expansion.test.tsx

diff --git a/.env.example b/.env.example
index b4b73bfd3a3..fdb38b67f7f 100644
--- a/.env.example
+++ b/.env.example
@@ -649,6 +649,10 @@ NEXT_PUBLIC_ENABLE_SOCKS5_PROXY=true
 # Legacy alias for OMNIROUTE_API_KEY.
 # ROUTER_API_KEY=
 
+# Enable the offline/local Issue Agent recorded-triage endpoint.
+# Used by: src/app/api/issue-agent/runs/route.ts. Default: disabled.
+# OMNIROUTE_ISSUE_AGENT_ENABLED=false
+
 # CLI remote-mode context/profile for `omniroute` commands (overrides the active
 # context in the local contexts store). Equivalent to the `--context <name>` flag.
 # Used by: bin/cli/program.mjs, bin/cli/api.mjs (remote mode).
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index d97161e470c..71c4d640b46 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -39,7 +39,7 @@ jobs:
         with:
           persist-credentials: false
           fetch-depth: 0
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
       - id: classify
@@ -82,7 +82,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -171,7 +171,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -280,7 +280,7 @@ jobs:
         with:
           persist-credentials: false
           fetch-depth: 0
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -388,7 +388,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -424,7 +424,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -452,7 +452,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -515,7 +515,7 @@ jobs:
         with:
           persist-credentials: false
           fetch-depth: 0
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
       - name: Fetch base branch
@@ -554,7 +554,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -602,7 +602,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -649,7 +649,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -710,7 +710,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -757,7 +757,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -802,7 +802,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -871,7 +871,7 @@ jobs:
       # (if-no-files-found: warn) — Sonar consumes the same file.
       - name: Upload coverage to Codecov (informational)
         if: always()
-        uses: codecov/codecov-action@04b047e8bb82a0c002c8312c1c880fbc6a999d45 # v5
+        uses: codecov/codecov-action@0fb7174895f61a3b6b78fc075e0cd60383518dac # v5
         with:
           files: coverage/lcov.info
           token: ${{ secrets.CODECOV_TOKEN }}
@@ -1045,7 +1045,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -1115,7 +1115,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -1138,7 +1138,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
diff --git a/.github/workflows/codeql.yml b/.github/workflows/codeql.yml
index e47f93b3f69..34dd99666aa 100644
--- a/.github/workflows/codeql.yml
+++ b/.github/workflows/codeql.yml
@@ -22,10 +22,10 @@ jobs:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
         with:
           persist-credentials: false
-      - uses: github/codeql-action/init@99df26d4f13ea111d4ec1a7dddef6063f76b97e9 # v4.37.0
+      - uses: github/codeql-action/init@7188fc363630916deb702c7fdcf4e481b751f97a # v4.37.1
         with:
           languages: javascript-typescript
           queries: security-extended
-      - uses: github/codeql-action/analyze@99df26d4f13ea111d4ec1a7dddef6063f76b97e9 # v4.37.0
+      - uses: github/codeql-action/analyze@7188fc363630916deb702c7fdcf4e481b751f97a # v4.37.1
         with:
           category: "/language:javascript-typescript"
diff --git a/.github/workflows/dast-smoke.yml b/.github/workflows/dast-smoke.yml
index e1b5c757e55..08a622fdbba 100644
--- a/.github/workflows/dast-smoke.yml
+++ b/.github/workflows/dast-smoke.yml
@@ -21,7 +21,7 @@ jobs:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
         with:
           persist-credentials: false
-      - uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
+      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7.0.0
         with:
           node-version: "24"
           cache: npm
diff --git a/.github/workflows/electron-release.yml b/.github/workflows/electron-release.yml
index b86cd50f69f..2a7fee0e9ae 100644
--- a/.github/workflows/electron-release.yml
+++ b/.github/workflows/electron-release.yml
@@ -88,7 +88,7 @@ jobs:
         with:
           persist-credentials: false
       - name: Setup Node
-        uses: actions/setup-node@v6
+        uses: actions/setup-node@v7
         with:
           node-version: 24
           cache: npm
diff --git a/.github/workflows/mutation-redundancy.yml b/.github/workflows/mutation-redundancy.yml
index e584bea4870..0960846c5fe 100644
--- a/.github/workflows/mutation-redundancy.yml
+++ b/.github/workflows/mutation-redundancy.yml
@@ -44,7 +44,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
diff --git a/.github/workflows/nightly-compat.yml b/.github/workflows/nightly-compat.yml
index 2abbb149d94..eabc2073b32 100644
--- a/.github/workflows/nightly-compat.yml
+++ b/.github/workflows/nightly-compat.yml
@@ -66,7 +66,7 @@ jobs:
         with:
           ref: ${{ needs.resolve-branch.outputs.target }}
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "26"
           cache: npm
@@ -93,7 +93,7 @@ jobs:
         with:
           ref: ${{ needs.resolve-branch.outputs.target }}
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ matrix.node }}
           cache: npm
diff --git a/.github/workflows/nightly-llm-security.yml b/.github/workflows/nightly-llm-security.yml
index f039690235b..6a9abac2948 100644
--- a/.github/workflows/nightly-llm-security.yml
+++ b/.github/workflows/nightly-llm-security.yml
@@ -15,7 +15,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with: { node-version: "24", cache: npm }
       - run: npm ci
       - name: Build CLI bundle
@@ -65,7 +65,7 @@ jobs:
         with:
           persist-credentials: false
         if: steps.gate.outputs.run == 'true'
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         if: steps.gate.outputs.run == 'true'
         with: { node-version: "24", cache: npm }
       - run: npm ci
diff --git a/.github/workflows/nightly-mutation.yml b/.github/workflows/nightly-mutation.yml
index 5dfbe345a7f..5ee377deb96 100644
--- a/.github/workflows/nightly-mutation.yml
+++ b/.github/workflows/nightly-mutation.yml
@@ -107,7 +107,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
@@ -151,7 +151,7 @@ jobs:
       - uses: actions/checkout@v6
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
       - name: Download all mutation reports
diff --git a/.github/workflows/nightly-property.yml b/.github/workflows/nightly-property.yml
index 776426573d4..2ef7a543aff 100644
--- a/.github/workflows/nightly-property.yml
+++ b/.github/workflows/nightly-property.yml
@@ -13,7 +13,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
diff --git a/.github/workflows/nightly-release-green.yml b/.github/workflows/nightly-release-green.yml
index 63d1b4a054f..b9b5ddc41ad 100644
--- a/.github/workflows/nightly-release-green.yml
+++ b/.github/workflows/nightly-release-green.yml
@@ -116,7 +116,7 @@ jobs:
           git checkout "$TARGET"
           git log -1 --oneline
 
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
@@ -228,7 +228,7 @@ jobs:
           fetch-depth: 0
           persist-credentials: false
 
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
diff --git a/.github/workflows/nightly-resilience.yml b/.github/workflows/nightly-resilience.yml
index af5a20e22da..f6adffeabf3 100644
--- a/.github/workflows/nightly-resilience.yml
+++ b/.github/workflows/nightly-resilience.yml
@@ -15,7 +15,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
@@ -29,7 +29,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
@@ -43,7 +43,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
@@ -94,7 +94,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "24"
           cache: npm
diff --git a/.github/workflows/nightly-schemathesis.yml b/.github/workflows/nightly-schemathesis.yml
index e430677638a..64adef8f50b 100644
--- a/.github/workflows/nightly-schemathesis.yml
+++ b/.github/workflows/nightly-schemathesis.yml
@@ -16,7 +16,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with: { node-version: "24", cache: npm }
       - run: npm ci
       - name: Build CLI bundle
diff --git a/.github/workflows/npm-publish.yml b/.github/workflows/npm-publish.yml
index 1edd3c09e08..692e6bd26c9 100644
--- a/.github/workflows/npm-publish.yml
+++ b/.github/workflows/npm-publish.yml
@@ -71,7 +71,7 @@ jobs:
           fetch-depth: 0
 
       - name: Setup Node.js
-        uses: actions/setup-node@v6
+        uses: actions/setup-node@v7
         with:
           node-version: ${{ env.NPM_PUBLISH_NODE_VERSION }}
           registry-url: https://registry.npmjs.org
@@ -265,7 +265,7 @@ jobs:
           # Full history needed for auto-bump: git diff against previous release tag
 
       - name: Setup Node.js
-        uses: actions/setup-node@v6
+        uses: actions/setup-node@v7
         with:
           node-version: ${{ env.NPM_PUBLISH_NODE_VERSION }}
           registry-url: https://registry.npmjs.org
diff --git a/.github/workflows/opencode-plugin-ci.yml b/.github/workflows/opencode-plugin-ci.yml
index 0925d4c7f8f..f1c99fe426c 100644
--- a/.github/workflows/opencode-plugin-ci.yml
+++ b/.github/workflows/opencode-plugin-ci.yml
@@ -35,7 +35,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ matrix.node }}
           cache: npm
@@ -52,7 +52,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "22"
           cache: npm
diff --git a/.github/workflows/opencode-provider-ci.yml b/.github/workflows/opencode-provider-ci.yml
index 52a206dd374..5fe7d1fe78d 100644
--- a/.github/workflows/opencode-provider-ci.yml
+++ b/.github/workflows/opencode-provider-ci.yml
@@ -35,7 +35,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ matrix.node }}
           cache: npm
@@ -51,7 +51,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: "20"
           cache: npm
diff --git a/.github/workflows/quality.yml b/.github/workflows/quality.yml
index e54c4ac4c25..87adbc6b0f4 100644
--- a/.github/workflows/quality.yml
+++ b/.github/workflows/quality.yml
@@ -36,7 +36,7 @@ jobs:
         with:
           persist-credentials: false
           fetch-depth: 0
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
       - id: classify
@@ -68,7 +68,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -99,7 +99,7 @@ jobs:
         with:
           fetch-depth: 0
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -225,7 +225,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -265,7 +265,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -299,7 +299,7 @@ jobs:
       - uses: actions/checkout@v7
         with:
           persist-credentials: false
-      - uses: actions/setup-node@v6
+      - uses: actions/setup-node@v7
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
@@ -343,7 +343,7 @@ jobs:
         with:
           fetch-depth: 0
           persist-credentials: false
-      - uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6
+      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v6
         with:
           node-version: ${{ env.CI_NODE_VERSION }}
           cache: npm
diff --git a/.github/workflows/wiki-sync.yml b/.github/workflows/wiki-sync.yml
index da22cccd8d1..9ef2cdee2d5 100644
--- a/.github/workflows/wiki-sync.yml
+++ b/.github/workflows/wiki-sync.yml
@@ -40,7 +40,7 @@ jobs:
         uses: actions/checkout@v7
 
       - name: Setup Node
-        uses: actions/setup-node@v6
+        uses: actions/setup-node@v7
         with:
           node-version: "24"
 
diff --git a/README.md b/README.md
index c6c656b921f..9559dd0a88a 100644
--- a/README.md
+++ b/README.md
@@ -892,7 +892,7 @@ Compression: aggressive (~50%) → double your free quota · Cost: $0/mo
 | OAuth token expired                       | Auto-refreshed; if stuck, delete + re-auth in Providers                            |
 | `unsupported_country_region_territory`    | Configure proxy in Settings → Proxy                                                |
 | Docker SQLite locks                       | Use `--stop-timeout 40` for clean WAL checkpoint                                   |
-| Node runtime errors                       | Use Node `>=22.0.0 <23` or `>=24.0.0 <27`                                          |
+| Node runtime errors                       | Use Node `>=22.22.2 <23` or `>=24.0.0 <27`                                         |
 
 🐛 **Reporting a bug?** Run `npm run system-info` and attach `system-info.txt`. 📖 [`docs/guides/TROUBLESHOOTING.md`](docs/guides/TROUBLESHOOTING.md)
 
@@ -936,7 +936,7 @@ Compression: aggressive (~50%) → double your free quota · Cost: $0/mo
 
 </div>
 
-- **Runtime**: Node.js 22.x or 24.x LTS (24 LTS recommended) — `>=22.0.0 <23 || >=24.0.0 <27`
+- **Runtime**: Node.js 22.x or 24.x LTS (24 LTS recommended) — `>=22.22.2 <23 || >=24.0.0 <27`
 - **Language**: TypeScript 6.0 — **100% TypeScript** across `src/` and `open-sse/` (zero `any` in core modules since v2.0)
 - **Framework**: Next.js 16 + React 19 + Tailwind CSS 4
 - **Database**: better-sqlite3 (SQLite) + LowDB (JSON legacy) — domain state, proxy logs, MCP audit, routing decisions, memory, skills
@@ -1251,7 +1251,7 @@ MIT License - see [LICENSE](LICENSE) for details.
 
 **[⬆ Back to top](#-omniroute)** · Built with ❤️ for the open-source AI community.
 
-<sub>OmniRoute v3.8.43 · Node ≥22.0.0 · MIT License · <a href="https://omniroute.online">omniroute.online</a></sub>
+<sub>OmniRoute v3.8.43 · Node ≥22.22.2 · MIT License · <a href="https://omniroute.online">omniroute.online</a></sub>
 
 </div>
 <!-- GitHub Discussions enabled for community Q&A -->
diff --git a/changelog.d/features/7318-router-eval-harness.md b/changelog.d/features/7318-router-eval-harness.md
new file mode 100644
index 00000000000..f88326e9e01
--- /dev/null
+++ b/changelog.d/features/7318-router-eval-harness.md
@@ -0,0 +1 @@
+- **feat(eval):** added a router-eval harness (`npm run eval:router`, `eval:router:compare`, `eval:router:patch-compare`, `eval:router:search`, `eval:router:trends`, `check:router-eval`) that replays routing decisions — from NDJSON corpora or the `usage_history`/`call_logs` SQLite tables — into an AIQ (success/latency/cost) score, compares baseline vs. candidate router configs with a retained-run regression gate, and ranks Pareto-optimal candidates; a sibling tool to the existing `eval:compression` harness (#7318 — thanks @KooshaPari).
diff --git a/changelog.d/features/provider-quota-connection-visibility.md b/changelog.d/features/provider-quota-connection-visibility.md
new file mode 100644
index 00000000000..46e18fd848b
--- /dev/null
+++ b/changelog.d/features/provider-quota-connection-visibility.md
@@ -0,0 +1,3 @@
+- Provider connections can now be shown or hidden individually on the Provider Quota page. The
+  visibility setting is available on each account in the provider detail view and does not affect
+  routing or account activation.
diff --git a/changelog.d/fixes/7032-auggie-model-ids-v032.md b/changelog.d/fixes/7032-auggie-model-ids-v032.md
new file mode 100644
index 00000000000..bec9929b7b2
--- /dev/null
+++ b/changelog.d/fixes/7032-auggie-model-ids-v032.md
@@ -0,0 +1 @@
+- **fix(providers):** Auggie (Augment CLI) model registry updated to the real v0.32.0 CLI model IDs, with a live `auggie model list` auto-discovery fallback for future renames; the old pre-v0.32.0 IDs (`claude-sonnet-4.6`, `claude-opus-4.6`, `claude-haiku-4.5`, `gemini-3.1-pro`, `gemini-3.0-flash`, the `gpt-5.4`/`gpt-5.5` high/medium variants) are a **breaking rename**, but a backward-compat alias map in `resolveAuggieModel()` transparently remaps them to their v0.32.0 equivalents so existing saved combos keep working without any manual update (#7032 — thanks @oyi77).
diff --git a/changelog.d/fixes/7490-align-engines-node.md b/changelog.d/fixes/7490-align-engines-node.md
new file mode 100644
index 00000000000..2b6839f7746
--- /dev/null
+++ b/changelog.d/fixes/7490-align-engines-node.md
@@ -0,0 +1 @@
+- fix(build): align `engines.node` supported range (>=22.22.2 <23 || >=24.0.0 <27) across package.json, lockfile, README engine references and the node-runtime support test, so install-time engine checks match the actually-tested runtimes (#7446)
diff --git a/changelog.d/fixes/7547-prefer-public-endpoint-url.md b/changelog.d/fixes/7547-prefer-public-endpoint-url.md
new file mode 100644
index 00000000000..387f0d09a1a
--- /dev/null
+++ b/changelog.d/fixes/7547-prefer-public-endpoint-url.md
@@ -0,0 +1 @@
+- **fix(dashboard):** Public and managed tunnel endpoints now take precedence over loopback URLs in dashboard setup and copyable API configuration ([#7547](https://github.com/diegosouzapw/OmniRoute/pull/7547)) — thanks @nguyenha935
diff --git a/changelog.d/fixes/pending-cli-service-detection.md b/changelog.d/fixes/pending-cli-service-detection.md
new file mode 100644
index 00000000000..f9752264d4d
--- /dev/null
+++ b/changelog.d/fixes/pending-cli-service-detection.md
@@ -0,0 +1 @@
+- **fix(cli):** CLI detection now refreshes stale cached results, reports discovered versions, and checks the Continue `cn` binary instead of assuming it is installed.
diff --git a/changelog.d/fixes/pending-react-flow-dark-theme.md b/changelog.d/fixes/pending-react-flow-dark-theme.md
new file mode 100644
index 00000000000..1e177ee7359
--- /dev/null
+++ b/changelog.d/fixes/pending-react-flow-dark-theme.md
@@ -0,0 +1 @@
+- **fix(ui):** Theme React Flow controls correctly in dark mode, improve idle connector contrast, and localize the provider topology legend.
diff --git a/changelog.d/maintenance/7334-incident-response-runbook.md b/changelog.d/maintenance/7334-incident-response-runbook.md
new file mode 100644
index 00000000000..32a2286cffa
--- /dev/null
+++ b/changelog.d/maintenance/7334-incident-response-runbook.md
@@ -0,0 +1 @@
+- **docs:** Add `docs/INCIDENT_RESPONSE.md` — a non-security incident-response runbook (severity ladder, first-15-minutes checklist, and per-failure-mode mitigation steps for provider outages, latency regressions, and auth/data-layer incidents) ([#7334](https://github.com/diegosouzapw/OmniRoute/pull/7334)) — thanks @KooshaPari
diff --git a/changelog.d/maintenance/7336-perf-latency-budgets-doc.md b/changelog.d/maintenance/7336-perf-latency-budgets-doc.md
new file mode 100644
index 00000000000..a74a2a670a0
--- /dev/null
+++ b/changelog.d/maintenance/7336-perf-latency-budgets-doc.md
@@ -0,0 +1 @@
+- **docs:** Add `docs/PERF_BUDGETS.md` — per-endpoint p50/p95/p99 latency, throughput, resource, and cold-start budget reference targets ([#7336](https://github.com/diegosouzapw/OmniRoute/pull/7336)) — thanks @KooshaPari
diff --git a/config/quality/complexity-baseline.json b/config/quality/complexity-baseline.json
index 6917dbb5e41..262870277a8 100644
--- a/config/quality/complexity-baseline.json
+++ b/config/quality/complexity-baseline.json
@@ -1,6 +1,7 @@
 {
   "_comment": "Catraca de complexidade (check-complexity.mjs, ESLint core rules complexity>=15 e max-lines-per-function>80 sobre src+open-sse+electron+bin via eslint.complexity.config.mjs). Conta total de violacoes; so pode cair. --update ratcheta.",
-  "count": 2058,
+  "count": 2059,
+  "_rebaseline_2026_07_18_pr7360_quota_visibility_resync": "2058->2059 (+1 vs recorded ceiling; measured 2056 fresh on release tip cab9e5f0c alone, so this ceiling still carries 2 units of un-banked slack from prior shrinkage — real regression from this merge is 2056->2059, +3). PR #7360 (JxnLexn) release-resync: merging origin/release/v3.8.49 to resolve the 3-file conflict (ConnectionRow.tsx/ConnectionsListPanel.tsx/useProviderConnections.ts) unions two already-compliant features in the same already-oversized god-component: release's confirm-delete-account wiring (#7361) and this PR's per-connection quota-visibility wiring. Diffed release-tip-only vs merged violation lists (scripts dumped via getComplexityEslintReport): most entries are the SAME pre-existing violations shifted a few lines (ConnectionRow/getStatusPresentation/inferErrorType — no count change) or marginally bigger (ConnectionRow function complexity 85->86, ConnectionsListPanel function 498->510 lines) from the two ConnectionRow call sites each gaining both PRs' multi-line JSX props. The 2 genuinely NEW crossings are the 'no tag' and 'tagged groups' .map() render callbacks in ConnectionsListPanel.tsx (83 and 85 lines, was <=80 on both parents individually) tipping over 80 lines specifically because both PRs' props land on the same call sites. No new logic was written during the resync itself (only import-statement unions); the growth is inherent to combining the two already-reviewed feature branches. Structural shrink tracked in #3501. Tighten via --update next cycle (true floor is 2056, not 2058).",
   "_rebaseline_2026_07_17_v3849_ownerprs_providers": "2056->2058 (+2). v3.8.49 owner-PR merge campaign own-growth: the new provider handlers/dispatch branches merged this cycle (freetheai/felo/notion/segmind/deepinfra/novita/msdesigner image+video handlers, each adding a format-dispatch guard) pushed cyclomatic violations 2056->2058. Fast-gates PR->release do not run the complexity ratchet, so this surfaced only on re-sync. Spread across the new leaf handlers (not a single extractable function); measured on the release tip. Structural shrink tracked in #3501.",
   "_rebaseline_2026_07_10_v3847_merge_burst": "2053->2054 (+1). Drift herdado do merge burst do dia em release/v3.8.47 (campanha /implement-prs: ~36 PRs mergeados — órfãos, features do dono, ports). O check:complexity NÃO roda no fast-path PR->release, então o ramo acumulou o +1 sem rebaselinar (mesma família de todos os rebaselines abaixo). Trust-but-verify: medido 2054 no tip da release pós-burst; a única função flagada nova é pré-existente (getResolvedModelCapabilities em modelCapabilities.ts, já >teto antes de #6714). Nenhum PR órfão/feature introduz violação NOVA — os fixes deste ciclo são complexity-net-zero. Rebaseline aprovado pelo dono (2026-07-10) para destravar o FQG dos ~7 órfãos verdes-exceto-complexity. Tighten via --update next cycle.",
   "_rebaseline_2026_07_10_gcf_v3_2": "2054->2056 (+2). PR feat/headroom-gcf-v3.2-nested-flattening: own growth from re-vendoring the GCF (Headroom) codec to spec v3.2 (nested flattening). The 2 new over-threshold functions are the v3.2 `>`-path flatten/unflatten walk in the vendored generic-profile encode/decode paths (open-sse/services/compression/engines/headroom/gcf/{generic,decode_generic}.ts). This is imported third-party code kept byte-faithful to upstream gcf-typescript, not extractable without diverging from the vendored source; local measures 2055 on the merged tree; frozen at 2056 = the base's CI-observed 2054 + this PR's 2 new functions, matching the documented local-vs-CI off-by-one convention (see _rebaseline_2026_07_02_v3844_ci_observed) so the GitHub runner stays green. Round-trip guarded by tests/unit/compression/headroom-smartcrusher.test.ts (deep-nested case). Structural shrink belongs upstream in gcf, not here.",
diff --git a/config/quality/dependency-allowlist.json b/config/quality/dependency-allowlist.json
index 5ecfb4329a5..6c7e2e3f452 100644
--- a/config/quality/dependency-allowlist.json
+++ b/config/quality/dependency-allowlist.json
@@ -114,6 +114,7 @@
     "safe-regex",
     "selfsigned",
     "size-limit",
+    "smol-toml",
     "socks",
     "sql.js",
     "sqlite-vec",
diff --git a/config/quality/file-size-baseline.json b/config/quality/file-size-baseline.json
index 46a56405b99..0c7d6db2a5f 100644
--- a/config/quality/file-size-baseline.json
+++ b/config/quality/file-size-baseline.json
@@ -1,4 +1,5 @@
 {
+  "_rebaseline_2026_07_18_pr7360_quota_visibility_resync": "PR #7360 (JxnLexn, per-connection Provider Quota visibility) release-resync: merging origin/release/v3.8.49 into the PR's own branch to resolve its 3-file conflict (ConnectionRow.tsx/ConnectionsListPanel.tsx/useProviderConnections.ts import unions) surfaced ProviderDetailPageClient.tsx 785->791 (measured via check-file-size.mjs's split('\\n').length; wc -l reports 790). Both parents were individually within the 786 cap: release tip alone measures 785 (its own confirm-delete-account #7361 wiring), and the PR's own pre-merge tip also measured 785 (its own quota-visibility wiring) — the two independent, already-cohesive features simply sum past the frozen ceiling once unioned in this same already-oversized god-component (tracked for decomposition under #3501). No new logic was written during the resync; the only edits were import-statement unions. Frozen raised 786->791.",
   "_rebaseline_2026_07_18_pr7653_chat_tracker_import": "PR #7653 merge-interaction growth: release moved chat.ts to its 1796 cap while this PR adds the single side-effect import 'quotaTrackersBatch.ts' (line 130) — chat.ts IS the canonical quota-fetcher registration point (codex/bailian/deepseek/openrouter/opencode/generic all import+register there), so the +1 is irreducible call-site wiring. 1796->1797. Covered by tests/unit/{agentrouter,v0,freemodel}-quota-fetcher.test.ts.",
   "_rebaseline_2026_07_17_pr7653_agentrouter_console_fields": "PR #7653 own growth (missing acceptance criterion: the AgentRouter quota tracker (#6850) read providerSpecificData.consoleApiKey/newApiUserId but neither field had dashboard UI for provider agentrouter — consoleApiKey was gated to bailian-coding-plan only and newApiUserId had zero UI). AddApiKeyModal.tsx 961->967 (+6) and EditConnectionModal.tsx 1278->1286 (+8) = import + a single <AgentrouterConsoleFields .../> render call plus the newApiUserId formData init field. The actual Input rendering (both consoleApiKey reuse + the new newApiUserId field) was EXTRACTED into a new leaf src/app/(dashboard)/dashboard/providers/[id]/components/modals/AgentrouterConsoleFields.tsx (48 LOC, <cap), mirroring the QuotaScrapingFields.tsx / GlmTeamQuotaFields.tsx precedent (#6351) so the frozen modals only carry the irreducible call-site wiring. Persist logic lives in connectionProviderSpecificData.ts (not frozen). Covered by tests/unit/agentrouter-connection-modal-fields.test.ts.",
   "_rebaseline_2026_07_17_v3849_6842_free_window_wiring": "PR #7651 (openrouter :free-window quota tracking) follow-up: the counter shipped built but never wired into the request pipeline, so combos kept spending guaranteed-429 requests on exhausted free-tier targets. Own growth: src/sse/services/auth.ts 2461->2462 (+1, irreducible at the existing model-aware preflight chokepoint — the `provider === \"codex\"` check that forwards requestedModel into the connection arg is extended to also cover `openrouter`, one added boolean + a doc comment, offset to a single net line by dropping the now-redundant inline condition). Enforcement itself lives in open-sse/services/openrouterQuotaFetcher.ts (not frozen) and the dispatch-time record/correct hooks live in open-sse/executors/base.ts (not frozen). Covered by tests/unit/openrouter-free-window-wiring-6842.test.ts.",
@@ -142,6 +143,7 @@
   "_rebaseline_2026_06_20_1449_1444_test_route": "Re-baseline providers test route.ts 842->887: combined growth of sibling fixes #1449 (bound OAuth connection-test probe with a timeout) + #1444 (label a deactivated account distinctly from a revoked token), both at the same connection-test chokepoint. Cohesive route handler; not extractable without hiding the test flow.",
   "_rebaseline_2026_06_20_1409_1294_models": "Re-baseline src/lib/db/models.ts 1184->1221: combined growth of sibling fixes #1409 (cascade-delete orphaned model aliases when a provider is removed) + #1294 (persist max_input_tokens/max_output_tokens on custom models), both adding CRUD at the existing models domain module. Cohesive db module; not extractable.",
   "_rebaseline_2026_06_20_4389_thinking_toolchoice": "Re-baseline base.ts 1387->1399 (#4389): tool_choice-forced thinking guard at the existing Claude wire-image injection chokepoint (effThinking gate avoids the Anthropic 400 when tool_choice forces a tool). Cohesive guard; structural shrink tracked in #3501.",
+  "_rebaseline_2026_07_18_6979_codex_test": "PR #6979 own growth: executor-codex.test.ts 1340->1347 (+7 = generalized ensureThinkingBudget assertion added to the existing codex thinking-budget cases). antigravity-test bump 942->977 REVERTED here: #7408's test split dropped that file to 888, so this PR's +35 fits under the original 942 frozen cap.",
   "cap": 800,
   "frozen": {
     "_rebaseline_2026_07_02_5816_qoder": "PR #5816 (@AgentKiller45, qoder PAT via qodercli): qoderCli.ts 666->989, new-above-cap frozen (owner-approved baseline freeze). The growth is the legitimate PAT job-token exchange + quota parsing CLI transport (the pure-JS Cosy path 500'd on every PAT request); extracting the spawn/parse helpers now would just add indirection to a contributor PR mid-merge. Test frozen also raised for this PR's coverage growth: providers-page-utils.test.ts 1052->1092. Additionally clears an inherited base-red from the already-merged #5933 (codex json_schema->text.format): translator-openai-responses-req.test.ts 1097->1172 (+75 regression tests, no offending branch left). All remain frozen (cannot grow further); release captain's rebaseline-at-release supersedes.",
@@ -198,7 +200,7 @@
     "open-sse/translator/request/openai-to-kiro.ts": 912,
     "open-sse/translator/response/openai-responses.ts": 1092,
     "open-sse/utils/cursorAgentProtobuf.ts": 1521,
-    "open-sse/utils/stream.ts": 2796,
+    "open-sse/utils/stream.ts": 2814,
     "src/app/(dashboard)/dashboard/HomePageClient.tsx": 1385,
     "src/app/(dashboard)/dashboard/analytics/ComboHealthTab.tsx": 1028,
     "src/app/(dashboard)/dashboard/api-manager/ApiManagerPageClient.tsx": 3120,
@@ -206,13 +208,14 @@
     "src/app/(dashboard)/dashboard/cache/page.tsx": 845,
     "src/app/(dashboard)/dashboard/cli-code/components/CodexToolCard.tsx": 900,
     "src/app/(dashboard)/dashboard/cloud-agents/page.tsx": 922,
-    "src/app/(dashboard)/dashboard/combos/page.tsx": 4655,
+    "_rebaseline_2026_07_15_7070_combos_memo": "PR #7070 (perf/p1-memo) own growth: src/app/(dashboard)/dashboard/combos/page.tsx 4655->4656 (+1 = React.memo wrapping of ComboCard). Covered by tests/unit/ui/combos-page-smoke.test.tsx.",
+    "src/app/(dashboard)/dashboard/combos/page.tsx": 4656,
     "src/app/(dashboard)/dashboard/costs/CostOverviewTab.tsx": 1495,
     "src/app/(dashboard)/dashboard/costs/quota-share/components/PoolWizard.tsx": 1007,
     "src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.tsx": 2612,
     "src/app/(dashboard)/dashboard/health/page.tsx": 1091,
     "src/app/(dashboard)/dashboard/playground/components/tabs/ApiTab.tsx": 847,
-    "src/app/(dashboard)/dashboard/providers/[id]/ProviderDetailPageClient.tsx": 786,
+    "src/app/(dashboard)/dashboard/providers/[id]/ProviderDetailPageClient.tsx": 791,
     "src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionRow.tsx": 942,
     "src/app/(dashboard)/dashboard/providers/[id]/components/modals/AddApiKeyModal.tsx": 967,
     "src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx": 1286,
@@ -314,7 +317,7 @@
     "tests/unit/db-settings-crud.test.ts": 941,
     "tests/unit/deepseek-web.test.ts": 1092,
     "tests/unit/executor-antigravity.test.ts": 942,
-    "tests/unit/executor-codex.test.ts": 1340,
+    "tests/unit/executor-codex.test.ts": 1347,
     "tests/unit/executor-default-base.test.ts": 1523,
     "tests/unit/grok-web.test.ts": 2437,
     "tests/unit/image-generation-handler.test.ts": 2019,
@@ -408,5 +411,6 @@
   "_rebaseline_2026_07_07_6534_chirag": "PR #6534 (@chirag127) own growth: open-sse/services/compression/strategySelector.ts ->1025. Owner-approved rebaseline. Frozen.",
   "_rebaseline_2026_07_08_6556_omniglyph_mode": "PR #6556 (omniglyph engine) own growth: open-sse/services/compression/strategySelector.ts 1025->1043 (+18 at the existing mode-dispatch chokepoints). Two single-mode branches (sync no-op + async resolve via the engine registry, mirroring the rtk single-mode pattern, B-MODE-ENGINE-DECOUPLE) plus the optional providerTransport field threaded through the three options types (gates transport-sensitive engines). The engine itself lives in engines/omniglyphAdapter.ts (<cap); the residual is cohesive dispatch/type wiring not extractable without hiding the dispatch boundary, mirroring the prior compression rebaselines (#5243/#5286/phase4b/#6534). Covered by tests/unit/compression/omniglyph-*.test.ts (22). Structural shrink of this file tracked in #3501.",
   "_rebaseline_2026_07_07_6546_chirag": "PR #6546 (@chirag127) own growth: src/sse/handlers/chatHelpers.ts ->876. Owner-approved rebaseline. Frozen.",
-  "_rebaseline_2026_07_07_6525_chirag_image_guard": "PR #6525 (@chirag127, #6457) own growth: chat.ts ->1778 (reject image-only models on /v1/chat/completions; stacks on #6515). Owner-approved. Frozen."
+  "_rebaseline_2026_07_07_6525_chirag_image_guard": "PR #6525 (@chirag127, #6457) own growth: chat.ts ->1778 (reject image-only models on /v1/chat/completions; stacks on #6515). Owner-approved. Frozen.",
+  "_rebaseline_2026_07_15_7045_perf_instrumentation": "PR #7045 (@oyi77) own growth: open-sse/utils/stream.ts 2796->2814 (+18) from performance.mark/measure instrumentation around the SSE dispatch chokepoint (b48ba21c4), a TextEncoder hoisting fix to avoid a per-chunk allocation on the hot path (c35e8a9b4), and clearing the fixed-name \"omni-request-body-size\" mark immediately after creation (babysit fix, addressing a review-flagged unbounded-growth leak in Node's global performance timeline). Cohesive wiring at the existing stream-dispatch chokepoint; not extractable. Covered by tests/unit/chatcore-streaming-pipeline.test.ts + tests/unit/stream-request-body-size-mark-7045.test.ts."
 }
diff --git a/config/quality/test-discovery-baseline.json b/config/quality/test-discovery-baseline.json
index baca8262965..e6cb0d2a49f 100644
--- a/config/quality/test-discovery-baseline.json
+++ b/config/quality/test-discovery-baseline.json
@@ -42,7 +42,6 @@
     "tests/unit/dashboard/batch/list-regression.test.tsx",
     "tests/unit/dashboard/batch/sanitization.test.tsx",
     "tests/unit/free-budget-card.test.tsx",
-    "tests/unit/free-pool-tab.test.tsx",
     "tests/unit/guardrails/visionBridgeRouter.test.tsx",
     "tests/unit/omni-skills-page.test.tsx",
     "tests/unit/shared-clipboard.test.tsx",
diff --git a/docs/INCIDENT_RESPONSE.md b/docs/INCIDENT_RESPONSE.md
new file mode 100644
index 00000000000..95cada8e84b
--- /dev/null
+++ b/docs/INCIDENT_RESPONSE.md
@@ -0,0 +1,194 @@
+# Incident Response Runbook — OmniRoute (2026-06-18)
+
+**Status**: Authoritative. The 71-pillar audit (L61) references this doc
+for the `Obs > 2.00` gate.
+**Owner**: observability-circle (lead: security-circle lead).
+**SLOs**: see `docs/PERF_BUDGETS.md` § 1 (top-level SLOs) and
+`ops/slos.yaml` (machine-readable form, generated by the Bifrost team).
+**Disclosure policy**: see `SECURITY.md` (vulnerability disclosure only,
+separate flow).
+
+This runbook is the operational playbook for **non-security** incidents:
+outages, latency regressions, error-budget burn, and provider-side
+failures. Vulnerability disclosure stays on `SECURITY.md`; do not route
+those through this runbook.
+
+---
+
+## 1. Severity ladder
+
+| Sev | Definition | Examples | Page on | Resolve by |
+|---|---|---|---|---|
+| **SEV-1** | User-visible outage; > 50 % of requests failing or > 2x SLO breach for 5 min. | Cluster down; auth layer broken; 5xx flood. | On-call P0 (immediate) | 4 h |
+| **SEV-2** | Significant degradation; 1.5–2x SLO breach for 15 min, or single-tenant impact. | Single provider down; p95 > 1.5x budget; rate-limit runaway. | On-call P1 (15 min) | 24 h |
+| **SEV-3** | Latent bug or near-miss; no current user impact but error budget at risk. | Memory leak trending up; circuit breaker tripping on one provider. | Slack `#omniroute-ops` (next standup) | 7 d |
+| **SEV-4** | Cosmetic / informational. | Log line noise; non-binding UI glitch. | Next weekly review | Next refactor cycle |
+
+**Burn-rate escalation** (per `docs/PERF_BUDGETS.md` § 1): 6x for 5 min
+is SEV-1; 2x for 1 h is SEV-2; sustained < 1x for 7 d demotes to SEV-3.
+
+---
+
+## 2. Detection sources
+
+| Source | Signal | Routing |
+|---|---|---|
+| Prometheus (`/metrics`) | Counter deltas (5xx, latency) | Alertmanager → PagerDuty |
+| Grafana SLO dashboards | SLO burn-rate panels | Slack `#omniroute-ops` |
+| Uptime probe (`/api/health/ping`) | 3 consecutive failures from 3 regions | Alertmanager → PagerDuty |
+| Dependabot | New CVE in dependency | GitHub issue + Slack `#security` |
+| User report (support@) | Manual triage | Slack `#omniroute-triage` |
+| Error budget burn alert | `slo_burn_rate > threshold` | Alertmanager |
+
+Prometheus and Alertmanager are configured in the deploy repo (see
+`docs/operations/DEPLOY.md` once published; currently inline in
+`docker-compose.prod.yml`).
+
+---
+
+## 3. First-15-minutes checklist
+
+When paged, the on-call engineer runs this checklist verbatim. **Do
+not** skip steps; each is timed.
+
+1. **0:00** — Acknowledge the page in PagerDuty. Stops the escalation
+   timer and notifies the secondary.
+2. **0:02** — Open the [SLO dashboard][dash] and the [incident
+   channel][chan] (`#inc-YYYY-MM-DD-slug`). Post a single-line ack
+   with the alert name and the time.
+3. **0:05** — Classify severity per § 1. If SEV-1 or SEV-2, declare
+   the incident in the channel and tag `@incident-commander`.
+4. **0:08** — Capture the alert payload, the most recent deploy SHA,
+   and the top 5 slow / erroring endpoints. Post to the channel.
+5. **0:12** — Decide: **mitigate first, root-cause later**. Choose
+   one of:
+   - **Roll back** to the last green deploy (`bin/rollback.sh vX.Y.Z`).
+   - **Failover** to the healthy replicas (Caddy LB removes the bad
+     replica automatically; verify with `curl /api/health/ping`).
+   - **Disable** the broken connection(s) via `PUT /api/providers/{connectionId}`
+     with body `{ "isActive": false }` (per-connection toggle, safe by
+     default; repeat per key/account — see § 4.1).
+6. **0:15** — Post the chosen mitigation in the channel. If the page
+   is still firing after 5 more minutes, escalate to the secondary.
+
+[chan]: TBD — set to your team's incident-chat channel (e.g. a Discord/Slack `#inc-*` channel); not provisioned by this repo.
+[dash]: TBD — set to your Grafana/observability dashboard URL; not provisioned by this repo.
+
+---
+
+## 4. Mitigation runbooks (per failure mode)
+
+### 4.1 Provider outage (single provider down)
+
+1. `PUT /api/providers/{connectionId}` with body `{ "isActive": false }` —
+   deactivates that connection; combo routing and account selection skip it
+   on the next request (`src/app/api/providers/[id]/route.ts`). There is no
+   single whole-provider kill switch — if the provider has more than one
+   key/account, repeat per connection, or let the automatic provider circuit
+   breaker trip on its own (`src/shared/utils/circuitBreaker.ts`,
+   `domain_circuit_breakers` table; see `docs/architecture/RESILIENCE_GUIDE.md`).
+2. Verify p95 returns to budget within 5 min.
+3. If all connections for a model are down, apply the same `isActive: false`
+   toggle to every connection offering that model — there is no separate
+   per-model disable endpoint. Combo routing's automatic Model Lockout
+   (`open-sse/services/accountFallback.ts`; see
+   `docs/architecture/RESILIENCE_GUIDE.md`) also skips a model that keeps
+   erroring, without manual action.
+4. Update the status page (if one is configured — see § 5) with a banner if
+   the outage exceeds 15 min.
+
+### 4.2 Cluster-wide latency regression
+
+1. Check the most recent deploy (`/api/version` returns the SHA).
+2. If p95 doubled vs the 7-day baseline, **roll back** to the prior
+   SHA via `bin/rollback.sh`.
+3. If the regression is provider-side, see § 4.1.
+
+### 4.3 Auth layer broken (5xx on /v1/responses for all keys)
+
+1. Check the authz-inventory endpoint:
+   `curl https://api.omniroute.dev/api/settings/authz-inventory | jq`.
+   It returns a route-tier inventory (`tiers`, `bypassEnabled`,
+   `bypassPrefixes`, `spawnCapablePrefixes`, `cors` — see
+   `src/app/api/settings/authz-inventory/route.ts`); there is no
+   `policies_active` field. A non-200 response, or a `tiers` array that
+   fails to populate, means the settings/DB layer the auth pipeline reads
+   from is down — not just a single bad key.
+2. If the endpoint itself errors or returns malformed data, restore the
+   settings store from the last good backup (`bin/restore-policies.sh <sha>`).
+3. If the endpoint is healthy but requests still 5xx for every key, verify
+   `JWT_SECRET` / `API_KEY_SECRET` are set and unchanged for this deploy,
+   and that `isValidApiKey` (`src/sse/services/auth.ts`) can reach the DB.
+4. Roll back if the cause is unclear.
+
+### 4.4 Data-layer incident (sqlite corruption, audit log gap)
+
+1. **Stop the cluster** (`docker compose -f docker-compose.prod.yml
+   stop`) — preventing further writes is more important than uptime.
+2. Snapshot the data volume (`bin/snapshot-data.sh`).
+3. Open a SEV-1; this is data-loss territory. Page the data-team.
+4. Restore from the last verified backup (see `docs/BACKUP.md` once
+   published; currently the runbook is `bin/restore-data.sh <sha>`).
+
+### 4.5 Security incident (vulnerability disclosure)
+
+**Stop.** This is the `SECURITY.md` path, not this runbook. Page the
+security on-call (`@security-team`); do not post details to
+`#omniroute-ops`.
+
+---
+
+## 5. Communication
+
+| Audience | Channel | Cadence | Owner |
+|---|---|---|---|
+| Engineering | `#inc-YYYY-MM-DD-slug` | Real-time | Incident commander |
+| Status page | TBD — not provisioned by this repo | Every 30 min during SEV-1/2 | On-call |
+| Customers (email) | TBD — set your announcement list/address | At SEV-1 start + resolution | Comms lead |
+| Upstream providers | Direct contact | At SEV-1 start | Vendor mgmt |
+| Postmortem | `docs/postmortem/YYYY-MM-DD-slug.md` | Within 5 business days | Incident commander |
+
+Postmortem template is at `docs/postmortem/TEMPLATE.md` (forthcoming; no
+dedicated ADR covers it yet — once written, register it in
+`docs/architecture/cluster-decisions.md` following this repo's 71-pillar/ADR
+numbering convention, e.g. ADR-041 there).
+
+---
+
+## 6. On-call rotation
+
+| Role | Primary | Secondary | Rotation |
+|---|---|---|---|
+| Engineering on-call | security-circle lead | @open-sse | Weekly, Mon 09:00 PDT |
+| Security on-call | @security-team | — | Weekly |
+| Data on-call | @db-team | — | Weekly |
+| Comms lead | @comms | — | As needed |
+
+**Handoff**: every Monday 09:00 PDT, the outgoing on-call posts a
+written handoff to the incoming in `#omniroute-ops-handoff` covering:
+open SEV-3/4 items, scheduled maintenance windows, and any
+in-flight mitigations.
+
+---
+
+## 7. Postmortem expectations
+
+- **Blameless**. People did the best they could with the information
+  they had. Focus on systems, signals, and decision points.
+- **Within 5 business days** of resolution. File via
+  `gh issue create --label postmortem --label SEV-1` (or `--label SEV-2`).
+- **Action items** must be assigned, dated, and tracked in
+  `docs/TECH_DEBT.md` (P0 < 30 d, P1 < 90 d per that doc's SLA).
+- **Mandatory attendees**: incident commander, on-call, any engineer
+  who touched the mitigation, and one person who was *not* involved
+  (fresh-eyes review).
+
+---
+
+## 8. Review log
+
+| Date | Reviewer | Change |
+|---|---|---|
+| 2026-06-18 | security-circle lead | Initial runbook; severity ladder + 15-min checklist + 4.1–4.5 mitigation runbooks. Closes 71-pillar audit L61 (1/3 → 2/3). |
+| 2026-07-18 | observability-circle | Corrected § 4.1/4.3 to the real provider-disable (`PUT /api/providers/{connectionId}`) and authz-inventory (`tiers`/`bypassEnabled`/`cors`, no `policies_active`) mechanisms; removed foreign branding and the nonexistent ADR-024/029 references. |
+| 2026-07-18 (planned) | observability-circle | Wire on-call rotation into PagerDuty schedule; add the postmortem template. |
diff --git a/docs/PERF_BUDGETS.md b/docs/PERF_BUDGETS.md
new file mode 100644
index 00000000000..ca7af81f12e
--- /dev/null
+++ b/docs/PERF_BUDGETS.md
@@ -0,0 +1,227 @@
+# Performance Budgets — OmniRoute (2026-06-18)
+
+**Status**: Authoritative. SLO targets that the 71-pillar audit (L13)
+references for the `Perf > 2.00` gate.
+**Methodology**: per-endpoint p50/p95/p99 latency budgets, plus a
+top-level availability SLO. Budgets are derived from the 3-replica
+Caddy + Redis topology (commit `038439fa7`); adjust on infra change.
+**Enforcement**: none yet. § 6 sketches a `benches/perf-gate.k6.js` k6
+script that would assert the SLOs below, but it is a design reference,
+not a committed file — no `bench/` or `benches/` directory exists in
+this repo today. This doc is a target-setting reference only until a
+CI gate is built as follow-up work.
+**Re-evaluation cadence**: quarterly, or on any major infra change.
+
+---
+
+## 1. Top-level SLOs
+
+| SLO | Target | Window | Page on breach |
+|---|---|---|---|
+| **Availability** (2xx or 4xx for /v1/* and /api/settings/*) | 99.9% | rolling 30 days | on-call P2 |
+| **Error budget burn rate** (1xx normalized rate) | < 2x for 1h, < 6x for 5m | 1h / 5m windows | on-call P1 |
+| **Aggregate p95 latency** (all /v1/*) | ≤ 1.5 s | rolling 5 min | on-call P2 |
+| **Aggregate p99 latency** (all /v1/*) | ≤ 4.0 s | rolling 5 min | on-call P2 |
+
+**Error budget**: 30-day window = 43.2 minutes of unavailability at
+99.9%. Burn rate > 2x is P2; > 6x is P1.
+
+---
+
+## 2. Per-endpoint latency budgets
+
+All budgets measured **server-side** (Next.js Route Handler entry to
+response start, or last byte for streaming). Stream endpoints are
+measured to time-of-first-byte (TTFB) since the body is incremental.
+
+### 2.1 Inference endpoints (the hot path)
+
+| Endpoint | Method | p50 | p95 | p99 | Notes |
+|---|---|---|---|---|---|
+| `/v1/responses` (non-stream) | POST | 800 ms | 1.8 s | 3.5 s | Includes translator + provider roundtrip |
+| `/v1/responses` (stream) | POST (TTFB) | 350 ms | 900 ms | 1.8 s | TTFB only; total duration unbounded |
+| `/v1/relay/chat/completions` (non-stream) | POST | 1.0 s | 2.2 s | 4.0 s | Includes per-(token,IP) rate-limit check |
+| `/v1/relay/chat/completions` (stream) | POST (TTFB) | 400 ms | 1.0 s | 2.0 s | |
+| `/v1/embeddings` | POST | 300 ms | 700 ms | 1.4 s | Pure provider roundtrip; cheap |
+| `/v1/rerank` | POST | 600 ms | 1.4 s | 2.8 s | |
+| `/v1/moderations` | POST | 250 ms | 600 ms | 1.2 s | Lightweight classification |
+| `/v1/audio/speech` | POST | 1.2 s | 3.0 s | 6.0 s | Audio synthesis is slow; budget reflects that |
+| `/v1/audio/transcriptions` | POST | 2.0 s | 5.0 s | 10.0 s | STT is bounded by audio duration + model size |
+| `/v1/images/generations` | POST | 4.0 s | 8.0 s | 15.0 s | Image gen is async-bound by provider |
+| `/v1/videos/generations` | POST (TTFB) | 600 ms | 1.5 s | 3.0 s | Async; client polls `/v1/videos/{id}` |
+| `/v1/music/generations` | POST | 3.0 s | 6.0 s | 12.0 s | |
+
+### 2.2 Files + batches
+
+| Endpoint | Method | p50 | p95 | p99 | Notes |
+|---|---|---|---|---|---|
+| `/v1/files` (GET) | GET | 80 ms | 200 ms | 400 ms | Cached list |
+| `/v1/files` (POST upload) | POST | 500 ms | 1.2 s | 2.5 s | 25 MB cap; multipart parse |
+| `/v1/files/{id}` (GET) | GET | 60 ms | 150 ms | 300 ms | |
+| `/v1/files/{id}` (DELETE) | DELETE | 80 ms | 200 ms | 400 ms | |
+| `/v1/files/{id}/content` (download) | GET | 100 ms | 300 ms | 600 ms | + per-MB throughput |
+| `/v1/batches` (GET) | GET | 150 ms | 400 ms | 800 ms | |
+| `/v1/batches` (POST create) | POST | 200 ms | 500 ms | 1.0 s | Validates input file then enqueues |
+| `/v1/batches/{id}` (GET) | GET | 100 ms | 300 ms | 600 ms | |
+| `/v1/batches/{id}` (DELETE) | DELETE | 100 ms | 300 ms | 600 ms | |
+| `/v1/batches/delete-completed` (POST) | POST | 400 ms | 1.0 s | 2.0 s | Mass delete; n rows |
+
+### 2.3 Agents
+
+| Endpoint | Method | p50 | p95 | p99 | Notes |
+|---|---|---|---|---|---|
+| `/v1/agents/health` | GET | 1.5 s | 4.5 s | 5.0 s | 5s per-provider timeout cap; expect 3-provider total |
+| `/v1/agents/credentials` | GET | 100 ms | 250 ms | 500 ms | Metadata only; values never returned |
+| `/v1/agents/tasks` (GET list) | GET | 150 ms | 400 ms | 800 ms | |
+| `/v1/agents/tasks` (POST create) | POST | 250 ms | 600 ms | 1.2 s | Just enqueues; doesn't run agent |
+| `/v1/agents/tasks/{id}` (GET) | GET | 100 ms | 300 ms | 600 ms | |
+| `/v1/agents/tasks/{id}` (DELETE) | DELETE | 150 ms | 400 ms | 800 ms | |
+
+### 2.4 Combos / me / providers
+
+| Endpoint | Method | p50 | p95 | p99 |
+|---|---|---|---|---|
+| `/v1/combos` | GET | 80 ms | 200 ms | 400 ms |
+| `/v1/me/status` | GET | 60 ms | 150 ms | 300 ms |
+| `/v1/providers/{provider}/models` | GET | 100 ms | 250 ms | 500 ms |
+
+### 2.5 Web / search
+
+| Endpoint | Method | p50 | p95 | p99 | Notes |
+|---|---|---|---|---|---|
+| `/v1/web/fetch` | POST | 1.5 s | 4.0 s | 8.0 s | 10s timeout cap; recurse depth 3 |
+| `/v1/search` | POST | 800 ms | 2.0 s | 4.0 s | Provider search latency varies |
+
+### 2.6 VSCode-CLI shim (token-scoped)
+
+These are the legacy passthrough paths. Budgets are tighter because
+they're called frequently by the VSCode-CLI extension in tight loops.
+
+| Endpoint | Method | p50 | p95 | p99 |
+|---|---|---|---|---|
+| `/v1/vscode/{token}/v1/chat/completions` | POST | 700 ms | 1.6 s | 3.0 s |
+| `/v1/vscode/{token}/v1/models` | GET | 60 ms | 150 ms | 300 ms |
+| `/v1/vscode/{token}/combos` | GET | 80 ms | 200 ms | 400 ms |
+| `/v1/vscode/{token}/chat/completions` (legacy) | POST | 700 ms | 1.6 s | 3.0 s |
+| `/v1/vscode/{token}/models` (legacy) | GET | 60 ms | 150 ms | 300 ms |
+| `/v1/vscode/{token}/responses` | POST | 800 ms | 1.8 s | 3.5 s |
+
+### 2.7 Management / settings
+
+Management endpoints are operator-only and not part of the hot path.
+Budgets are set conservatively; breaches don't page on-call but do
+flag in the weekly perf review.
+
+| Endpoint group | p50 | p95 | p99 |
+|---|---|---|---|
+| `/api/settings/*` (GET) | 100 ms | 300 ms | 600 ms |
+| `/api/settings/*` (POST/PATCH/DELETE) | 200 ms | 500 ms | 1.0 s |
+| `/api/keys/*` (CRUD) | 150 ms | 400 ms | 800 ms |
+| `/api/quota/*` (CRUD) | 150 ms | 400 ms | 800 ms |
+| `/api/monitoring/health` (heavy) | 500 ms | 1.5 s | 3.0 s |
+
+### 2.8 Public probes
+
+| Endpoint | Method | p50 | p95 | p99 |
+|---|---|---|---|---|
+| `/api/health/ping` | GET | 5 ms | 20 ms | 50 ms |
+| `/api/version` | GET | 5 ms | 20 ms | 50 ms |
+| `/api/docs` | GET | 20 ms | 80 ms | 200 ms (HTML shell, no provider call) |
+
+---
+
+## 3. Throughput targets
+
+| Tier | Per-replica RPS | Cluster RPS (3 replicas) | Notes |
+|---|---|---|---|
+| Inference (non-stream) | 50 RPS | 150 RPS | Bounded by provider quota + translator CPU |
+| Inference (stream) | 25 concurrent streams | 75 streams | Bounded by Node event-loop + memory |
+| Embeddings | 200 RPS | 600 RPS | Cheap |
+| Files (upload) | 10 RPS | 30 RPS | Multipart parse + DB write |
+| Files (download) | 100 RPS | 300 RPS | Static-content via Next.js |
+| Combos / me / providers | 500 RPS | 1,500 RPS | Cached |
+| WebSocket | 100 concurrent connections | 300 | Per-IP cap 5 |
+
+**Cluster ceiling** (all endpoints combined, sustained): ~1,000 RPS
+before p95 latency begins to climb. Scale horizontally beyond that
+by adding replicas; the Caddy LB is stateless.
+
+---
+
+## 4. Resource budgets
+
+| Resource | Per-replica cap | Notes |
+|---|---|---|
+| RSS memory | 1.5 GB | Spikes during audio/video gen; expect brief 2 GB |
+| Event-loop lag (p99) | 50 ms | Alert via `clinic doctor` regression |
+| Heap retained | 800 MB | Old-gen GC tuning in `node --max-old-space-size` |
+| File descriptors | 2,000 | `ulimit -n 4096` recommended at host |
+| DB connections (sql.js) | 1 per replica | sql.js is in-process; no pool needed |
+| Redis connections | 20 per replica | Pooled; idle reaped at 5 min |
+
+---
+
+## 5. Cold-start budget
+
+Next.js App Router cold-start on a fresh container:
+
+| Phase | Budget |
+|---|---|
+| Container start → HTTP listening | ≤ 800 ms |
+| First request TTFB (warm) | ≤ 200 ms |
+| Translator registry bootstrap | ≤ 500 ms (one-time, first /v1/responses) |
+
+**Measurement script**: `bin/cold-start-bench.sh` (already in the repo
+since v3.8.36; `bin/` is the canonical scripts dir).
+
+---
+
+## 6. Regression gate (k6 reference, not yet implemented)
+
+The sketch below shows how a future `benches/perf-gate.k6.js` script
+would assert the SLOs above. Nothing in this section is committed or
+wired into CI today — it is a design reference for follow-up work, not
+a running gate.
+
+```javascript
+// benches/perf-gate.k6.js — pseudo-code; not yet committed
+import http from 'k6/http';
+import { check, Trend } from 'k6';
+
+const responsesTTFB = new Trend('v1_responses_ttfb', true);
+
+export const options = {
+  scenarios: {
+    smoke: {
+      executor: 'constant-vus',
+      vus: 10,
+      duration: '1m',
+    },
+  },
+  thresholds: {
+    'http_req_duration{endpoint:v1_responses}': ['p(95)<1800', 'p(99)<3500'],
+    'http_req_failed': ['rate<0.01'],
+    'v1_responses_ttfb': ['p(95)<900'],
+  },
+};
+
+export default function () {
+  const res = http.post(`${__ENV.BASE_URL}/api/v1/responses`, JSON.stringify({
+    model: 'gpt-4o-mini',
+    input: 'ping',
+  }), { headers: { 'Authorization': `Bearer ${__ENV.API_KEY}` }});
+  check(res, { 'status is 200': (r) => r.status === 200 });
+  responsesTTFB.add(res.timings.waiting);
+}
+```
+
+---
+
+## 7. Review log
+
+| Date | Reviewer | Change |
+|---|---|---|
+| 2026-06-18 | security-circle lead | Initial per-endpoint budgets derived from 3-replica Caddy + Redis topology |
+| 2026-07-18 | observability-circle | Clarified this doc ships zero enforcement today (no `bench/`/`benches/` dir, no CI gate) and fixed the stale "not yet committed" claim about `bin/cold-start-bench.sh` (present since v3.8.36). |
+| 2026-07-18 (planned) | observability-circle | Wire `benches/perf-gate.k6.js` into CI; gate on p95 + p99 breach |
+| 2026-09-18 (planned) | observability-circle | Quarterly review; adjust after real-traffic baseline data |
diff --git a/docs/compression/COMPRESSION_GUIDE.md b/docs/compression/COMPRESSION_GUIDE.md
index 326a042a1c1..314458f2874 100644
--- a/docs/compression/COMPRESSION_GUIDE.md
+++ b/docs/compression/COMPRESSION_GUIDE.md
@@ -93,6 +93,8 @@ RTK mode is optimized for verbose tool outputs that appear in coding-agent sessi
   TypeScript/Vite/Webpack builds, ESLint/Biome/Prettier, npm audit/installs, Docker logs, infra
   output, and generic shell output
 - Applies JSON filter packs from `open-sse/services/compression/engines/rtk/filters/`
+- Imports RTK TOML schema v1 filters from project or global `filters.toml` files, with inline-test
+  validation and trust-gating for project files
 - Ships 49 built-in filters with inline verify samples
 - Removes ANSI control sequences, progress bars, repeated lines, and non-actionable noise
 - Preserves failures, errors, warnings, changed files, summaries, and the tail of long output
diff --git a/docs/compression/RTK_COMPRESSION.md b/docs/compression/RTK_COMPRESSION.md
index 4457cda3da1..6ac7f0d2d77 100644
--- a/docs/compression/RTK_COMPRESSION.md
+++ b/docs/compression/RTK_COMPRESSION.md
@@ -52,28 +52,59 @@ class is not enough.
 
 RTK loads filters in this order:
 
-1. Project filters from `.rtk/filters.json`, only when trusted.
-2. Global filters from `DATA_DIR/rtk/filters.json`.
+1. Project filters from `.rtk/filters.toml` and `.rtk/filters.json`, only when trusted.
+2. Global filters from `DATA_DIR/rtk/filters.toml` and `DATA_DIR/rtk/filters.json`.
 3. Built-in filters from `open-sse/services/compression/engines/rtk/filters/`.
 
+Within the same scope, RTK TOML schema v1 filters take precedence over OmniRoute JSON filters. TOML
+`match_command` expressions are checked before command-type matching so an imported command-specific
+filter can override a broader filter in that scope. Project scope still takes precedence over global
+scope, regardless of file format.
+
 Project filters are intentionally trust-gated because regex filters can change how tool output is
 shown to agents. A project filter file is accepted when one of these is true:
 
 - `rtkConfig.trustProjectFilters` is `true`.
 - `OMNIROUTE_RTK_TRUST_PROJECT_FILTERS=1` is set.
-- `.rtk/trust.json` contains the SHA-256 hash of `.rtk/filters.json`.
+- `.rtk/trust.json` contains the matching SHA-256 hash for the project filter file.
 
 Trust file example:
 
 ```json
 {
-  "filtersSha256": "0123456789abcdef..."
+  "filtersSha256": "0123456789abcdef...",
+  "filtersTomlSha256": "fedcba9876543210..."
 }
 ```
 
+The hashes are separate: `filtersSha256` trusts `.rtk/filters.json`, while `filtersTomlSha256`
+trusts `.rtk/filters.toml`. Editing either file invalidates only its own trust entry. Global files
+are administrator-installed and use the existing global-filter trust behavior.
+
 Custom filters can be one filter object or an array of filter objects. Invalid custom filters are
 skipped and reported by `/api/context/rtk/filters` diagnostics. Invalid built-in filters fail fast.
 
+## RTK TOML schema v1 compatibility
+
+OmniRoute can parse, validate, test, and install declarative filter files using RTK TOML schema v1.
+The supported fields are `description`, `match_command`, `strip_ansi`, `filter_stderr`,
+`strip_lines_matching`, `keep_lines_matching`, `replace`, `match_output`, `truncate_lines_at`,
+`head_lines`, `tail_lines`, `max_lines`, `on_empty`, and `[[tests.<filter>]]` inline tests.
+Unknown fields, invalid or unsafe regular expressions, simultaneous strip/keep rules, files over
+1 MiB, and references to unknown filters are rejected. A file whose inline tests fail can be
+validated for inspection but cannot be installed or loaded. Custom-file load failures remain
+fail-open: the invalid file is skipped and the remaining filters continue to work.
+
+OmniRoute receives tool output after the client has already captured it, so `filter_stderr = true`
+cannot change process capture. The field is accepted as a no-op and validation returns a warning.
+This is intentionally described as **RTK TOML schema v1 compatibility**, not full compatibility
+with the RTK executable, shell hooks, Rust command implementations, or its trust-store layout.
+
+The dashboard's advanced RTK view accepts pasted or uploaded TOML. Validation is read-only.
+Installation writes `DATA_DIR/rtk/filters.toml` atomically with restrictive permissions and refreshes
+the live filter catalog without a restart. Replacing an existing file requires explicit `overwrite`
+confirmation and creates `DATA_DIR/rtk/filters.toml.bak` first.
+
 ## Filter DSL
 
 Filters use the JSON schema described in [Compression Rules Format](./COMPRESSION_RULES_FORMAT.md).
@@ -216,6 +247,7 @@ round-trips through the same store and survives a restart.
 | `/api/context/rtk/config`          | GET    | Read RTK config                              |
 | `/api/context/rtk/config`          | PUT    | Update RTK config                            |
 | `/api/context/rtk/filters`         | GET    | List filter catalog and load diagnostics     |
+| `/api/context/rtk/import`          | POST   | Validate or install RTK TOML schema v1 files |
 | `/api/context/rtk/test`            | POST   | Preview RTK compression for one text payload |
 | `/api/context/rtk/raw-output/[id]` | GET    | Read retained redacted raw output            |
 | `/api/compression/preview`         | POST   | Preview any compression mode                 |
@@ -253,6 +285,18 @@ Compression preview payload:
 
 Management routes require dashboard management auth or the matching API-key policy.
 
+RTK TOML validation payload:
+
+```json
+{
+  "action": "validate",
+  "content": "schema_version = 1\n\n[filters.my-tool]\nmatch_command = \"^my-tool\\\\b\"\nmax_lines = 20\n"
+}
+```
+
+Use `"action": "install"` to install the validated file globally. Add `"overwrite": true` only
+after reviewing and confirming replacement of an existing global file.
+
 ## Raw Output Recovery
 
 RTK normally returns only compressed text. For debugging, `rawOutputRetention` can retain redacted
diff --git a/docs/frameworks/MCP-SERVER.md b/docs/frameworks/MCP-SERVER.md
index 1f953da93e0..73fcf367afc 100644
--- a/docs/frameworks/MCP-SERVER.md
+++ b/docs/frameworks/MCP-SERVER.md
@@ -6,13 +6,9 @@ lastUpdated: 2026-06-28
 
 # OmniRoute MCP Server Documentation
 
-> Model Context Protocol server with 94 tools across routing, cache, compression, memory, skills, proxy, pool, and context source operations.
+> Model Context Protocol server with 104 tools across routing, cache, compression, memory, skills, proxy, pool, and context source operations.
 >
-> Source of truth: `open-sse/mcp-server/schemas/tools.ts` (34 base) + `memoryTools.ts` (3) + `skillTools.ts` (4) + `agentSkillTools.ts` (3) + `poolTools.ts` (6) + `gamificationTools.ts` (8) + `pluginTools.ts` (8) + `notionTools.ts` (6) + `obsidianTools.ts` (22) = **94** (`TOTAL_MCP_TOOL_COUNT`). Tool registration and scope wiring lives in `open-sse/mcp-server/server.ts`.
-
-![MCP tool inventory (94 tools by category)](../diagrams/exported/mcp-tools-94.svg)
-
-> Source: [diagrams/mcp-tools-94.mmd](../diagrams/mcp-tools-94.mmd) (regenerate via `npm run docs:render-diagrams`).
+> Source of truth: `open-sse/mcp-server/server.ts` computes **104 unique tools** with `countUniqueMcpTools()`: 42 canonical definitions (including the six CCR lifecycle tools and the agent-skills trio), plus memory (3), skills (4), GitHub skills (3), pool (6), gamification (8), plugins (8), Notion (6), Obsidian (22), and two RTK-only compression tools.
 
 ## Installation
 
@@ -110,7 +106,7 @@ Cursor, Cline, and compatible MCP client setup.
 | `omniroute_cache_stats` | `read:cache`  | Semantic cache, prompt-cache, and idempotency stats |
 | `omniroute_cache_flush` | `write:cache` | Flush cache globally or by signature/model          |
 
-## Compression Tools (5)
+## Compression Tools (13)
 
 | Tool                                | Scopes              | Description                                                                                                              |
 | :---------------------------------- | :------------------ | :----------------------------------------------------------------------------------------------------------------------- |
@@ -119,6 +115,20 @@ Cursor, Cline, and compatible MCP client setup.
 | `omniroute_set_compression_engine`  | `write:compression` | Pick the active engine (off/caveman/rtk/stacked) and Caveman/RTK intensity                                               |
 | `omniroute_list_compression_combos` | `read:compression`  | List named compression combos and their engine pipelines                                                                 |
 | `omniroute_compression_combo_stats` | `read:compression`  | Analytics grouped by compression combo and engine                                                                        |
+| `omniroute_ccr_store`               | `write:compression` | Store caller-isolated content in the bounded in-memory CCR store and return a marker plus `ccr://` reference             |
+| `omniroute_ccr_retrieve`            | `read:compression`  | Retrieve CCR content in full or with head, tail, lines, grep, and stats modes                                            |
+| `omniroute_ccr_inspect`             | `read:compression`  | Inspect caller-owned CCR metadata without returning content                                                              |
+| `omniroute_ccr_list`                | `read:compression`  | List paginated metadata for caller-owned CCR blocks                                                                      |
+| `omniroute_ccr_delete`              | `write:compression` | Delete a caller-owned CCR block                                                                                          |
+| `omniroute_ccr_stats`               | `read:compression`  | Report caller-scoped memory usage, lifecycle counters, and store limits                                                  |
+| `omniroute_rtk_discover`            | `read:compression`  | Discover recurring noise in opt-in RTK output samples                                                                    |
+| `omniroute_rtk_learn`               | `read:compression`  | Generate a reviewable RTK filter draft from opt-in samples                                                               |
+
+CCR entries are in-memory only and disappear on restart. Each block is limited to 2 MiB, each
+principal to 16 MiB, and the global store to 64 MiB. Entries default to a 24-hour TTL (maximum
+seven days). Full MCP retrieval is limited to 256 KiB; larger blocks remain available through the
+ranged and grep modes. Storage, retrieval, listing, inspection, deletion, and stats are isolated by
+the authenticated API-key principal. Audit records contain hashes and size metadata, never content.
 
 `omniroute_compression_status` reports MCP description compression separately under
 `analytics.mcpDescriptionCompression`. Those values are metadata-size estimates for MCP listable
@@ -127,7 +137,7 @@ receipts and are marked with `source: "mcp_metadata_estimate"`.
 
 ### MCP Accessibility Tree Filter (v3.8.0)
 
-Separate from the 5 compression tools above, OmniRoute includes a post-execution filter that
+Separate from the compression tools above, OmniRoute includes a post-execution filter that
 compresses the **tool results** of MCP browser/accessibility tools before they are returned to the
 agent. This filter is not itself a tool — it runs transparently on any tool result that contains
 verbose accessibility-tree or browser-snapshot text (≥2000 chars).
@@ -217,7 +227,7 @@ See [AGENT-SKILLS.md](./AGENT-SKILLS.md) for the full catalog and how external a
 
 ## Related Frameworks (v3.8.0)
 
-The MCP tool inventory above (94 tools = 34 core + 3 memory + 4 skills + 3 agent-skills + 6 pool + 8 gamification + 8 plugins + 6 notion + 22 obsidian) is intentionally
+The MCP tool inventory above (104 unique tools, computed by `countUniqueMcpTools()`) is intentionally
 scoped to runtime routing/cache/compression/memory/skills/proxy/context-source operations. Two adjacent
 frameworks ship alongside the MCP server in v3.8.0 and are documented separately:
 
@@ -331,7 +341,7 @@ MCP tool, prompt, and resource registries can compress descriptions at registrat
 
 Description compression shrinks each tool's metadata; **tool-cardinality reduction** goes one step further by reducing _how many_ tools are announced at all. Advertising fewer tools in the `tools/list` manifest cuts the per-request token cost the client's model pays for the tool catalog ("layer 5" compression). The implementation is a pure, stateless filter in `open-sse/mcp-server/toolCardinality.ts` (`reduceToolManifest`), wired into the registration loop in `createMcpServer()` (`open-sse/mcp-server/server.ts`).
 
-**Opt-in, off by default.** The filter only runs when at least one of two environment variables is set; with neither set, all 94 tools are announced unchanged.
+**Opt-in, off by default.** The filter only runs when at least one of two environment variables is set; with neither set, all 104 tools are announced unchanged.
 
 | Variable         | Mode                                                                                    |
 | :--------------- | :-------------------------------------------------------------------------------------- |
diff --git a/docs/openapi.yaml b/docs/openapi.yaml
index 994154d41d5..6c9241849b0 100644
--- a/docs/openapi.yaml
+++ b/docs/openapi.yaml
@@ -2201,6 +2201,36 @@ paths:
         "200":
           description: RTK filter catalog and diagnostics
 
+  /api/context/rtk/import:
+    post:
+      tags: [Compression]
+      summary: Validate or install an RTK TOML schema v1 filter file
+      security:
+        - ManagementSessionAuth: []
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              required: [action, content]
+              additionalProperties: false
+              properties:
+                action:
+                  type: string
+                  enum: [validate, install]
+                content:
+                  type: string
+                  maxLength: 1048576
+                overwrite:
+                  type: boolean
+                  description: Replace an existing global file and create a backup
+      responses:
+        "200":
+          description: Filter metadata, inline-test outcomes, warnings, and installation status
+        "400":
+          description: Invalid TOML, schema, regular expression, inline test, or install request
+
   /api/context/rtk/test:
     post:
       tags: [Compression]
diff --git a/docs/reference/ENVIRONMENT.md b/docs/reference/ENVIRONMENT.md
index 9b39ccee2ed..6527b4b3031 100644
--- a/docs/reference/ENVIRONMENT.md
+++ b/docs/reference/ENVIRONMENT.md
@@ -411,6 +411,7 @@ detection above).
 | `OMNIROUTE_API_KEY`                             | _(unset)_                                           | MCP/A2A modules                                             | API key for internal MCP tool and A2A skill calls.                                                                                                                      |
 | `OMNIROUTE_API_KEY_ID`                          | _(unset)_                                           | `open-sse/mcp-server/audit.ts`                              | Key ID for MCP audit log attribution.                                                                                                                                   |
 | `ROUTER_API_KEY`                                | _(unset)_                                           | Legacy                                                      | Legacy alias for `OMNIROUTE_API_KEY`.                                                                                                                                   |
+| `OMNIROUTE_ISSUE_AGENT_ENABLED`                 | `false`                                             | `src/app/api/issue-agent/runs/route.ts`                     | Enables the offline/local Issue Agent recorded-triage endpoint. Leave disabled unless explicitly running local recorded-triage workflows.                              |
 | `OMNIROUTE_CONTEXT`                             | _(active context)_                                  | `bin/cli/program.mjs`, `bin/cli/api.mjs`                    | CLI remote-mode context/profile for `omniroute` commands; overrides the active context in the local contexts store. Equivalent to `--context <name>`.                   |
 | `OMNIROUTE_MCP_ENFORCE_SCOPES`                  | `true`                                              | `open-sse/mcp-server/server.ts`                             | Enforce scope-based access control on MCP tool calls.                                                                                                                   |
 | `OMNIROUTE_MCP_SCOPES`                          | _(all)_                                             | `open-sse/mcp-server/server.ts`                             | Comma-separated scopes: `admin`, `combos`, `health`, `models`, `routing`, `budget`, `metrics`, `pricing`, `memory`, `skills`.                                           |
diff --git a/docs/reference/PROVIDER_PLUGIN_MANIFEST.md b/docs/reference/PROVIDER_PLUGIN_MANIFEST.md
index 9f91dc39eda..9db239b371d 100644
--- a/docs/reference/PROVIDER_PLUGIN_MANIFEST.md
+++ b/docs/reference/PROVIDER_PLUGIN_MANIFEST.md
@@ -21,6 +21,14 @@ OmniRoute advertises that URL to Bifrost and CLIProxyAPI via the
 `OMNIROUTE_PROVIDER_MANIFEST_URL` when the sidecar needs a public or container
 network URL instead of the local request origin.
 
+## Refreshing the Manifest
+
+The HTTP endpoint returns `Cache-Control: public, max-age=60` and a strong
+`ETag`. A sidecar should retain the last validated manifest and send its ETag
+in `If-None-Match` when refreshing. A `304 Not Modified` response has no body;
+the sidecar keeps its cached manifest. If no validated cached manifest exists,
+the sidecar must issue an unconditional request instead of accepting a `304`.
+
 ## Goal
 
 Move provider metadata toward a plugin contract so the hot request path can
diff --git a/docs/sessions/20260714-issue-agent-executable-triage/00_SESSION_OVERVIEW.md b/docs/sessions/20260714-issue-agent-executable-triage/00_SESSION_OVERVIEW.md
new file mode 100644
index 00000000000..378c4454e94
--- /dev/null
+++ b/docs/sessions/20260714-issue-agent-executable-triage/00_SESSION_OVERVIEW.md
@@ -0,0 +1,36 @@
+# Issue-Agent Executable Triage: Session Overview
+
+Machine status: `in_progress`
+Updated at: `2026-07-14`
+Issue: `https://github.com/diegosouzapw/OmniRoute/issues/5980`
+PR: `https://github.com/diegosouzapw/OmniRoute/pull/7002`
+
+## Goal
+
+Deliver GitHub issue #5980 as a production issue-agent workflow. The workflow
+must execute recorded GitHub triage through OmniRoute routing, persist a complete
+audit trail, return an actionable result, and cover all terminal outcomes.
+
+## Current State
+
+| artifact_id | requirement                                                            | status                           | current evidence                                                                                     | next proof                                                   |
+| ----------- | ---------------------------------------------------------------------- | -------------------------------- | ---------------------------------------------------------------------------------------------------- | ------------------------------------------------------------ |
+| AC1         | configured provider/model/policy use normal chat routing               | `implemented_pending_acceptance` | `623e0d541`, `fa2c1d7c6`; real-route test invokes the issue-agent route and mocks only provider HTTP | prove routing-policy semantics and terminal failure handling |
+| AC2         | persist lifecycle, request, output, usage/cost/runtime, terminal error | `not_started`                    | audit JSONL currently records only pre-execution run context                                         | lifecycle persistence tests                                  |
+| AC3         | return actionable triage result                                        | `not_started`                    | route forwards raw completion body                                                                   | result contract and integration test                         |
+| AC4         | success, provider failure, timeout, budget stop                        | `not_started`                    | only success-route coverage exists                                                                   | terminal-outcome test matrix                                 |
+| release     | CI/review evidence                                                     | `in_progress`                    | route-validation and focused tests have prior passing evidence                                       | rerun final gates on PR head                                 |
+
+## Decisions
+
+| decision_id | decision                                                                         | rationale                                                                                                      | status        |
+| ----------- | -------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------- | ------------- |
+| DEC-001     | Use the in-process `POST` export from `/api/v1/chat/completions`                 | preserves existing admission, initialization, guardrails, and provider routing                                 | `implemented` |
+| DEC-002     | Keep issue-agent execution opt-in with `OMNIROUTE_ISSUE_AGENT_ENABLED=true`      | prevents unrequested autonomous execution                                                                      | `implemented` |
+| DEC-003     | Treat AC1 as incomplete until policy and error semantics are verified end-to-end | request construction alone does not prove the chat route consumes the policy or returns correct terminal state | `active`      |
+
+## Traceability
+
+The canonical WBS is `03_DAG_WBS.md`; the canonical QA matrix is
+`06_TESTING_STRATEGY.md`. Every status change must identify its commit SHA,
+exact command, observed result, and PR head.
diff --git a/docs/sessions/20260714-issue-agent-executable-triage/01_RESEARCH.md b/docs/sessions/20260714-issue-agent-executable-triage/01_RESEARCH.md
new file mode 100644
index 00000000000..d0af543d875
--- /dev/null
+++ b/docs/sessions/20260714-issue-agent-executable-triage/01_RESEARCH.md
@@ -0,0 +1,28 @@
+# Issue-Agent Executable Triage: Research
+
+Machine status: `complete_for_current_phase`
+
+## In-Repository Findings
+
+| research_id | source                                           | finding                                                                                                                                            | consequence                                                                                        |
+| ----------- | ------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------- |
+| RES-001     | `src/app/api/issue-agent/runs/route.ts`          | the endpoint validates body, rejects unsupported mode/disabled execution, builds recorded context, writes audit JSONL, then delegates non-dry runs | execution behavior is centralized at the issue-agent route                                         |
+| RES-002     | `src/app/api/v1/chat/completions/route.ts`       | standard chat entrypoint exports `POST` and owns the normal chat request path                                                                      | AC1 must exercise this export rather than a fake internal seam                                     |
+| RES-003     | `src/lib/issueAgent/execution.ts`                | provider and model are resolved into the chat request; policy is only encoded as `X-OmniRoute-Mode`                                                | an implementation review must establish that this header is a consumed routing-policy contract     |
+| RES-004     | `src/lib/issueAgent/audit.ts`                    | audit persistence occurs before execution and writes run context/steps only                                                                        | AC2 is unsatisfied: no transition, completion, usage/cost/runtime, or terminal-error record exists |
+| RES-005     | `tests/unit/issue-agent-route-execution.test.ts` | live route test initializes isolated DB, calls the actual issue-agent `POST`, and mocks only `globalThis.fetch` at provider boundary               | strong AC1 path evidence, but it verifies success only and does not prove policy consumption       |
+
+## Validation Evidence
+
+| evidence_id | command                                                                                                                                      | observed                                                      | scope                                                 | evidence_sha |
+| ----------- | -------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------- | ----------------------------------------------------- | ------------ |
+| EVD-001     | `bun test tests/unit/issue-agent-execution.test.ts tests/unit/issue-agent-route-execution.test.ts tests/unit/issue-agent-runs-route.test.ts` | prior focused run reported green                              | AC1 focused path                                      | `fa2c1d7c6`  |
+| EVD-002     | `npm run check:route-validation:t06`                                                                                                         | prior run reported pass                                       | request route validation                              | `e6a63eb33`  |
+| EVD-003     | `npm run typecheck:core`                                                                                                                     | unresolved `omniglyph` declarations outside issue-agent paths | release gate blocked by pre-existing unrelated errors | pre-existing |
+
+## Research Conclusions
+
+The normal chat route is correctly selected as the AC1 integration seam. The
+remaining design work must use a persisted run-lifecycle model rather than
+extending the pre-execution JSONL row. No external API research was needed:
+the implementation uses existing in-repository routes and provider adapters.
diff --git a/docs/sessions/20260714-issue-agent-executable-triage/02_SPECIFICATIONS.md b/docs/sessions/20260714-issue-agent-executable-triage/02_SPECIFICATIONS.md
new file mode 100644
index 00000000000..636b543a54d
--- /dev/null
+++ b/docs/sessions/20260714-issue-agent-executable-triage/02_SPECIFICATIONS.md
@@ -0,0 +1,43 @@
+# Issue-Agent Executable Triage: Specifications
+
+Machine status: `in_progress`
+
+## Acceptance Contract
+
+| ac_id | requirement                                                                                                                             | acceptance evidence                                                                                                                 | status                           |
+| ----- | --------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------- | -------------------------------- |
+| AC1   | Non-dry recorded triage executes through normal chat routing with selected provider, model, and policy                                  | actual issue-agent route reaches chat `POST`; provider-boundary mock observes selected target; policy is proven consumed by routing | `implemented_pending_acceptance` |
+| AC2   | Persist `accepted`, `running`, and terminal state plus sanitized request/prompt, model output, usage, cost, runtime, and terminal error | durable queryable record contains each field for success and failures                                                               | `pending`                        |
+| AC3   | API returns a useful, structured triage result derived from model output                                                                | response has stable triage schema and is not a raw opaque provider payload                                                          | `pending`                        |
+| AC4   | Tests cover success, provider/model failure, timeout, and budget stop                                                                   | each outcome asserts HTTP response and persisted terminal record                                                                    | `pending`                        |
+
+## API Contract (Target)
+
+| field               | rule                                                                                                                       |
+| ------------------- | -------------------------------------------------------------------------------------------------------------------------- |
+| `mode`              | must be `recorded-triage`                                                                                                  |
+| execution selection | accepts configured `provider`, `model`, `routingPolicy`, and bounded `timeoutMs`                                           |
+| `runId`             | stable execution identifier returned for every accepted run                                                                |
+| result              | includes structured triage decision/summary/actions and execution metadata                                                 |
+| errors              | return sanitized terminal error with explicit terminal status; never leak provider credentials or unredacted issue content |
+
+## Persistence Contract (Target)
+
+| field group    | required values                                                                                           |
+| -------------- | --------------------------------------------------------------------------------------------------------- |
+| identity       | run ID, issue URL/repository/number, mode, timestamps                                                     |
+| lifecycle      | `accepted`, `running`, `succeeded`, `failed`, `timed_out`, or `budget_stopped` with transition timestamps |
+| input          | redacted recorded context and rendered prompt fingerprint/content according to retention policy           |
+| routing        | requested provider/model/policy and resolved execution target                                             |
+| output         | sanitized model output and structured triage result                                                       |
+| accounting     | input/output/total tokens, cost, and runtime when available                                               |
+| terminal error | normalized code/message for failure, timeout, and budget stop                                             |
+
+## Assumptions, Risks, Uncertainties
+
+| aru_id  | type        | statement                                                                                | mitigation                                                                | status |
+| ------- | ----------- | ---------------------------------------------------------------------------------------- | ------------------------------------------------------------------------- | ------ |
+| ARU-001 | risk        | `X-OmniRoute-Mode` may not be a consumed routing-policy input in the chat route          | trace the policy contract and test an observable policy effect            | `open` |
+| ARU-002 | risk        | current catch maps all thrown execution errors to HTTP 400 and does not persist them     | introduce typed terminal outcomes and persistence before response mapping | `open` |
+| ARU-003 | risk        | current audit row is emitted before execution and cannot represent final execution state | replace/extend with append-only lifecycle records or durable run storage  | `open` |
+| ARU-004 | uncertainty | provider response metadata may differ by adapter                                         | normalize accounting fields and preserve unknowns explicitly              | `open` |
diff --git a/docs/sessions/20260714-issue-agent-executable-triage/03_DAG_WBS.md b/docs/sessions/20260714-issue-agent-executable-triage/03_DAG_WBS.md
new file mode 100644
index 00000000000..284a885c5c9
--- /dev/null
+++ b/docs/sessions/20260714-issue-agent-executable-triage/03_DAG_WBS.md
@@ -0,0 +1,25 @@
+# Issue-Agent Executable Triage: DAG and WBS
+
+Machine status: `in_progress`
+
+| id      | phase       | acceptance criterion                                             | status   | source paths                                                                        | test paths                                                                           | evidence_sha                | depends_on                |
+| ------- | ----------- | ---------------------------------------------------------------- | -------- | ----------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------ | --------------------------- | ------------------------- |
+| WBS-001 | contract    | AC1-AC4                                                          | complete | `src/app/api/issue-agent/runs/route.ts`                                             | `tests/unit/issue-agent-runs-route.test.ts`                                          | `a4378a26d`                 | -                         |
+| WBS-002 | execution   | AC1: execute through normal chat-completions routing/policy seam | pending  | `src/app/api/issue-agent/runs/route.ts`; `src/app/api/v1/chat/completions/route.ts` | `tests/unit/issue-agent-runs-route.test.ts`                                          | `e6a` (reconciled baseline) | WBS-001                   |
+| WBS-003 | persistence | AC2: persist lifecycle, input, output, usage, and terminal error | pending  | `src/lib/issueAgent/*`; `src/app/api/issue-agent/runs/route.ts`                     | `tests/unit/issue-agent-audit.test.ts`; `tests/unit/issue-agent-runner.test.ts`      | `e6a` (reconciled baseline) | WBS-002                   |
+| WBS-004 | result      | AC3: return an actionable triage result from execution           | pending  | `src/lib/issueAgent/*`; `src/app/api/issue-agent/runs/route.ts`                     | `tests/unit/issue-agent-runner.test.ts`; `tests/unit/issue-agent-runs-route.test.ts` | `e6a` (reconciled baseline) | WBS-002, WBS-003          |
+| WBS-005 | acceptance  | AC4: cover success, provider failure, timeout, and budget stop   | pending  | `src/lib/issueAgent/*`                                                              | `tests/unit/issue-agent-*.test.ts`                                                   | `e6a` (reconciled baseline) | WBS-002, WBS-003, WBS-004 |
+| WBS-006 | release     | PR validation and maintainer review                              | pending  | `.github/workflows/*`                                                               | CI checks                                                                            | `a4378a26d`                 | WBS-005                   |
+
+## Dependency Graph
+
+`WBS-001 -> WBS-002 -> WBS-003 -> WBS-004 -> WBS-005 -> WBS-006`
+
+`a4378a26d` is a prerequisite validation repair: it validates the issue-agent request body through the shared route validator and passes `npm run check:route-validation:t06` (535 routes). It does not satisfy AC1-AC4.
+
+## Machine Evidence Contract
+
+Every WBS item must maintain: `id`, `acceptance_criterion`, `status`, `source_paths`, `test_paths`, `command`, `expected`, `observed`, `evidence_sha`, `updated_at`, and `pr_url`.
+
+PR: `https://github.com/diegosouzapw/OmniRoute/pull/7002`  
+Issue: `https://github.com/diegosouzapw/OmniRoute/issues/5980`
diff --git a/docs/sessions/20260714-issue-agent-executable-triage/04_IMPLEMENTATION_STRATEGY.md b/docs/sessions/20260714-issue-agent-executable-triage/04_IMPLEMENTATION_STRATEGY.md
new file mode 100644
index 00000000000..1c591749739
--- /dev/null
+++ b/docs/sessions/20260714-issue-agent-executable-triage/04_IMPLEMENTATION_STRATEGY.md
@@ -0,0 +1,37 @@
+# Issue-Agent Executable Triage: Implementation Strategy
+
+Machine status: `in_progress`
+
+## Phase Plan
+
+| phase | work package                                                | dependency        | exit evidence                                                           | status        |
+| ----- | ----------------------------------------------------------- | ----------------- | ----------------------------------------------------------------------- | ------------- |
+| P1    | verify/finish routing-policy contract and failure semantics | existing AC1 seam | actual chat route test proves policy consumption and non-2xx mapping    | `in_progress` |
+| P2    | introduce durable execution lifecycle persistence           | P1                | records transitions, request/prompt, output, accounting, terminal error | `pending`     |
+| P3    | normalize actionable triage result                          | P2                | stable API result schema derived from completion                        | `pending`     |
+| P4    | implement terminal outcome controls                         | P2                | provider failure, timeout, budget stop transition tests                 | `pending`     |
+| P5    | release validation and PR review                            | P1-P4             | focused tests, route gate, relevant typecheck/CI evidence               | `pending`     |
+
+## Architecture
+
+1. Keep `src/app/api/issue-agent/runs/route.ts` as the API adapter: validation,
+   feature gate, and response formatting only.
+2. Keep the standard chat `POST` as the routing boundary; do not add a parallel
+   provider invocation path.
+3. Extract lifecycle persistence and result normalization into focused
+   `src/lib/issueAgent/` modules. Do not overload the existing pre-execution audit
+   writer with unrelated transport behavior.
+4. Use typed execution outcomes so provider failure, abort/timeout, and budget
+   termination are distinguishable before HTTP mapping and persistence.
+5. Add tests from the actual route down to a mocked external provider boundary;
+   use unit tests for pure normalization and lifecycle state transitions.
+
+## Quality Controls
+
+| control            | command or review                                     | threshold                                                  |
+| ------------------ | ----------------------------------------------------- | ---------------------------------------------------------- |
+| route contract     | `npm run check:route-validation:t06`                  | pass                                                       |
+| AC1 route behavior | focused `bun test` issue-agent route/execution suites | policy and provider/model assertions pass                  |
+| AC2-AC4            | lifecycle/result/terminal-outcome suites              | all required states persist and API matches                |
+| static safety      | `npm run typecheck:core`                              | distinguish new failures from existing `omniglyph` blocker |
+| patch integrity    | `git diff --check origin/main...HEAD`                 | pass                                                       |
diff --git a/docs/sessions/20260714-issue-agent-executable-triage/05_KNOWN_ISSUES.md b/docs/sessions/20260714-issue-agent-executable-triage/05_KNOWN_ISSUES.md
new file mode 100644
index 00000000000..6c3f0fbe0e3
--- /dev/null
+++ b/docs/sessions/20260714-issue-agent-executable-triage/05_KNOWN_ISSUES.md
@@ -0,0 +1,21 @@
+# Issue-Agent Executable Triage: Known Issues
+
+Machine status: `open`
+
+| issue_id | severity | status | evidence                                                                                                                        | impact                                                                                                                 | resolution owner          |
+| -------- | -------- | ------ | ------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------- | ------------------------- |
+| KI-001   | P1       | `open` | `execution.ts` places `routingPolicy` in `X-OmniRoute-Mode`; the researched chat route has no observed consumer in the AC1 path | AC1 does not yet prove configured routing policy affects routing                                                       | AC1 implementation/review |
+| KI-002   | P1       | `open` | issue-agent route catches execution errors and returns `{ error }` with HTTP 400 after writing only pre-execution audit         | provider failure, timeout, and budget stop lack correct terminal semantics and persistence                             | AC2/AC4 implementation    |
+| KI-003   | P1       | `open` | `audit.ts` serializes only run context/steps before execution                                                                   | AC2 fields for lifecycle, prompt, output, token/cost/runtime, and error are missing                                    | AC2 implementation        |
+| KI-004   | P1       | `open` | API returns raw `completion.body`                                                                                               | AC3 has no stable actionable triage result contract                                                                    | AC3 implementation        |
+| KI-005   | P2       | `open` | `npm run typecheck:core` has unresolved `omniglyph` declarations in `open-sse/services/compression/*`                           | full typecheck cannot be used as issue-agent completion evidence until separately resolved or excluded with provenance | release validation        |
+
+## Resolved/Verified
+
+| issue_id | status     | evidence                                                                                                                      |
+| -------- | ---------- | ----------------------------------------------------------------------------------------------------------------------------- |
+| KI-R001  | `verified` | `fa2c1d7c6` adds an isolated test that invokes the actual issue-agent route and mocks only provider HTTP for the success path |
+| KI-R002  | `verified` | `e6a63eb33` applies shared request-body validation to the issue-agent route; prior route-validation gate passed               |
+
+No workaround in this document changes the acceptance contract. Open P1 items
+block declaring AC1-AC4 complete.
diff --git a/docs/sessions/20260714-issue-agent-executable-triage/06_TESTING_STRATEGY.md b/docs/sessions/20260714-issue-agent-executable-triage/06_TESTING_STRATEGY.md
new file mode 100644
index 00000000000..006908ef71a
--- /dev/null
+++ b/docs/sessions/20260714-issue-agent-executable-triage/06_TESTING_STRATEGY.md
@@ -0,0 +1,25 @@
+# Issue-Agent Executable Triage: Testing Strategy
+
+Machine status: `in_progress`
+
+## QA Matrix
+
+| qa_id  | AC           | scenario                                                            | command                                                                                    | expected                                              | observed                                | status        | evidence_sha |
+| ------ | ------------ | ------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | ----------------------------------------------------- | --------------------------------------- | ------------- | ------------ |
+| QA-001 | prerequisite | request schema validation                                           | `npm run check:route-validation:t06`                                                       | all routes pass                                       | 535 routes scanned; pass                | pass          | `a4378a26d`  |
+| QA-002 | AC1          | selected provider/model/policy reaches normal chat-completions seam | `bun test tests/unit/issue-agent-runs-route.test.ts`                                       | captured request uses configured routing inputs       | not implemented                         | pending       | `e6a`        |
+| QA-003 | AC2          | run lifecycle persists input, output, usage, terminal error         | `bun test tests/unit/issue-agent-audit.test.ts tests/unit/issue-agent-runner.test.ts`      | durable records for every terminal state              | not implemented                         | pending       | `e6a`        |
+| QA-004 | AC3          | successful execution returns actionable triage output               | `bun test tests/unit/issue-agent-runner.test.ts tests/unit/issue-agent-runs-route.test.ts` | output derives from routed execution, not placeholder | not implemented                         | pending       | `e6a`        |
+| QA-005 | AC4          | provider/model failure                                              | `bun test tests/unit/issue-agent-runner.test.ts`                                           | failed lifecycle and sanitized error persisted        | missing coverage                        | pending       | `e6a`        |
+| QA-006 | AC4          | timeout                                                             | `bun test tests/unit/issue-agent-runner.test.ts`                                           | timed-out lifecycle and terminal error persisted      | missing coverage                        | pending       | `e6a`        |
+| QA-007 | AC4          | budget stop                                                         | `bun test tests/unit/issue-agent-runner.test.ts`                                           | budget stop is explicit and persisted                 | missing coverage                        | pending       | `e6a`        |
+| QA-008 | release      | core type safety                                                    | `npm run typecheck:core`                                                                   | pass                                                  | pending rerun after dependency recovery | pending       | `e6a`        |
+| QA-009 | release      | whitespace integrity                                                | `git diff --check origin/main...HEAD`                                                      | no errors                                             | passed before remote rewrite            | pass/reverify | `a4378a26d`  |
+
+## Test Rules
+
+Tests must mock only the external provider boundary. AC1 must exercise the in-process `POST` export from `src/app/api/v1/chat/completions/route.ts` so admission, policy, translator initialization, and routing remain in the execution path. Each terminal outcome asserts both API behavior and persisted audit state.
+
+## Evidence Requirements
+
+Before a WBS item is marked complete, record the exact command output, commit SHA, test identifiers, and whether the test environment had a lockfile-compatible dependency set. The current recovered environment has incomplete dependencies due to `npm ci` disk exhaustion; no pending test may be reported as passing until rerun.
diff --git a/electron/package-lock.json b/electron/package-lock.json
index b5fda4c818d..65430313322 100644
--- a/electron/package-lock.json
+++ b/electron/package-lock.json
@@ -1,18 +1,18 @@
 {
   "name": "omniroute-desktop",
-  "version": "3.8.46",
+  "version": "3.8.49",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omniroute-desktop",
-      "version": "3.8.46",
+      "version": "3.8.49",
       "license": "MIT",
       "dependencies": {
         "electron-updater": "^6.8.9"
       },
       "devDependencies": {
-        "electron": "^43.1.0",
+        "electron": "^43.1.1",
         "electron-builder": "^26.15.3"
       },
       "engines": {
@@ -1367,9 +1367,9 @@
       }
     },
     "node_modules/electron": {
-      "version": "43.1.0",
-      "resolved": "https://registry.npmjs.org/electron/-/electron-43.1.0.tgz",
-      "integrity": "sha512-DPfxpQLd4NL3BJ8DBxYAfmLUKKesF5Rx9dQx5FyczAP8bhOPScjHE48GArVeXu68LlAainuwkmQTQvdZwpIIAQ==",
+      "version": "43.1.1",
+      "resolved": "https://registry.npmjs.org/electron/-/electron-43.1.1.tgz",
+      "integrity": "sha512-I5c5vfuVvaXpWx3IZdwvXgxQW44+e7OP1wXGVQkogLeSFSkUZ6sLCcWV05AdEcs65AO5tAIJJwbp7ixw+LdarA==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
diff --git a/electron/package.json b/electron/package.json
index b3b5f1a268e..fcae7078db2 100644
--- a/electron/package.json
+++ b/electron/package.json
@@ -28,7 +28,7 @@
     "electron-updater": "^6.8.9"
   },
   "devDependencies": {
-    "electron": "^43.1.0",
+    "electron": "^43.1.1",
     "electron-builder": "^26.15.3"
   },
   "overrides": {
diff --git a/next.config.mjs b/next.config.mjs
index b3f7887bb5c..fa46e3adbd7 100644
--- a/next.config.mjs
+++ b/next.config.mjs
@@ -9,8 +9,8 @@ const distDir = process.env.NEXT_DIST_DIR || ".build/next";
 const projectRoot = dirname(fileURLToPath(import.meta.url));
 const scriptSrc =
   process.env.NODE_ENV === "development"
-    ? "script-src 'self' 'unsafe-inline' 'unsafe-eval' blob:"
-    : "script-src 'self' 'unsafe-inline' 'unsafe-eval' blob:";
+    ? "script-src 'self' 'unsafe-inline' 'unsafe-eval' blob: https://static.cloudflareinsights.com"
+    : "script-src 'self' 'unsafe-inline' 'unsafe-eval' blob: https://static.cloudflareinsights.com";
 const contentSecurityPolicy = [
   "default-src 'self'",
   "base-uri 'self'",
diff --git a/open-sse/config/agyModels.ts b/open-sse/config/agyModels.ts
index d836888b7ff..919b97e9573 100644
--- a/open-sse/config/agyModels.ts
+++ b/open-sse/config/agyModels.ts
@@ -19,7 +19,7 @@ export const AGY_PUBLIC_MODELS = Object.freeze([
   {
     id: "claude-opus-4-6-thinking",
     name: "Claude Opus 4.6 (Thinking)",
-    contextLength: 200000,
+    contextLength: 1048576,
     maxOutputTokens: 65536,
     supportsReasoning: true,
     supportsVision: true,
@@ -28,7 +28,7 @@ export const AGY_PUBLIC_MODELS = Object.freeze([
   {
     id: "claude-sonnet-4-6",
     name: "Claude Sonnet 4.6 (Thinking)",
-    contextLength: 200000,
+    contextLength: 1048576,
     maxOutputTokens: 65536,
     supportsReasoning: true,
     supportsVision: true,
diff --git a/open-sse/config/antigravityModelAliases.ts b/open-sse/config/antigravityModelAliases.ts
index a88ab804c02..3feebf495b1 100644
--- a/open-sse/config/antigravityModelAliases.ts
+++ b/open-sse/config/antigravityModelAliases.ts
@@ -7,7 +7,7 @@ export const ANTIGRAVITY_PUBLIC_MODELS = Object.freeze([
   {
     id: "claude-sonnet-5",
     name: "Claude Sonnet 5 (Thinking)",
-    contextLength: 200000,
+    contextLength: 1048576,
     maxOutputTokens: 65536,
     supportsReasoning: true,
     supportsVision: true,
@@ -16,7 +16,7 @@ export const ANTIGRAVITY_PUBLIC_MODELS = Object.freeze([
   {
     id: "claude-opus-4-6-thinking",
     name: "Claude Opus 4.6 (Thinking)",
-    contextLength: 200000,
+    contextLength: 1048576,
     maxOutputTokens: 65536,
     supportsReasoning: true,
     supportsVision: true,
@@ -25,7 +25,7 @@ export const ANTIGRAVITY_PUBLIC_MODELS = Object.freeze([
   {
     id: "claude-sonnet-4-6",
     name: "Claude Sonnet 4.6 (Thinking)",
-    contextLength: 200000,
+    contextLength: 1048576,
     maxOutputTokens: 65536,
     supportsReasoning: true,
     supportsVision: true,
diff --git a/open-sse/config/freeModelCatalog.data.ts b/open-sse/config/freeModelCatalog.data.ts
index 6b807affb19..f1725e012e5 100644
--- a/open-sse/config/freeModelCatalog.data.ts
+++ b/open-sse/config/freeModelCatalog.data.ts
@@ -285,6 +285,7 @@ export const FREE_MODEL_BUDGETS: FreeModelBudget[] = [
   { provider: "nscale", modelId: "meta-llama/Llama-4-Scout-17B-16E-Instruct", displayName: "meta-llama/Llama-4-Scout-17B-16E-Instruct", monthlyTokens: 0, creditTokens: 5000000, freeType: "one-time-initial", poolKey: "nscale", tos: "caution" },
   { provider: "nscale", modelId: "meta-llama/Llama-3.3-70B-Instruct", displayName: "meta-llama/Llama-3.3-70B-Instruct", monthlyTokens: 0, creditTokens: 5000000, freeType: "one-time-initial", poolKey: "nscale", tos: "caution" },
   { provider: "nvidia", modelId: "z-ai/glm-5.1", displayName: "GLM 5.1", monthlyTokens: 0, creditTokens: 0, freeType: "one-time-initial", poolKey: "nvidia", tos: "caution" },
+  { provider: "nvidia", modelId: "z-ai/glm-5.2", displayName: "GLM 5.2", monthlyTokens: 0, creditTokens: 0, freeType: "one-time-initial", poolKey: "nvidia", tos: "caution" },
   { provider: "nvidia", modelId: "minimaxai/minimax-m2.7", displayName: "MiniMax M2.7", monthlyTokens: 0, creditTokens: 0, freeType: "one-time-initial", poolKey: "nvidia", tos: "caution" },
   { provider: "nvidia", modelId: "google/gemma-4-31b-it", displayName: "Gemma 4 31B", monthlyTokens: 0, creditTokens: 0, freeType: "one-time-initial", poolKey: "nvidia", tos: "caution" },
   { provider: "nvidia", modelId: "mistralai/mistral-small-4-119b-2603", displayName: "Mistral Small 4 2603", monthlyTokens: 0, creditTokens: 0, freeType: "one-time-initial", poolKey: "nvidia", tos: "caution" },
diff --git a/open-sse/config/imageRegistry.ts b/open-sse/config/imageRegistry.ts
index 109d494b038..45803f41229 100644
--- a/open-sse/config/imageRegistry.ts
+++ b/open-sse/config/imageRegistry.ts
@@ -9,11 +9,16 @@ import { LMARENA_DIRECT_IMAGE_MODELS } from "./providers/registry/lmarena/direct
 import { SEGMIND_IMAGE_PROVIDER } from "./providers/registry/segmind/imageModels.ts";
 import { KIE_IMAGE_MODELS } from "./providers/registry/kie/imageModels.ts";
 import { FREEPIK_IMAGE_PROVIDER } from "./providers/registry/freepik/index.ts";
+import { STABILITY_AI_IMAGE_MODELS } from "./providers/registry/stability-ai/imageModels.ts";
+import { GEMINI_IMAGEN_PROVIDER } from "./providers/registry/gemini/imageModels.ts";
 
 interface ImageModelEntry {
   id: string;
   name: string;
   inputModalities?: string[];
+  // See STABILITY_AI_IMAGE_MODELS for why this exists: some models accept "text"
+  // but mechanically require an image regardless.
+  imageRequired?: boolean;
   description?: string;
   isMarket?: boolean;
 }
@@ -38,6 +43,7 @@ interface ImageModelAliasEntry {
   name: string;
   listInCatalog: boolean;
   inputModalities?: string[];
+  imageRequired?: boolean;
   description?: string;
 }
 
@@ -126,6 +132,13 @@ function findImageModelConfig(providerId, modelId) {
   return provider.models.find((model) => model.id === modelId) || null;
 }
 
+// Kept out of getImageModelEntry() (which sits at the complexity-ratchet cap) — an
+// alias can override imageRequired directly, else it falls back to its target
+// model's own flag. Consumers coerce the result with Boolean(), so no `?? false`.
+function resolveAliasImageRequired(alias, modelConfig) {
+  return alias.imageRequired ?? modelConfig?.imageRequired;
+}
+
 export const IMAGE_PROVIDERS: Record<string, ImageProviderConfig> = {
   openai: {
     id: "openai",
@@ -280,6 +293,10 @@ export const IMAGE_PROVIDERS: Record<string, ImageProviderConfig> = {
     supportedSizes: ["1024x1024"],
   },
 
+  // Google AI Studio Imagen family — dedicated :predict endpoint, not generateContent.
+  // See providers/registry/gemini/imageModels.ts for the full rationale.
+  gemini: GEMINI_IMAGEN_PROVIDER,
+
   //Curruntly no models serving
   nebius: {
     id: "nebius",
@@ -472,32 +489,7 @@ export const IMAGE_PROVIDERS: Record<string, ImageProviderConfig> = {
     authType: "apikey",
     authHeader: "bearer",
     format: "stability-ai",
-    models: [
-      { id: "stable-image-ultra", name: "Stable Image Ultra" },
-      { id: "stable-image-core", name: "Stable Image Core" },
-      { id: "sd3.5-large-turbo", name: "sd3.5-large-turbo" },
-      { id: "sd3.5-large", name: "sd3.5-large" },
-      { id: "sd3.5-medium", name: "sd3.5-medium" },
-      { id: "sd3.5-flash", name: "sd3.5-flash" },
-      { id: "erase", name: "Erase", inputModalities: ["image"] },
-      { id: "inpaint", name: "Inpaint", inputModalities: ["text", "image"] },
-      { id: "outpaint", name: "Outpaint", inputModalities: ["text", "image"] },
-      { id: "remove-background", name: "Remove Background", inputModalities: ["image"] },
-      { id: "search-and-replace", name: "Search and Replace", inputModalities: ["text", "image"] },
-      { id: "search-and-recolor", name: "Search and Recolor", inputModalities: ["text", "image"] },
-      {
-        id: "replace-background-and-relight",
-        name: "Replace Background and Relight",
-        inputModalities: ["text", "image"],
-      },
-      { id: "creative", name: "Creative Upscale", inputModalities: ["text", "image"] },
-      { id: "fast", name: "Fast Upscale", inputModalities: ["image"] },
-      { id: "conservative", name: "Conservative Upscale", inputModalities: ["image"] },
-      { id: "sketch", name: "Sketch Control", inputModalities: ["text", "image"] },
-      { id: "structure", name: "Structure Control", inputModalities: ["text", "image"] },
-      { id: "style", name: "Style Control", inputModalities: ["text", "image"] },
-      { id: "style-transfer", name: "Style Transfer", inputModalities: ["text", "image"] },
-    ],
+    models: STABILITY_AI_IMAGE_MODELS,
     supportedSizes: ["1024x1024", "1024x1280", "1280x1024"],
   },
 
@@ -620,7 +612,9 @@ export const IMAGE_PROVIDERS: Record<string, ImageProviderConfig> = {
   // beyond this seed list.
   huggingface: {
     id: "huggingface",
-    baseUrl: "https://api-inference.huggingface.co/models",
+    // HF retired api-inference.huggingface.co; text-to-image now routes through
+    // router.huggingface.co with the hf-inference provider pinned in the path.
+    baseUrl: "https://router.huggingface.co/hf-inference/models",
     authType: "apikey",
     authHeader: "bearer",
     format: "huggingface-image",
@@ -766,6 +760,7 @@ export function getImageModelEntry(modelStr) {
       provider: alias.provider,
       model: alias.model,
       inputModalities: alias.inputModalities || modelConfig?.inputModalities || ["text"],
+      imageRequired: resolveAliasImageRequired(alias, modelConfig),
       description: alias.description || modelConfig?.description || undefined,
     };
   }
@@ -780,6 +775,18 @@ export function getImageModelEntry(modelStr) {
     provider,
     model,
     inputModalities: modelConfig.inputModalities || ["text"],
+    imageRequired: modelConfig.imageRequired,
     description: modelConfig.description || undefined,
   };
 }
+
+/**
+ * An image input is only MANDATORY for edit-only models — those whose modalities
+ * are `["image"]` with no `"text"`. Models listing both `["text", "image"]` accept
+ * an image but can also run pure text-to-image, so they must NOT be gated on an
+ * image input (that gate previously blocked 41 dual-modality t2i models).
+ */
+export function modalitiesRequireImageInput(inputModalities) {
+  const list = Array.isArray(inputModalities) ? inputModalities : ["text"];
+  return list.includes("image") && !list.includes("text");
+}
diff --git a/open-sse/config/nvidiaHostedModels.snapshot.json b/open-sse/config/nvidiaHostedModels.snapshot.json
new file mode 100644
index 00000000000..29bb66e0ad5
--- /dev/null
+++ b/open-sse/config/nvidiaHostedModels.snapshot.json
@@ -0,0 +1,18 @@
+[
+  "deepseek-ai/deepseek-v4-pro",
+  "google/gemma-4-31b-it",
+  "minimaxai/minimax-m2.7",
+  "mistralai/devstral-2-123b-instruct-2512",
+  "mistralai/mistral-large-3-675b-instruct-2512",
+  "mistralai/mistral-small-4-119b-2603",
+  "nvidia/nemotron-3-super-120b-a12b",
+  "openai/gpt-oss-120b",
+  "openai/gpt-oss-20b",
+  "poolside/laguna-xs-2.1",
+  "qwen/qwen3.5-122b-a10b",
+  "qwen/qwen3.5-397b-a17b",
+  "stepfun-ai/step-3.5-flash",
+  "thinkingmachines/inkling",
+  "z-ai/glm-5.1",
+  "z-ai/glm-5.2"
+]
diff --git a/open-sse/config/providers/index.ts b/open-sse/config/providers/index.ts
index ecdd7a82b5c..286ed60fbdd 100644
--- a/open-sse/config/providers/index.ts
+++ b/open-sse/config/providers/index.ts
@@ -123,6 +123,7 @@ import { gitlawbProvider } from "./registry/gitlawb/index.ts";
 import { liquidProvider } from "./registry/liquid/index.ts";
 import { deepinfraProvider } from "./registry/deepinfra/index.ts";
 import { agyProvider } from "./registry/agy/index.ts";
+import { agnesProvider } from "./registry/agnes/index.ts";
 import { udioProvider } from "./registry/udio/index.ts";
 import { longcatProvider } from "./registry/longcat/index.ts";
 import { vertex_partnerProvider } from "./registry/vertex/partner/index.ts";
@@ -318,6 +319,7 @@ export const REGISTRY: Record<string, RegistryEntry> = {
   liquid: liquidProvider,
   deepinfra: deepinfraProvider,
   agy: agyProvider,
+  agnes: agnesProvider,
   udio: udioProvider,
   longcat: longcatProvider,
   "vertex-partner": vertex_partnerProvider,
diff --git a/open-sse/config/providers/registry/agnes/index.ts b/open-sse/config/providers/registry/agnes/index.ts
new file mode 100644
index 00000000000..c3120a562c0
--- /dev/null
+++ b/open-sse/config/providers/registry/agnes/index.ts
@@ -0,0 +1,26 @@
+import type { RegistryEntry } from "../../shared.ts";
+import { buildOpenAiCompatibleRegistryEntry } from "../../shared.ts";
+
+export const agnesProvider: RegistryEntry = buildOpenAiCompatibleRegistryEntry({
+  id: "agnes",
+  baseUrl: "https://apihub.agnes-ai.com/v1/chat/completions",
+  models: [
+    {
+      id: "agnes-2.0-flash",
+      name: "Agnes 2.0 Flash",
+      contextLength: 524288,
+      maxOutputTokens: 65536,
+      supportsReasoning: true,
+      supportsVision: true,
+      toolCalling: true,
+      interleavedField: "reasoning_content",
+    },
+    {
+      id: "agnes-1.5-flash",
+      name: "Agnes 1.5 Flash",
+      contextLength: 262144,
+      maxOutputTokens: 65536,
+      supportsVision: true,
+    },
+  ],
+});
diff --git a/open-sse/config/providers/registry/auggie/index.ts b/open-sse/config/providers/registry/auggie/index.ts
index ee346426a0e..1415c627a7a 100644
--- a/open-sse/config/providers/registry/auggie/index.ts
+++ b/open-sse/config/providers/registry/auggie/index.ts
@@ -3,6 +3,8 @@ import type { RegistryEntry } from "../../shared.ts";
 // Augment / Auggie CLI — local no-auth provider. The executor spawns the
 // user's local `auggie` binary (auth handled entirely by `auggie login`);
 // OmniRoute never stores credentials for this connection.
+//
+// Model IDs sourced from `auggie model list` on auggie v0.32.0.
 export const auggieProvider: RegistryEntry = {
   id: "auggie",
   alias: "aug",
@@ -13,26 +15,38 @@ export const auggieProvider: RegistryEntry = {
   authHeader: "none",
   defaultContextLength: 200000,
   models: [
-    // Claude
-    { id: "claude-sonnet-4.6", name: "Claude Sonnet 4.6", contextLength: 200000 },
-    {
-      id: "claude-sonnet-4.6-thinking",
-      name: "Claude Sonnet 4.6 Thinking",
-      contextLength: 200000,
-    },
-    { id: "claude-opus-4.6", name: "Claude Opus 4.6", contextLength: 200000 },
-    { id: "claude-haiku-4.5", name: "Claude Haiku 4.5", contextLength: 200000 },
-    // Gemini
-    { id: "gemini-3.1-pro", name: "Gemini 3.1 Pro", contextLength: 1000000 },
-    { id: "gemini-3.0-flash", name: "Gemini 3 Flash", contextLength: 1000000 },
-    // GPT-5.x
-    { id: "gpt-5.5-high", name: "GPT-5.5 High", contextLength: 200000 },
-    { id: "gpt-5.5-medium", name: "GPT-5.5 Medium", contextLength: 200000 },
-    { id: "gpt-5.4-high", name: "GPT-5.4 High", contextLength: 200000 },
-    { id: "gpt-5.4-medium", name: "GPT-5.4 Medium", contextLength: 200000 },
-    // Kimi
+    // ── Anthropic Claude ────────────────────────────────────────────────
+    { id: "sonnet4.6", name: "Sonnet 4.6", contextLength: 200000 },
+    { id: "fable-5", name: "Claude Fable 5", contextLength: 200000 },
+    { id: "haiku4.5", name: "Haiku 4.5", contextLength: 200000 },
+    { id: "sonnet4.5", name: "Sonnet 4.5", contextLength: 200000 },
+    { id: "sonnet4.6-500k", name: "Sonnet 4.6 (500K)", contextLength: 500000 },
+    { id: "sonnet5-high", name: "Claude Sonnet 5", contextLength: 200000 },
+    { id: "sonnet5-500k", name: "Claude Sonnet 5 (500K)", contextLength: 500000 },
+    { id: "opus4.5", name: "Opus 4.5", contextLength: 200000 },
+    { id: "opus4.6", name: "Opus 4.6", contextLength: 200000 },
+    { id: "opus4.6-500k", name: "Opus 4.6 (500K)", contextLength: 500000 },
+    { id: "opus4.7", name: "Opus 4.7", contextLength: 200000 },
+    { id: "opus4.7-500k", name: "Opus 4.7 (500K)", contextLength: 500000 },
+    { id: "opus4.8", name: "Opus 4.8", contextLength: 200000 },
+    // ── Gemini ──────────────────────────────────────────────────────────
+    { id: "gemini-3.1-pro-preview", name: "Gemini 3.1 Pro", contextLength: 1000000 },
+    // ── OpenAI GPT ──────────────────────────────────────────────────────
+    { id: "gpt5", name: "GPT-5", contextLength: 200000 },
+    { id: "gpt5.1", name: "GPT-5.1", contextLength: 200000 },
+    { id: "gpt5.2", name: "GPT-5.2", contextLength: 200000 },
+    { id: "gpt5.4", name: "GPT-5.4", contextLength: 200000 },
+    { id: "gpt5.4-mini", name: "GPT-5.4 Mini", contextLength: 200000 },
+    { id: "gpt5.5", name: "GPT-5.5", contextLength: 200000 },
+    { id: "gpt5.6-luna", name: "GPT-5.6 Luna", contextLength: 200000 },
+    { id: "gpt5.6-sol", name: "GPT-5.6 Sol", contextLength: 200000 },
+    { id: "gpt5.6-terra", name: "GPT-5.6 Terra", contextLength: 200000 },
+    // ── Others ──────────────────────────────────────────────────────────
+    { id: "glm-5.2", name: "GLM 5.2", contextLength: 200000 },
     { id: "kimi-k2.6", name: "Kimi K2.6", contextLength: 131000 },
-    // Prism (Augment's in-house model)
-    { id: "prism", name: "Augment Prism", contextLength: 200000 },
+    { id: "kimi-k2.7", name: "Kimi K2.7 Code", contextLength: 131000 },
+    // ── Augment Prism (composite routers) ───────────────────────────────
+    { id: "prism-a", name: "Prism (Claude + Gemini)", contextLength: 200000 },
+    { id: "prism-b", name: "Prism (GPT + Kimi)", contextLength: 200000 },
   ],
 };
diff --git a/open-sse/config/providers/registry/claude/index.ts b/open-sse/config/providers/registry/claude/index.ts
index 19b2c5a5096..d142872de84 100644
--- a/open-sse/config/providers/registry/claude/index.ts
+++ b/open-sse/config/providers/registry/claude/index.ts
@@ -81,7 +81,7 @@ export const claudeProvider: RegistryEntry = {
       id: "claude-sonnet-4-6",
       name: "Claude 4.6 Sonnet",
       supportsXHighEffort: false,
-      contextLength: 200000,
+      contextLength: 1000000,
       maxOutputTokens: 64000,
     },
     {
diff --git a/open-sse/config/providers/registry/gemini/imageModels.ts b/open-sse/config/providers/registry/gemini/imageModels.ts
new file mode 100644
index 00000000000..bc7002da013
--- /dev/null
+++ b/open-sse/config/providers/registry/gemini/imageModels.ts
@@ -0,0 +1,32 @@
+/**
+ * Google AI Studio (Gemini API) Imagen family image-generation provider entry.
+ *
+ * Uses the dedicated `:predict` endpoint (handled by format "google-imagen"), NOT
+ * generateContent — so only imagen-* models belong here; gemini flash-image /
+ * nano-banana route through /v1/chat/completions instead. The models are also
+ * surfaced live via ListModels; this seed makes them addressable on
+ * /v1/images/generations. Note: Imagen requires a billing-enabled Google project —
+ * free-tier keys get 403 / quota 0. The handler builds `{baseUrl}/{model}:predict`.
+ *
+ * Extracted out of imageRegistry.ts (which sits right at the 800-line file-size
+ * cap) so the catalog lives in its own semantic family module, following the same
+ * pattern as `providers/registry/stability-ai/imageModels.ts` and
+ * `providers/registry/segmind/imageModels.ts`. Co-located with the existing
+ * `gemini/index.ts` chat-provider entry — same provider id, different
+ * modality/consumer (chat registry vs image registry), mirroring the
+ * `kie/index.ts` + `kie/imageModels.ts` split.
+ */
+export const GEMINI_IMAGEN_PROVIDER = {
+  id: "gemini",
+  alias: "gemini",
+  baseUrl: "https://generativelanguage.googleapis.com/v1beta/models",
+  authType: "apikey",
+  authHeader: "x-goog-api-key",
+  format: "google-imagen",
+  models: [
+    { id: "imagen-4.0-generate-001", name: "Imagen 4" },
+    { id: "imagen-4.0-ultra-generate-001", name: "Imagen 4 Ultra" },
+    { id: "imagen-4.0-fast-generate-001", name: "Imagen 4 Fast" },
+  ],
+  supportedSizes: ["1024x1024", "1792x1024", "1024x1792"],
+};
diff --git a/open-sse/config/providers/registry/opencode/go/index.ts b/open-sse/config/providers/registry/opencode/go/index.ts
index c8e575b185b..e3dd4b4d80a 100644
--- a/open-sse/config/providers/registry/opencode/go/index.ts
+++ b/open-sse/config/providers/registry/opencode/go/index.ts
@@ -17,14 +17,22 @@ export const opencode_goProvider: RegistryEntry = {
     // glm-5.2 is now advertised and Kimi chat traffic must route through
     // `kimi-k2.7-code` (the live API rejects the plain `kimi-k2.7` alias for
     // `/chat/completions`, even though the docs config example uses it).
-    { id: "glm-5.2", name: "GLM-5.2" },
+    // GLM-5.2 — base model + effort-tier aliases (#6922).
+    // OpencodeExecutor rewrites the alias to the canonical id and injects
+    // reasoning_effort, mirroring the deepseek-v4-pro-* pattern.
+    { id: "glm-5.2", name: "GLM-5.2", supportsReasoning: true },
+    { id: "glm-5.2-high", name: "GLM-5.2 (high effort)", supportsReasoning: true },
+    { id: "glm-5.2-max", name: "GLM-5.2 (max effort)", supportsReasoning: true },
     { id: "glm-5.1", name: "GLM-5.1" },
     { id: "glm-5", name: "GLM-5" },
     { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
     { id: "kimi-k2.6", name: "Kimi K2.6" },
     { id: "kimi-k2.5", name: "Kimi K2.5" },
-    { id: "mimo-v2.5-pro", name: "MiMo-V2.5-Pro" },
-    { id: "mimo-v2.5", name: "MiMo-V2.5" },
+    // MiMo-V2.5 — base model + effort-tier aliases (#6922).
+    { id: "mimo-v2.5-pro", name: "MiMo-V2.5-Pro", supportsReasoning: true },
+    { id: "mimo-v2.5", name: "MiMo-V2.5", supportsReasoning: true },
+    { id: "mimo-v2.5-high", name: "MiMo-V2.5 (high effort)", supportsReasoning: true },
+    { id: "mimo-v2.5-max", name: "MiMo-V2.5 (max effort)", supportsReasoning: true },
     // #3110: MiniMax M3 via OpenCode Go tier
     {
       id: "minimax-m3",
diff --git a/open-sse/config/providers/registry/stability-ai/imageModels.ts b/open-sse/config/providers/registry/stability-ai/imageModels.ts
new file mode 100644
index 00000000000..4a5b20623e9
--- /dev/null
+++ b/open-sse/config/providers/registry/stability-ai/imageModels.ts
@@ -0,0 +1,76 @@
+/**
+ * Stability AI image-generation model catalog.
+ *
+ * Extracted out of imageRegistry.ts (which sits right at the 800-line file-size
+ * cap) so the catalog lives in its own semantic family module, following the same
+ * pattern as `providers/registry/kie/imageModels.ts` and
+ * `providers/registry/segmind/imageModels.ts`. See `imageRegistry.ts`'s
+ * `stability-ai` entry for baseUrl/auth/format wiring.
+ *
+ * `imageRequired: true` marks the dedicated edit/control/upscale endpoints
+ * (STABILITY_EDIT_ENDPOINTS in open-sse/handlers/imageGeneration.ts) that accept a
+ * text prompt but mechanically require an input image regardless —
+ * modalitiesRequireImageInput() alone can't tell them apart from flexible
+ * dual-modality generation models (BFL Kontext, Together, NVIDIA, LMArena,
+ * NanoGPT), which correctly allow pure text-to-image.
+ */
+
+export interface StabilityImageModelEntry {
+  id: string;
+  name: string;
+  inputModalities?: string[];
+  imageRequired?: boolean;
+}
+
+export const STABILITY_AI_IMAGE_MODELS: StabilityImageModelEntry[] = [
+  { id: "stable-image-ultra", name: "Stable Image Ultra" },
+  { id: "stable-image-core", name: "Stable Image Core" },
+  { id: "sd3.5-large-turbo", name: "sd3.5-large-turbo" },
+  { id: "sd3.5-large", name: "sd3.5-large" },
+  { id: "sd3.5-medium", name: "sd3.5-medium" },
+  { id: "sd3.5-flash", name: "sd3.5-flash" },
+  { id: "erase", name: "Erase", inputModalities: ["image"] },
+  { id: "inpaint", name: "Inpaint", inputModalities: ["text", "image"], imageRequired: true },
+  { id: "outpaint", name: "Outpaint", inputModalities: ["text", "image"], imageRequired: true },
+  { id: "remove-background", name: "Remove Background", inputModalities: ["image"] },
+  {
+    id: "search-and-replace",
+    name: "Search and Replace",
+    inputModalities: ["text", "image"],
+    imageRequired: true,
+  },
+  {
+    id: "search-and-recolor",
+    name: "Search and Recolor",
+    inputModalities: ["text", "image"],
+    imageRequired: true,
+  },
+  {
+    id: "replace-background-and-relight",
+    name: "Replace Background and Relight",
+    inputModalities: ["text", "image"],
+    imageRequired: true,
+  },
+  {
+    id: "creative",
+    name: "Creative Upscale",
+    inputModalities: ["text", "image"],
+    imageRequired: true,
+  },
+  { id: "fast", name: "Fast Upscale", inputModalities: ["image"] },
+  { id: "conservative", name: "Conservative Upscale", inputModalities: ["image"] },
+  { id: "sketch", name: "Sketch Control", inputModalities: ["text", "image"], imageRequired: true },
+  {
+    id: "structure",
+    name: "Structure Control",
+    inputModalities: ["text", "image"],
+    imageRequired: true,
+  },
+  { id: "style", name: "Style Control", inputModalities: ["text", "image"], imageRequired: true },
+  {
+    id: "style-transfer",
+    name: "Style Transfer",
+    inputModalities: ["text", "image"],
+    imageRequired: true,
+  },
+];
diff --git a/open-sse/executors/antigravity.ts b/open-sse/executors/antigravity.ts
index e3fc66b378b..c3d4e4fbe8b 100644
--- a/open-sse/executors/antigravity.ts
+++ b/open-sse/executors/antigravity.ts
@@ -62,6 +62,10 @@ import {
   markCreditsExhausted,
   type SafeAntigravityLog,
 } from "./antigravity/executeAttempt.ts";
+import {
+  handleAntigravityFallbackChainError,
+  handleAntigravityFallback400,
+} from "./antigravity/proFallbackChain.ts";
 import {
   generateAntigravityRequestId,
   getAntigravityEnvelopeUserAgent,
@@ -365,7 +369,18 @@ function sanitizeAntigravityGeminiRequest(
   const geminiTools = buildGeminiTools(request.tools);
   if (geminiTools) {
     clean.tools = geminiTools;
-    clean.toolConfig = { functionCallingConfig: { mode: "VALIDATED" } };
+    // #6914: Preserve includeServerSideToolInvocations from the raw request's
+    // toolConfig when present (set by transformRequest when tools exist). The
+    // sanitize whitelist would otherwise rebuild toolConfig without it.
+    const rawToolConfig = asRecord(request.toolConfig);
+    const rawFnConfig = asRecord(rawToolConfig?.functionCallingConfig);
+    const includeServerSide = rawFnConfig?.includeServerSideToolInvocations === true;
+    clean.toolConfig = {
+      functionCallingConfig: {
+        mode: "VALIDATED",
+        ...(includeServerSide ? { includeServerSideToolInvocations: true } : {}),
+      },
+    };
   } else if (asRecord(request.toolConfig)) {
     clean.toolConfig = request.toolConfig;
   }
@@ -641,7 +656,7 @@ export class AntigravityExecutor extends BaseExecutor {
       safetySettings: getAntigravitySafetySettings(normalizedRequest?.safetySettings),
       toolConfig:
         Array.isArray(normalizedRequest?.tools) && normalizedRequest.tools.length > 0
-          ? { functionCallingConfig: { mode: "VALIDATED" } }
+          ? { functionCallingConfig: { mode: "VALIDATED", includeServerSideToolInvocations: true } }
           : normalizedRequest?.toolConfig,
     };
 
@@ -1015,7 +1030,28 @@ export class AntigravityExecutor extends BaseExecutor {
     let firstResult: Awaited<ReturnType<AntigravityExecutor["executeOnce"]>> | null = null;
     for (let i = 0; i < chain.length; i++) {
       const candidate = chain[i];
-      const result = await this.executeOnce(input, candidate);
+      let result: Awaited<ReturnType<AntigravityExecutor["executeOnce"]>>;
+      try {
+        result = await this.executeOnce(input, candidate);
+      } catch (error) {
+        const outcome = handleAntigravityFallbackChainError(
+          input,
+          error,
+          candidate,
+          i,
+          chain,
+          firstResult,
+          resolvedUpstreamId
+        );
+        switch (outcome.action) {
+          case "throw":
+            throw outcome.error;
+          case "return":
+            return outcome.result;
+          default:
+            continue;
+        }
+      }
 
       // Success (or any non-400) on a candidate → return immediately.
       if (result.response.status !== HTTP_STATUS.BAD_REQUEST) {
@@ -1023,23 +1059,18 @@ export class AntigravityExecutor extends BaseExecutor {
       }
 
       // Remember the FIRST 400 so the exhausted-chain case surfaces the original error.
-      if (i === 0) firstResult = result;
-
-      const isLast = i === chain.length - 1;
-      if (!isLast) {
-        input.log?.debug?.(
-          "AG_PRO_FALLBACK",
-          `400 on "${candidate}" — retrying with next Pro candidate "${chain[i + 1]}"`
-        );
-        continue;
-      }
-
-      // Chain exhausted: surface the FIRST candidate's sanitized 400.
-      input.log?.warn?.(
-        "AG_PRO_FALLBACK",
-        `Pro fallback chain exhausted (all ${chain.length} candidates 400'd) for "${resolvedUpstreamId}"`
+      if (!firstResult) firstResult = result;
+
+      const outcome400 = handleAntigravityFallback400(
+        input,
+        result,
+        firstResult,
+        candidate,
+        i,
+        chain,
+        resolvedUpstreamId
       );
-      return firstResult ?? result;
+      if (outcome400.action === "return") return outcome400.result;
     }
 
     // Unreachable (loop always returns), but keeps the type checker happy.
diff --git a/open-sse/executors/antigravity/proFallbackChain.ts b/open-sse/executors/antigravity/proFallbackChain.ts
new file mode 100644
index 00000000000..ec5b710c1f8
--- /dev/null
+++ b/open-sse/executors/antigravity/proFallbackChain.ts
@@ -0,0 +1,104 @@
+// Pure Pro-family fallback-chain decision helpers for the Antigravity executor (#7290):
+// decide what execute()'s per-candidate loop does after executeOnce() throws or
+// returns a 400, without depending on executor instance state (no `this`).
+// Extracted from antigravity.ts (file-size cap) -- mirrors the existing
+// antigravity/sseCollect.ts submodule pattern.
+import type { ExecuteInput } from "../base.ts";
+
+/** Shape of one execute()/executeOnce() result (kept local to avoid importing the class). */
+export type AntigravityExecuteResult = {
+  response: Response;
+  url: string;
+  headers: Record<string, string>;
+  transformedBody: unknown;
+};
+
+/** True for an aborted request (caller disconnect) — never retried across candidates. */
+export function isAntigravityAbortError(input: ExecuteInput, error: unknown): boolean {
+  return Boolean(
+    input.signal?.aborted ||
+      (error instanceof DOMException && error.name === "AbortError") ||
+      (error instanceof Error && error.name === "AbortError")
+  );
+}
+
+export type AntigravityFallbackChainErrorOutcome =
+  | { action: "throw"; error: unknown }
+  | { action: "return"; result: AntigravityExecuteResult }
+  | { action: "continue" };
+
+/**
+ * Decide what execute()'s Pro-fallback loop does after executeOnce() THROWS for one
+ * candidate: propagate an abort immediately, retry the next candidate, surface the
+ * first 400 if the chain is exhausted, or throw a chain-exhausted error.
+ */
+export function handleAntigravityFallbackChainError(
+  input: ExecuteInput,
+  error: unknown,
+  candidate: string,
+  i: number,
+  chain: readonly string[],
+  firstResult: AntigravityExecuteResult | null,
+  resolvedUpstreamId: string
+): AntigravityFallbackChainErrorOutcome {
+  // Abort signal (user disconnect) — propagate immediately, do not retry.
+  if (isAntigravityAbortError(input, error)) {
+    return { action: "throw", error };
+  }
+  if (i < chain.length - 1) {
+    input.log?.debug?.(
+      "AG_PRO_FALLBACK",
+      `Exception on "${candidate}" (${error instanceof Error ? error.message : String(error)}) -- retrying with next Pro candidate "${chain[i + 1]}"`
+    );
+    return { action: "continue" };
+  }
+  // Last candidate also threw -- return original 400 if available, otherwise throw.
+  if (firstResult) {
+    input.log?.warn?.(
+      "AG_PRO_FALLBACK",
+      `Pro fallback chain exhausted (last candidate threw, but first candidate returned 400) for "${resolvedUpstreamId}". Returning original 400.`
+    );
+    return { action: "return", result: firstResult };
+  }
+  return {
+    action: "throw",
+    error: new Error(
+      `Pro fallback chain exhausted (all ${chain.length} candidates failed). Last error: ${error instanceof Error ? error.message : String(error)}`
+    ),
+  };
+}
+
+export type AntigravityFallback400Outcome =
+  | { action: "return"; result: AntigravityExecuteResult }
+  | { action: "continue" };
+
+/**
+ * Decide what execute()'s Pro-fallback loop does after one candidate returns a 400:
+ * retry the next candidate, or (chain exhausted) surface the first candidate's
+ * sanitized 400.
+ */
+export function handleAntigravityFallback400(
+  input: ExecuteInput,
+  result: AntigravityExecuteResult,
+  firstResult: AntigravityExecuteResult | null,
+  candidate: string,
+  i: number,
+  chain: readonly string[],
+  resolvedUpstreamId: string
+): AntigravityFallback400Outcome {
+  const isLast = i === chain.length - 1;
+  if (!isLast) {
+    input.log?.debug?.(
+      "AG_PRO_FALLBACK",
+      `400 on "${candidate}" — retrying with next Pro candidate "${chain[i + 1]}"`
+    );
+    return { action: "continue" };
+  }
+
+  // Chain exhausted: surface the FIRST candidate's sanitized 400.
+  input.log?.warn?.(
+    "AG_PRO_FALLBACK",
+    `Pro fallback chain exhausted (all ${chain.length} candidates 400'd) for "${resolvedUpstreamId}"`
+  );
+  return { action: "return", result: firstResult ?? result };
+}
diff --git a/open-sse/executors/antigravity/sseCollect.ts b/open-sse/executors/antigravity/sseCollect.ts
index d42bab72e96..7c6e7957871 100644
--- a/open-sse/executors/antigravity/sseCollect.ts
+++ b/open-sse/executors/antigravity/sseCollect.ts
@@ -96,9 +96,23 @@ export function processAntigravitySSEPayload(
             collected.textContent += part.text;
           }
         }
+        // Native Gemini function calls. Non-streaming responses (and some
+        // streaming ones) carry the tool call as `part.functionCall` rather than
+        // the textual `[Tool call: ...]` markdown. Without this, a tool-only
+        // response produced empty content and a 502 Provider error (#7037).
+        if (part.functionCall && typeof part.functionCall.name === "string") {
+          addAntigravityTextualToolCall(collected, {
+            name: part.functionCall.name,
+            args: part.functionCall.args ?? {},
+          });
+        }
       }
     }
-    if (candidate?.finishReason) {
+    // Preserve a tool-call finish reason: once a native `part.functionCall`
+    // (or textual tool call) has populated `toolCalls`, the candidate's own
+    // finish reason (often STOP) must not clobber it (#7037 — a tool-only
+    // response would otherwise report STOP and lose its tool-call signal).
+    if (candidate?.finishReason && collected.toolCalls.length === 0) {
       collected.finishReason = normalizeOpenAICompatibleFinishReasonString(
         String(candidate.finishReason).toLowerCase()
       );
diff --git a/open-sse/executors/auggie.ts b/open-sse/executors/auggie.ts
index b4034cbb2da..f3c98c8d8f9 100644
--- a/open-sse/executors/auggie.ts
+++ b/open-sse/executors/auggie.ts
@@ -38,8 +38,114 @@ const AUGGIE_URL = "auggie://cli/stdio";
 // untrusted-input sink. We only ever pass a model that is declared in the
 // registry entry — this closes flag-smuggling (a `model` starting with "-" would
 // otherwise be parsed by auggie as an option) and unknown-model passthrough.
+//
+// The static registry (shipped with the code) is checked first.  On first use
+// the executor also spawns `auggie model list` at runtime and merges any IDs it
+// finds — this lets the allowlist stay current when auggie adds or renames
+// models without a code update.
 const AUGGIE_MODEL_ALLOWLIST: ReadonlySet<string> = new Set(auggieProvider.models.map((m) => m.id));
 const DEFAULT_AUGGIE_MODEL = auggieProvider.models[0]?.id ?? "claude-sonnet-4.6";
+// ─── Model alias map (backward compat for saved combos) ─────────────────────
+// Old model IDs from before the v0.32.0 registry update; each maps to the
+// equivalent v0.32.0 ID so existing combos continue to work after the rename.
+const AUGGIE_MODEL_ALIASES: ReadonlyMap<string, string> = new Map([
+  // Claude
+  ["claude-sonnet-4.6", "sonnet4.6"],
+  ["claude-sonnet-4.6-thinking", "sonnet4.6"],
+  ["claude-opus-4.6", "opus4.6"],
+  ["claude-haiku-4.5", "haiku4.5"],
+  // Gemini
+  ["gemini-3.1-pro", "gemini-3.1-pro-preview"],
+  ["gemini-3.0-flash", "gemini-3.1-pro-preview"],
+  // GPT-5.x (high/medium split was synthetic — v0.32.0 has a single ID per version)
+  ["gpt-5.5-high", "gpt5.5"],
+  ["gpt-5.5-medium", "gpt5.5"],
+  ["gpt-5.4-high", "gpt5.4"],
+  ["gpt-5.4-medium", "gpt5.4"],
+]);
+
+/**
+ * Live model cache populated by `initAuggieModels()`.
+ * - `null`  = not yet attempted
+ * - `Set`   = successfully fetched IDs (possibly empty)
+ */
+let liveModelSet: Set<string> | null = null;
+
+/**
+ * Spawn `auggie model list`, parse `[model-id]` entries, and merge them into
+ * the live allowlist so the executor accepts models auggie recognises even
+ * when the static registry has not been updated yet.
+ *
+ * Safe to call repeatedly: only the first call spawns the process; subsequent
+ * calls are a no-op (including after a failed fetch — `liveModelSet` is set to
+ * an empty set so we don't retry every request).
+ */
+export async function initAuggieModels(
+  signal?: AbortSignal | null,
+  timeoutMs = 8000
+): Promise<void> {
+  if (liveModelSet !== null) return;
+  let bin: string;
+  try {
+    bin = resolveAuggieBin();
+  } catch {
+    liveModelSet = new Set();
+    return;
+  }
+  const child = spawn(bin, ["model", "list"], {
+    env: process.env,
+    stdio: ["ignore", "pipe", "pipe"],
+    shell: false,
+    windowsHide: true,
+  });
+  const fragments: string[] = [];
+  child.stdout.on("data", (d: Buffer) => fragments.push(d.toString("utf8")));
+  let settled = false;
+  const settle = (result: Set<string>) => {
+    if (settled) return;
+    settled = true;
+    liveModelSet = result;
+  };
+  const timer = setTimeout(() => {
+    if (!child.killed) child.kill("SIGKILL");
+    settle(new Set());
+  }, timeoutMs);
+  const onAbort = () => {
+    if (!child.killed) child.kill("SIGKILL");
+    clearTimeout(timer);
+    settle(new Set());
+  };
+  if (signal) {
+    if (signal.aborted) {
+      clearTimeout(timer);
+      settle(new Set());
+      return;
+    }
+    signal.addEventListener("abort", onAbort, { once: true });
+  }
+  try {
+    const code = await new Promise<number | null>((resolve, reject) => {
+      child.on("close", resolve);
+      child.on("error", (e: Error) => reject(e));
+    });
+    clearTimeout(timer);
+    signal?.removeEventListener("abort", onAbort);
+    if (code !== 0) {
+      settle(new Set());
+      return;
+    }
+    const ids = new Set<string>();
+    for (const line of fragments.join("").split("\n")) {
+      const m = line.match(/\[([^\]]+)\]/);
+      if (m) ids.add(m[1]);
+    }
+    settle(ids.size > 0 ? ids : new Set());
+  } catch {
+    clearTimeout(timer);
+    settle(new Set());
+    signal?.removeEventListener("abort", onAbort);
+  }
+}
 
 type AuggieModelResolution = { ok: true; model: string } | { ok: false; error: string };
 
@@ -47,6 +153,9 @@ type AuggieModelResolution = { ok: true; model: string } | { ok: false; error: s
  * Validate + resolve the requested model against the registry allowlist.
  * Rejects flag-smuggling (leading "-") and any id not declared in the registry.
  * An empty/absent model resolves to the registry's first (default) model.
+ *
+ * Note: `initAuggieModels()` must be called at least once before this function
+ * sees live-discovered models (the executor's `execute()` does this).
  */
 export function resolveAuggieModel(model: unknown): AuggieModelResolution {
   const requested = typeof model === "string" ? model.trim() : "";
@@ -57,15 +166,20 @@ export function resolveAuggieModel(model: unknown): AuggieModelResolution {
       error: `Invalid Auggie model "${requested}": model must not start with "-".`,
     };
   }
-  if (!AUGGIE_MODEL_ALLOWLIST.has(requested)) {
-    return {
-      ok: false,
-      error: `Unknown Auggie model "${requested}". Supported models: ${[
-        ...AUGGIE_MODEL_ALLOWLIST,
-      ].join(", ")}.`,
-    };
-  }
-  return { ok: true, model: requested };
+  // Backward-compat alias: resolve old model IDs → v0.32.0 equivalents.
+  // This lets saved combos referencing the old names keep working.
+  const requestedAlias = AUGGIE_MODEL_ALIASES.get(requested);
+  if (requestedAlias) return { ok: true, model: requestedAlias };
+  // Static registry — always authoritative for the shipped set.
+  if (AUGGIE_MODEL_ALLOWLIST.has(requested)) return { ok: true, model: requested };
+  // Live-discovered models (if loaded) extend the static list.
+  if (liveModelSet?.has(requested)) return { ok: true, model: requested };
+  const known = [...AUGGIE_MODEL_ALLOWLIST];
+  if (liveModelSet) known.push(...liveModelSet);
+  return {
+    ok: false,
+    error: `Unknown Auggie model "${requested}". Supported models: ${known.join(", ")}.`,
+  };
 }
 
 /**
@@ -151,6 +265,7 @@ export function buildAuggiePrompt(messages: OpenAIMsg[]): string {
       }
     }
     if (!text.trim()) continue;
+
     if (role === "system") {
       lines.push(`[System]\n${text}`);
     } else if (role === "assistant") {
@@ -252,7 +367,6 @@ export class AuggieExecutor extends BaseExecutor {
   ): Promise<Partial<ProviderCredentials> | null> {
     return null;
   }
-
   async execute({ model, body, stream, signal, log }: ExecuteInput): Promise<{
     response: Response;
     url: string;
@@ -265,6 +379,9 @@ export class AuggieExecutor extends BaseExecutor {
     const auggieBin = resolveAuggieBin();
     const wantsStream = stream !== false;
 
+    // On first execution, try to discover model IDs the local auggie recognises.
+    // Best-effort: missing/inactive CLI falls through to the static list.
+    await initAuggieModels(signal);
     // Argument-injection defense: never forward an unvalidated model into the argv.
     const modelResolution = resolveAuggieModel(model);
     if (!modelResolution.ok) {
@@ -601,3 +718,13 @@ function buildAuggieSseError(message: string): Response {
     },
   });
 }
+
+// ─── Test helpers ──────────────────────────────────────────────────────────
+
+/**
+ * Reset the live model cache for testing.
+ * Not exported from the package index.
+ */
+export function __resetAuggieModels(): void {
+  liveModelSet = null;
+}
diff --git a/open-sse/executors/base/reasoningEffort.ts b/open-sse/executors/base/reasoningEffort.ts
index dcabf6ef260..dac59232856 100644
--- a/open-sse/executors/base/reasoningEffort.ts
+++ b/open-sse/executors/base/reasoningEffort.ts
@@ -56,6 +56,86 @@ export function supportsMaxEffortForProvider(provider: string, model: string): b
   return isClaude || isOpencodeGoDeepSeek || isOllamaCloud;
 }
 
+// ── Effort carrier helpers (#7044) ──────────────────────────────────────────
+// OmniRoute carries the requested effort on up to three shapes:
+//   1. top-level `reasoning_effort`        — OpenAI / OmniRoute-internal
+//   2. `reasoning.effort`                  — OpenAI Responses shape
+//   3. `output_config.effort`              — Anthropic Messages native (Claude Code / Claude passthrough)
+// Carrier (3) was previously invisible to this sanitizer, so a native Claude request
+// carrying `output_config.effort: "xhigh"` reached providers that don't accept xhigh
+// (e.g. claude-sonnet-4-6, supportsXHighEffort=false) unchanged → HTTP 400 (#7044).
+interface EffortCarriers {
+  reasoning: Record<string, unknown> | null;
+  outputConfig: Record<string, unknown> | null;
+  hasTopLevelReasoningEffort: boolean;
+  hasReasoningEffort: boolean;
+  hasOutputConfigEffort: boolean;
+  effort: unknown;
+}
+
+function readEffortCarriers(b: Record<string, unknown>): EffortCarriers {
+  const reasoning =
+    b.reasoning && typeof b.reasoning === "object" && !Array.isArray(b.reasoning)
+      ? (b.reasoning as Record<string, unknown>)
+      : null;
+  const outputConfig =
+    b.output_config && typeof b.output_config === "object" && !Array.isArray(b.output_config)
+      ? (b.output_config as Record<string, unknown>)
+      : null;
+  const hasTopLevelReasoningEffort = Object.prototype.hasOwnProperty.call(b, "reasoning_effort");
+  const hasReasoningEffort = !!(
+    reasoning && Object.prototype.hasOwnProperty.call(reasoning, "effort")
+  );
+  const hasOutputConfigEffort = !!(
+    outputConfig && Object.prototype.hasOwnProperty.call(outputConfig, "effort")
+  );
+  const effort = b.reasoning_effort ?? reasoning?.effort ?? outputConfig?.effort;
+  return {
+    reasoning,
+    outputConfig,
+    hasTopLevelReasoningEffort,
+    hasReasoningEffort,
+    hasOutputConfigEffort,
+    effort,
+  };
+}
+
+/** Write a normalized effort value back to every carrier that was present. */
+function writeEffortValue(
+  b: Record<string, unknown>,
+  value: string,
+  c: EffortCarriers
+): Record<string, unknown> {
+  const next: Record<string, unknown> = { ...b };
+  if (c.hasTopLevelReasoningEffort) next.reasoning_effort = value;
+  if (c.hasReasoningEffort && c.reasoning) next.reasoning = { ...c.reasoning, effort: value };
+  if (c.hasOutputConfigEffort && c.outputConfig)
+    next.output_config = { ...c.outputConfig, effort: value };
+  return next;
+}
+
+/** Strip the effort field from every carrier that was present. */
+function stripEffortValue(
+  b: Record<string, unknown>,
+  c: EffortCarriers
+): Record<string, unknown> {
+  const next: Record<string, unknown> = { ...b };
+  if (c.hasTopLevelReasoningEffort) delete next.reasoning_effort;
+  if (c.hasReasoningEffort && c.reasoning) {
+    const r: Record<string, unknown> = { ...c.reasoning };
+    delete r.effort;
+    if (Object.keys(r).length === 0) delete next.reasoning;
+    else next.reasoning = r;
+  }
+  if (c.hasOutputConfigEffort && c.outputConfig) {
+    const oc: Record<string, unknown> = { ...c.outputConfig };
+    delete oc.effort;
+    if (Object.keys(oc).length === 0) delete next.output_config;
+    else next.output_config = oc;
+  }
+  return next;
+}
+
 export function sanitizeReasoningEffortForProvider(
   body: unknown,
   provider: string,
@@ -64,14 +144,9 @@ export function sanitizeReasoningEffortForProvider(
 ): unknown {
   if (!body || typeof body !== "object" || Array.isArray(body)) return body;
   const b = body as Record<string, unknown>;
-  const reasoning =
-    b.reasoning && typeof b.reasoning === "object" && !Array.isArray(b.reasoning)
-      ? (b.reasoning as Record<string, unknown>)
-      : null;
-  const hasTopLevelReasoningEffort = Object.prototype.hasOwnProperty.call(b, "reasoning_effort");
-  const effort = b.reasoning_effort ?? reasoning?.effort;
-  if (effort === undefined) return body;
-  const effortStr = typeof effort === "string" ? effort.toLowerCase() : "";
+  const c = readEffortCarriers(b);
+  if (c.effort === undefined) return body;
+  const effortStr = typeof c.effort === "string" ? c.effort.toLowerCase() : "";
   const modelStr = model || "";
 
   const githubOptIn =
@@ -84,15 +159,7 @@ export function sanitizeReasoningEffortForProvider(
       "REASONING_SANITIZE",
       `${provider}/${modelStr}: removed unsupported reasoning_effort`
     );
-    const next: Record<string, unknown> = { ...b };
-    delete next.reasoning_effort;
-    if (reasoning) {
-      const r = { ...reasoning };
-      delete r.effort;
-      if (Object.keys(r).length === 0) delete next.reasoning;
-      else next.reasoning = r;
-    }
-    return next;
+    return stripEffortValue(b, c);
   }
 
   // Native DeepSeek (api.deepseek.com) — V4 thinking mode accepts reasoning_effort
@@ -111,10 +178,7 @@ export function sanitizeReasoningEffortForProvider(
         "REASONING_SANITIZE",
         `deepseek/${modelStr}: normalized reasoning_effort ${effortStr} → ${mapped}`
       );
-      const next: Record<string, unknown> = { ...b };
-      if (hasTopLevelReasoningEffort) next.reasoning_effort = mapped;
-      if (reasoning) next.reasoning = { ...reasoning, effort: mapped };
-      return next;
+      return writeEffortValue(b, mapped, c);
     }
     return body;
   }
@@ -131,14 +195,7 @@ export function sanitizeReasoningEffortForProvider(
       "REASONING_SANITIZE",
       `${provider}/${modelStr}: normalized reasoning_effort max → xhigh`
     );
-    const next: Record<string, unknown> = { ...b };
-    if (hasTopLevelReasoningEffort) {
-      next.reasoning_effort = "xhigh";
-    }
-    if (reasoning) {
-      next.reasoning = { ...reasoning, effort: "xhigh" };
-    }
-    return next;
+    return writeEffortValue(b, "xhigh", c);
   }
 
   if (shouldDowngradeXHigh || shouldDowngradeMax) {
@@ -146,14 +203,7 @@ export function sanitizeReasoningEffortForProvider(
       "REASONING_SANITIZE",
       `${provider}/${modelStr}: downgraded reasoning_effort ${effortStr} → high`
     );
-    const next: Record<string, unknown> = { ...b };
-    if (hasTopLevelReasoningEffort) {
-      next.reasoning_effort = "high";
-    }
-    if (reasoning) {
-      next.reasoning = { ...reasoning, effort: "high" };
-    }
-    return next;
+    return writeEffortValue(b, "high", c);
   }
 
   return body;
diff --git a/open-sse/executors/default.ts b/open-sse/executors/default.ts
index 9fad2305a46..5ce3772f8d4 100644
--- a/open-sse/executors/default.ts
+++ b/open-sse/executors/default.ts
@@ -562,9 +562,7 @@ export class DefaultExecutor extends BaseExecutor {
       withDefaults &&
       typeof withDefaults === "object" &&
       !Array.isArray(withDefaults) &&
-      (this.provider === "cerebras" ||
-        this.provider === "mistral" ||
-        this.provider === "nvidia") &&
+      (this.provider === "cerebras" || this.provider === "mistral" || this.provider === "nvidia") &&
       Object.prototype.hasOwnProperty.call(withDefaults, "client_metadata")
     ) {
       const withoutClientMetadata = { ...(withDefaults as Record<string, unknown>) };
@@ -732,10 +730,8 @@ export class DefaultExecutor extends BaseExecutor {
       }
     }
 
-    // ClinePass reasoning models burn all of max_tokens on the thinking phase
-    // when the budget is too small, leaving content empty (finish_reason:
-    // "length"). Bump max_tokens to a safe floor when reasoning is enabled and
-    // the budget is undersized. CLINEPASS-GATED — no-op for every other provider.
+    // Reasoning models burn all of max_tokens on the thinking phase when the budget is too
+    // small, leaving content empty (finish_reason: "length"); applies to all providers (#6912).
     if (typeof withDefaults === "object" && withDefaults !== null) {
       this.ensureThinkingBudget(withDefaults as Record<string, unknown>, model);
     }
@@ -760,12 +756,10 @@ export class DefaultExecutor extends BaseExecutor {
     return withDefaults;
   }
 
-  // ClinePass / OpenRouter-style thinking models leave content empty when the
-  // reasoning budget consumes all of max_tokens. Bump max_tokens to a safe
-  // minimum only when reasoning is enabled and the budget is undersized.
-  // CLINEPASS-GATED: returns early for every other provider.
+  // Reasoning models (ClinePass, OpenRouter, etc.) leave content empty when the reasoning
+  // budget consumes all of max_tokens; bump max_tokens to a safe minimum when undersized.
   ensureThinkingBudget(body: Record<string, unknown>, model: string): Record<string, unknown> {
-    if (!body || this.provider !== "clinepass") return body;
+    if (!body) return body;
 
     const outboundModel = typeof body.model === "string" ? body.model : model;
     const entry = getRegistryEntry(this.provider);
@@ -789,10 +783,15 @@ export class DefaultExecutor extends BaseExecutor {
     const target = Math.min(MIN_TOKENS, maxOutput);
     const current = body.max_tokens ?? body.max_completion_tokens;
 
+    // #6912: keep whichever token key transformRequest already set (o1/o3/o4/gpt-5 use
+    // max_completion_tokens) instead of re-introducing max_tokens alongside it.
+    const tokenKey =
+      body.max_completion_tokens !== undefined ? "max_completion_tokens" : "max_tokens";
+
     if (typeof current !== "number" || current <= 0) {
-      body.max_tokens = target;
+      body[tokenKey] = target;
     } else if (current < MIN_TOKENS && current < maxOutput) {
-      body.max_tokens = MIN_TOKENS;
+      body[tokenKey] = MIN_TOKENS;
     }
     return body;
   }
diff --git a/open-sse/executors/duckduckgo-web.ts b/open-sse/executors/duckduckgo-web.ts
index bf653932c78..0d5107013f9 100644
--- a/open-sse/executors/duckduckgo-web.ts
+++ b/open-sse/executors/duckduckgo-web.ts
@@ -8,6 +8,60 @@ import type { Session } from "../services/sessionPool/session.ts";
 import { tryBackedChat } from "../services/browserBackedChat.ts";
 import { sanitizeErrorMessage } from "../utils/error.ts";
 
+// Issue #6999: Lightweight circuit breaker for the DuckDuckGo executor.
+// After CB_THRESHOLD consecutive failures (429, 5xx, or network errors),
+// the breaker "opens" for CB_COOLDOWN_MS — during that window every request
+// fast-fails with 503 instead of hammering the upstream. A single success
+// resets the failure counter. Half-open probing happens naturally: once the
+// cooldown expires the breaker closes and the next request is a real probe.
+export const CB_THRESHOLD = 5;
+export const CB_COOLDOWN_MS = 30_000;
+
+interface CircuitBreakerState {
+  failures: number;
+  openedAt: number;
+}
+
+const circuitBreaker: CircuitBreakerState = { failures: 0, openedAt: 0 };
+
+export function cbIsOpen(): boolean {
+  if (circuitBreaker.openedAt === 0) return false;
+  if (Date.now() - circuitBreaker.openedAt >= CB_COOLDOWN_MS) {
+    // Cooldown elapsed — half-open: allow the next request through.
+    circuitBreaker.openedAt = 0;
+    return false;
+  }
+  return true;
+}
+
+export function cbRecordFailure(): void {
+  circuitBreaker.failures++;
+  if (circuitBreaker.failures >= CB_THRESHOLD && circuitBreaker.openedAt === 0) {
+    circuitBreaker.openedAt = Date.now();
+    console.warn(
+      `[DDG-CB] Circuit breaker opened after ${circuitBreaker.failures} consecutive failures — fast-failing for ${CB_COOLDOWN_MS}ms`
+    );
+  }
+}
+
+export function cbRecordSuccess(): void {
+  if (circuitBreaker.failures > 0) {
+    circuitBreaker.failures = 0;
+  }
+}
+
+// Test-only: direct read/write access to the module-level breaker singleton
+// so tests can exercise open/half-open/closed transitions without waiting
+// CB_COOLDOWN_MS in real time. Not used by production code.
+export function __setDdgCircuitBreakerStateForTests(failures: number, openedAt: number): void {
+  circuitBreaker.failures = failures;
+  circuitBreaker.openedAt = openedAt;
+}
+
+export function __getDdgCircuitBreakerStateForTests(): CircuitBreakerState {
+  return { ...circuitBreaker };
+}
+
 export const DUCKDUCKGO_BASE = "https://duckduckgo.com";
 // #4037: the live DuckDuckGo AI Chat backend is served from duckduckgo.com. The
 // status/chat fetches, Origin, and Referer must all use this host so the request's
@@ -389,6 +443,13 @@ export class DuckDuckGoWebExecutor extends BaseExecutor {
       return errorResponse(400, "No messages provided");
     }
 
+    // Issue #6999: Circuit breaker fast-fail. If DDG has been consistently
+    // failing, short-circuit with 503 so the combo engine can immediately
+    // fail over to the next provider instead of waiting for timeouts.
+    if (cbIsOpen()) {
+      return errorResponse(503, "DuckDuckGo circuit breaker open — upstream unavailable");
+    }
+
     // Browser-backed path: opt-in via OMNIROUTE_BROWSER_POOL=on or
     // WEB_COOKIE_USE_BROWSER=1. Routes the chat through a shared
     // Playwright/Cloakbrowser page so DDG's VQD challenge is solved by
@@ -508,6 +569,7 @@ export class DuckDuckGoWebExecutor extends BaseExecutor {
 
       if (chatResponse.status === 429) {
         if (pool && session) pool.reportCooldown(session);
+        cbRecordFailure();
         return await this.processResponse(chatResponse, isStreaming, hasTools, requestedTools);
       }
 
@@ -523,6 +585,7 @@ export class DuckDuckGoWebExecutor extends BaseExecutor {
 
       if (chatResponse.status >= 500) {
         if (pool && session) pool.reportDead(session);
+        cbRecordFailure();
         return errorResponse(502, "Upstream error");
       }
 
@@ -544,11 +607,13 @@ export class DuckDuckGoWebExecutor extends BaseExecutor {
         }
       }
 
+      cbRecordSuccess();
       return result;
     } catch (error) {
       if (pool && session) {
         pool.reportCooldown(session);
       }
+      cbRecordFailure();
 
       if (error instanceof DOMException && error.name === "AbortError") {
         return errorResponse(499, "Request cancelled");
diff --git a/open-sse/executors/opencode.ts b/open-sse/executors/opencode.ts
index 2453199b8f1..babae8a710d 100644
--- a/open-sse/executors/opencode.ts
+++ b/open-sse/executors/opencode.ts
@@ -41,17 +41,37 @@ const OPENCODE_COOLDOWN_MAX_MS = 60_000;
 const EFFORT_LEVELS = ["low", "medium", "high", "max"] as const;
 
 /**
- * Parse a DeepSeek V4 Pro model string with an effort-level suffix.
+ * Models on opencode-go that support effort-tier aliases. Each entry maps the
+ * canonical base id to the set of effort suffixes the upstream supports.
+ *
+ * - deepseek-v4-pro: all four tiers (low/medium/high/max)
+ * - glm-5.2: high/max only (Z.AI maps these through the reasoning plane;
+ *   low/medium are not supported on the OpenAI transport)
+ * - mimo-v2.5: high/max only (same reasoning; Xiaomi MiMo does not document
+ *   low/medium effort tiers)
+ */
+const EFFORT_TIERS: Record<string, readonly string[]> = {
+  "deepseek-v4-pro": EFFORT_LEVELS,
+  "glm-5.2": ["high", "max"],
+  "mimo-v2.5": ["high", "max"],
+};
+
+/**
+ * Parse a model string with an effort-level suffix.
  * e.g. "deepseek-v4-pro-low" → { baseModel: "deepseek-v4-pro", effort: "low" }
- * Returns null if the model doesn't match the pattern.
+ *      "glm-5.2-high"         → { baseModel: "glm-5.2", effort: "high" }
+ * Returns null if the model doesn't match any known effort-tier pattern.
  */
-function parseDeepSeekEffortLevel(model: string): { baseModel: string; effort: string } | null {
+export function parseEffortLevel(model: string): { baseModel: string; effort: string } | null {
   const m = String(model || "");
-  const matchedLevel = EFFORT_LEVELS.find((level) => m.endsWith(`-${level}`));
-  if (!matchedLevel) return null;
-  const baseModel = m.slice(0, -matchedLevel.length - 1);
-  if (baseModel.toLowerCase() !== "deepseek-v4-pro") return null;
-  return { baseModel: "deepseek-v4-pro", effort: matchedLevel };
+  for (const [baseModel, levels] of Object.entries(EFFORT_TIERS)) {
+    for (const level of levels) {
+      if (m === `${baseModel}-${level}`) {
+        return { baseModel, effort: level };
+      }
+    }
+  }
+  return null;
 }
 
 export class OpencodeExecutor extends BaseExecutor {
@@ -316,7 +336,7 @@ export class OpencodeExecutor extends BaseExecutor {
     }
     if (modifiedBody && typeof modifiedBody === "object" && !Array.isArray(modifiedBody)) {
       const mb = modifiedBody as Record<string, unknown>;
-      const parsed = parseDeepSeekEffortLevel(model);
+      const parsed = parseEffortLevel(model);
       if (parsed) {
         mb.model = parsed.baseModel;
         if (mb.reasoning_effort === undefined) {
diff --git a/open-sse/executors/theoldllm.ts b/open-sse/executors/theoldllm.ts
index 64e985b96c3..8acc471b3a9 100644
--- a/open-sse/executors/theoldllm.ts
+++ b/open-sse/executors/theoldllm.ts
@@ -108,6 +108,23 @@ export function mapModel(model: string): string {
 const TOKEN_SEED = "oldllm-client-2026";
 const UA_PREFIX = CHROME_UA.slice(0, 20); // "Mozilla/5.0 (Windows"
 
+type TheOldLlmProxy = {
+  type?: string;
+  host: string;
+  port: number;
+  username?: string | null;
+  password?: string | null;
+} | null;
+
+interface TheOldLlmFetchDependencies {
+  resolveProxy: () => Promise<TheOldLlmProxy>;
+  runWithProxy: <T>(proxy: TheOldLlmProxy, request: () => Promise<T>) => Promise<T>;
+  fetch: typeof fetch;
+  hasBlockingProxyAssignment?: () => boolean;
+}
+
+class TheOldLlmProxyUnavailableError extends Error {}
+
 export function generateRequestToken(): string {
   const n = Date.now();
   const e = `${n}-${TOKEN_SEED}-${UA_PREFIX}`;
@@ -127,17 +144,31 @@ export const tokenCache: { value: string; expiresAt: number } = { value: "", exp
 
 // ── Direct Node.js fetch ──────────────────────────────────────────────────
 
-async function directFetch(
+export async function fetchTheOldLlmWithProviderProxy(
   reqBody: Record<string, unknown>,
-  signal?: AbortSignal | null
+  signal: AbortSignal,
+  dependencies?: TheOldLlmFetchDependencies
 ): Promise<Response> {
-  const controller = new AbortController();
-  const timer = setTimeout(() => controller.abort(), 120_000);
-  const onSignal = signal ? () => controller.abort(signal.reason) : undefined;
-  signal?.addEventListener("abort", onSignal!, { once: true });
+  let deps = dependencies;
+  if (!deps) {
+    const [
+      { resolveProxyForProvider, hasBlockingProxyAssignmentForProvider },
+      { runWithProxyContext },
+    ] = await Promise.all([import("../../src/lib/db/proxies"), import("../utils/proxyFetch.ts")]);
+    deps = {
+      resolveProxy: () => resolveProxyForProvider("theoldllm"),
+      runWithProxy: runWithProxyContext,
+      fetch: globalThis.fetch,
+      hasBlockingProxyAssignment: () => hasBlockingProxyAssignmentForProvider("theoldllm"),
+    };
+  }
 
-  try {
-    return await fetch(API_URL, {
+  const proxy = await deps.resolveProxy();
+  if (!proxy && deps.hasBlockingProxyAssignment?.()) {
+    throw new TheOldLlmProxyUnavailableError("No active proxy is available for The Old LLM");
+  }
+  return deps.runWithProxy(proxy, () =>
+    deps.fetch(API_URL, {
       method: "POST",
       headers: {
         "Content-Type": "application/json",
@@ -146,14 +177,41 @@ async function directFetch(
         "User-Agent": CHROME_UA,
       },
       body: JSON.stringify(reqBody),
-      signal: controller.signal,
-    });
+      signal,
+    })
+  );
+}
+
+async function directFetch(
+  reqBody: Record<string, unknown>,
+  signal?: AbortSignal | null
+): Promise<Response> {
+  const controller = new AbortController();
+  const timer = setTimeout(() => controller.abort(), 120_000);
+  const onSignal = signal ? () => controller.abort(signal.reason) : undefined;
+  signal?.addEventListener("abort", onSignal!, { once: true });
+
+  try {
+    // No-auth providers do not have a connection row, so chatCore cannot apply
+    // a connection-scoped proxy context for them. Resolve the provider/global
+    // assignment explicitly; otherwise The Old LLM always leaks out through the
+    // VPS address and Vercel's bot protection denies every model.
+    return await fetchTheOldLlmWithProviderProxy(reqBody, controller.signal);
   } finally {
     clearTimeout(timer);
     if (onSignal) signal?.removeEventListener("abort", onSignal);
   }
 }
 
+export function isVercelMitigationResponse(response: Response, body: string): boolean {
+  const mitigation = response.headers.get("x-vercel-mitigated")?.toLowerCase();
+  if (mitigation === "deny" || mitigation === "challenge") return true;
+  return (
+    (response.status === 403 || response.status === 429) &&
+    /vercel security checkpoint|"message"\s*:\s*"forbidden"/i.test(body)
+  );
+}
+
 function isTokenRejected(status: number, body: string): boolean {
   if (status === 401 || status === 403) return true;
   try {
@@ -211,6 +269,45 @@ function buildErrorResponse(status: number, body: string): string {
   });
 }
 
+function buildVercelMitigationError(): string {
+  return JSON.stringify({
+    error: {
+      message:
+        "The Old LLM is blocked by Vercel for this server egress IP. Configure a residential provider or global proxy for 'theoldllm' and retry.",
+      type: "upstream_access_denied",
+      code: "THEOLDLLM_VERCEL_MITIGATED",
+    },
+  });
+}
+
+function buildProxyUnavailableError(): string {
+  return JSON.stringify({
+    error: {
+      message:
+        "The Old LLM proxy assignment has no active proxies. Configure or enable a proxy and retry.",
+      type: "proxy_unavailable",
+      code: "THEOLDLLM_PROXY_UNAVAILABLE",
+    },
+  });
+}
+
+async function fetchUpstreamWithRetry(
+  reqBody: Record<string, unknown>,
+  signal: AbortSignal | null | undefined,
+  log: ExecuteInput["log"]
+): Promise<{ response: Response; body: string; vercelMitigated: boolean }> {
+  let response = await directFetch(reqBody, signal);
+  let body = await response.text();
+  let vercelMitigated = isVercelMitigationResponse(response, body);
+  if (!vercelMitigated && isTokenRejected(response.status, body)) {
+    log?.warn?.("THEOLDLLM", `Token rejected (${response.status}), retrying with fresh token…`);
+    response = await directFetch(reqBody, signal);
+    body = await response.text();
+    vercelMitigated = isVercelMitigationResponse(response, body);
+  }
+  return { response, body, vercelMitigated };
+}
+
 // ── Executor ──────────────────────────────────────────────────────────────
 
 export class TheOldLlmExecutor extends BaseExecutor {
@@ -237,27 +334,37 @@ export class TheOldLlmExecutor extends BaseExecutor {
     return body;
   }
 
+  private executionResult(input: ExecuteInput, response: Response, body: unknown) {
+    return {
+      response,
+      url: API_URL,
+      headers: this.buildHeaders(input.credentials),
+      transformedBody: body,
+    };
+  }
+
   async testConnection(
     _credentials: ProviderCredentials,
     _signal?: AbortSignal | null,
     log?: ExecuteInput["log"]
   ): Promise<boolean> {
     try {
-      const resp = await fetch(API_URL, {
-        method: "POST",
-        headers: {
-          "Content-Type": "application/json",
-          "X-Client-Version": "3.8.4",
-          "X-Request-Token": generateRequestToken(),
-          "User-Agent": CHROME_UA,
-        },
-        body: JSON.stringify({
+      const resp = await directFetch(
+        {
           model: "GPT_5_4",
           messages: [{ role: "user", content: "ping" }],
           stream: false,
-        }),
-        signal: _signal ?? undefined,
-      });
+        },
+        _signal
+      );
+      const body = await resp.text();
+      if (!resp.ok && isVercelMitigationResponse(resp, body)) {
+        log?.warn?.(
+          "THEOLDLLM",
+          "Vercel blocked this egress IP; configure a residential provider proxy"
+        );
+        return false;
+      }
       return resp.status === 200;
     } catch {
       log?.warn?.("THEOLDLLM", "testConnection network error");
@@ -297,56 +404,55 @@ export class TheOldLlmExecutor extends BaseExecutor {
         stream: true,
       };
 
-      let upstream = await directFetch(reqBody, signal);
-      let finalBody = await upstream.text();
-
-      if (isTokenRejected(upstream.status, finalBody)) {
-        log?.warn?.("THEOLDLLM", `Token rejected (${upstream.status}), retrying with fresh token…`);
-        upstream = await directFetch(reqBody, signal);
-        finalBody = await upstream.text();
-      }
+      const {
+        response: upstream,
+        body: finalBody,
+        vercelMitigated,
+      } = await fetchUpstreamWithRetry(reqBody, signal, log);
 
       if (upstream.status === 200 && finalBody) {
         const payload = stream ? finalBody : buildChatCompletion(parseSseContent(finalBody), model);
-        return {
-          response: new Response(encoder.encode(payload), {
+        return this.executionResult(
+          input,
+          new Response(encoder.encode(payload), {
             status: 200,
             headers: {
               "Content-Type": stream ? "text/event-stream" : "application/json",
               "Cache-Control": "no-cache",
             },
           }),
-          url: API_URL,
-          headers: this.buildHeaders(input.credentials),
-          transformedBody: body,
-        };
+          body
+        );
       }
 
-      return {
-        response: new Response(encoder.encode(buildErrorResponse(upstream.status, finalBody)), {
+      const errorPayload = vercelMitigated
+        ? buildVercelMitigationError()
+        : buildErrorResponse(upstream.status, finalBody);
+      return this.executionResult(
+        input,
+        new Response(encoder.encode(errorPayload), {
           status: upstream.status,
           headers: { "Content-Type": "application/json" },
         }),
-        url: API_URL,
-        headers: this.buildHeaders(input.credentials),
-        transformedBody: body,
-      };
+        body
+      );
     } catch (err) {
+      const proxyUnavailable = err instanceof TheOldLlmProxyUnavailableError;
       const msg = err instanceof Error ? err.message : String(err);
       log?.error?.("THEOLDLLM", `Executor error: ${msg}`);
-      return {
-        response: new Response(
-          encoder.encode(
-            JSON.stringify({
-              error: { message: msg, type: "upstream_error", code: "EXECUTOR_ERROR" },
-            })
-          ),
-          { status: 502, headers: { "Content-Type": "application/json" } }
-        ),
-        url: API_URL,
-        headers: this.buildHeaders(input.credentials),
-        transformedBody: body,
-      };
+      const errorPayload = proxyUnavailable
+        ? buildProxyUnavailableError()
+        : JSON.stringify({
+            error: { message: msg, type: "upstream_error", code: "EXECUTOR_ERROR" },
+          });
+      return this.executionResult(
+        input,
+        new Response(encoder.encode(errorPayload), {
+          status: proxyUnavailable ? 503 : 502,
+          headers: { "Content-Type": "application/json" },
+        }),
+        body
+      );
     }
   }
 }
diff --git a/open-sse/handlers/chatCore.ts b/open-sse/handlers/chatCore.ts
index 83a3eb71520..abcd3a9c0bf 100644
--- a/open-sse/handlers/chatCore.ts
+++ b/open-sse/handlers/chatCore.ts
@@ -14,6 +14,7 @@ import { buildPostCallGuardrailContext } from "./chatCore/postCallGuardrailConte
 import { storeSemanticCacheResponse } from "./chatCore/semanticCacheStore.ts";
 import { buildNonStreamingResponseHeaders } from "./chatCore/nonStreamingResponseHeaders.ts";
 import { buildNonStreamingJsonResponse } from "./chatCore/nonStreamingJsonResponse.ts";
+import { enforceOutputTokenBudget } from "./chatCore/outputTokenBudget.ts";
 import { maybeConvertJsonBodyToSse } from "./chatCore/jsonBodyToSse.ts";
 import { assembleStreamingResponseHeaders } from "./chatCore/streamingResponseHeaders.ts";
 import { storeStreamingSemanticCacheResponse } from "./chatCore/streamingSemanticCacheStore.ts";
@@ -125,7 +126,11 @@ import {
   formatProviderError,
   sanitizeErrorMessage,
 } from "../utils/error.ts";
-import { reportMalformed200, detectMalformedNonStream } from "../utils/diagnostics.ts";
+import {
+  reportMalformed200,
+  detectMalformedNonStream,
+  describeMalformedNonStream,
+} from "../utils/diagnostics.ts";
 import {
   checkTokenLimits,
   recordTokenUsage,
@@ -140,6 +145,7 @@ import {
   STREAM_READINESS_TIMEOUT_MS,
   ANTIGRAVITY_PRE_RESPONSE_TIMEOUT_CODE,
   STREAM_RECOVERY,
+  DEFAULT_MAX_TOKENS,
 } from "../config/constants.ts";
 import { createRecoverableStream, makeContinuationBody } from "../services/streamRecovery.ts";
 import {
@@ -301,7 +307,12 @@ import {
   resolveComboContextLimit,
 } from "../services/contextManager.ts";
 import { resolveBackgroundTaskRedirect } from "./chatCore/backgroundRedirect.ts";
-import type { CompressionConfig, CompressionPipelineStep } from "../services/compression/types.ts";
+import type {
+  CompressionConfig,
+  CompressionPipelineStep,
+  CompressionResult,
+} from "../services/compression/types.ts";
+import { generateSessionId } from "../services/sessionManager.ts";
 import { prepareWebSearchFallbackBody } from "../services/webSearchFallback.ts";
 import { resolveInterceptSearch } from "@/lib/db/interceptionRules";
 import {
@@ -1334,7 +1345,8 @@ export async function handleChatCore({
         // that selectCompressionStrategy can only partially apply via the mode string.
         const cacheCtx = { provider, targetFormat, model: effectiveModel, connectionCacheOverride };
         const compressionConfig = resolveCacheAwareConfig(config, compressionInputBody, cacheCtx);
-        const result = await applyCompressionAsync(compressionInputBody, mode, {
+        const compressionPrincipalId = apiKeyInfo?.id ? String(apiKeyInfo.id) : undefined;
+        const compressionOptions = {
           model: effectiveModel,
           // #7237: feed the AUTHORITATIVE capability (model spec / models.dev sync / DB
           // override, with the conservative model-id fragment heuristic only as its
@@ -1348,10 +1360,11 @@ export async function handleChatCore({
             .supportsVision,
           // Rota direta oficial ('anthropic') vs agregadores: o engine omniglyph
           // exige 'direct' — agregadores redimensionam imagens (medido 2026-07-06).
-          providerTransport: provider === "anthropic" ? "direct" : "aggregator",
+          providerTransport:
+            provider === "anthropic" ? ("direct" as const) : ("aggregator" as const),
           config: compressionConfig,
           cachingContext: cacheCtx,
-          principalId: apiKeyInfo?.id ? String(apiKeyInfo.id) : undefined,
+          principalId: compressionPrincipalId,
           // F3.3: stream per-engine progress live (best-effort) before compression.completed.
           onEngineStep: (s) => {
             try {
@@ -1375,7 +1388,52 @@ export async function handleChatCore({
               // best-effort live event — never fail the request
             }
           },
-        });
+        };
+        const runCompression = (input: Record<string, unknown>) =>
+          applyCompressionAsync(input, mode, compressionOptions);
+        let result: CompressionResult;
+        if (compressionConfig.liveZone?.enabled === true) {
+          const { applyLiveZoneCompression } = await import("../services/compression/liveZone.ts");
+          const explicitSessionId =
+            clientRawRequest?.headers && typeof clientRawRequest.headers.get === "function"
+              ? clientRawRequest.headers.get("x-omniroute-session-id")
+              : getHeaderValueCaseInsensitive(
+                  clientRawRequest?.headers ?? null,
+                  "x-omniroute-session-id"
+                );
+          const liveZoneSessionId =
+            explicitSessionId ||
+            generateSessionId(compressionInputBody, {
+              provider,
+              connectionId: getCurrentConnectionId() ?? undefined,
+            }) ||
+            undefined;
+          result = await applyLiveZoneCompression(
+            compressionInputBody,
+            {
+              principalId: compressionPrincipalId,
+              sessionId: liveZoneSessionId,
+              variant: {
+                mode,
+                provider,
+                model: effectiveModel,
+                config: compressionConfig,
+                cachePrefix: {
+                  system: compressionInputBody.system,
+                  systemInstruction: compressionInputBody.systemInstruction,
+                  system_instruction: compressionInputBody.system_instruction,
+                  instructions: compressionInputBody.instructions,
+                  tools: compressionInputBody.tools,
+                  toolChoice: compressionInputBody.tool_choice,
+                },
+              },
+              ttlMinutes: compressionConfig.cacheMinutes,
+            },
+            runCompression
+          );
+        } else {
+          result = await runCompression(compressionInputBody);
+        }
         if (result.stats) {
           const annotation = formatCompressionAnnotation(result.stats);
           if (annotation) {
@@ -1441,6 +1499,7 @@ export async function handleChatCore({
               cavemanOutputModeIntensity,
               log,
             });
+            await compressionAnalyticsWritePromise;
           } else {
             // Compression was attempted (mode active, engines ran) but produced no
             // recordable saving — e.g. a Stacked RTK→Caveman pipeline on already-compact
@@ -1463,6 +1522,7 @@ export async function handleChatCore({
               },
               "no_savings"
             );
+            await compressionAnalyticsWritePromise;
           }
 
           if (result.compressed) {
@@ -1493,6 +1553,7 @@ export async function handleChatCore({
           cavemanOutputModeIntensity,
           log,
         });
+        await compressionAnalyticsWritePromise;
       }
       emitOutputStyleTelemetry({
         outputStyleResult,
@@ -1632,6 +1693,56 @@ export async function handleChatCore({
     );
   }
 
+  // Re-check the concrete target after all compression passes. Combo compatibility
+  // filtering is advisory and may preserve an all-incompatible pool; this is the
+  // hard boundary that prevents a too-large prompt (or a negative token budget)
+  // from reaching an OpenAI-compatible upstream such as NVIDIA NIM.
+  const finalCompressionBody = body
+    ? adaptBodyForCompression(body as Record<string, unknown>).body
+    : null;
+  const finalMessages =
+    finalCompressionBody?.messages ||
+    body?.contents ||
+    body?.request?.contents ||
+    (body?.input && typeof body.input === "object" && !Array.isArray(body.input)
+      ? body.input
+      : []);
+  const finalEstimatedInputTokens =
+    estimateTokens(finalMessages) +
+    (Array.isArray(body?.tools) ? estimateTokens(body.tools) : 0) +
+    estimateTokens(body?.system) +
+    estimateTokens(body?.instructions);
+  const finalContextLimit = getTokenLimit(provider, effectiveModel);
+  const outputBudget = enforceOutputTokenBudget(
+    body as Record<string, unknown>,
+    finalEstimatedInputTokens,
+    finalContextLimit,
+    targetFormat === FORMATS.CLAUDE && sourceFormat !== FORMATS.CLAUDE ? DEFAULT_MAX_TOKENS : 0
+  );
+  if (!outputBudget.ok) {
+    const message =
+      `Input exceeds the context window for ${provider}/${effectiveModel}: ` +
+      `estimated ${outputBudget.estimatedInputTokens} input tokens, limit ${outputBudget.contextLimit}. ` +
+      "Reduce the prompt or route to a model with a larger context window.";
+    log?.warn?.("CONTEXT", message);
+    trackPendingRequest(model, provider, connectionId, false);
+    return createErrorResult(
+      HTTP_STATUS.BAD_REQUEST,
+      message,
+      null,
+      "context_length_exceeded",
+      "invalid_request_error"
+    );
+  }
+  if (outputBudget.adjustedFields.length > 0) {
+    log?.info?.(
+      "CONTEXT",
+      `Adjusted invalid or oversized output token fields (${outputBudget.adjustedFields.join(", ")}); ` +
+        `${outputBudget.availableOutputTokens} tokens remain for output`
+    );
+  }
+  body = outputBudget.body;
+
   let translatedBody = body;
   const isClaudePassthrough = sourceFormat === FORMATS.CLAUDE && targetFormat === FORMATS.CLAUDE;
   const isClaudeCodeCompatible = isClaudeCodeCompatibleProvider(provider);
@@ -3384,7 +3495,13 @@ export async function handleChatCore({
           // otherwise degenerate into a 429 rate-limit storm). Connection stays
           // active since only the specific model is unavailable. (#6827)
           const notFoundCooldownMs = COOLDOWN_MS.notFound;
-          lockModel(provider, errorConnectionId, currentModel, "model_not_found", notFoundCooldownMs);
+          lockModel(
+            provider,
+            errorConnectionId,
+            currentModel,
+            "model_not_found",
+            notFoundCooldownMs
+          );
           console.warn(
             `[provider] Node ${errorConnectionId} model not found (${statusCode}) for ${currentModel} - locking model for ${Math.ceil(notFoundCooldownMs / 1000)}s (connection stays active)`
           );
@@ -4070,7 +4187,11 @@ export async function handleChatCore({
         connectionId,
         status: `FAILED ${HTTP_STATUS.BAD_GATEWAY}`,
       }).catch(() => {});
-      const malformedMessage = `[${provider}/${model}] returned an empty response (no usable choices/output)`;
+      const malformed = describeMalformedNonStream(translatedResponse, malformedTranslatedReason);
+      const malformedMessage = `[${provider}/${model}] ${malformed.message}`;
+      const malformedClientBody = buildErrorBody(HTTP_STATUS.BAD_GATEWAY, malformedMessage);
+      malformedClientBody.error.code = malformed.code;
+      malformedClientBody.error.type = malformed.type;
       persistAttemptLogs({
         status: HTTP_STATUS.BAD_GATEWAY,
         tokens: usage,
@@ -4079,14 +4200,20 @@ export async function handleChatCore({
         providerResponse: looksLikeSSE
           ? { _streamed: true, _format: "sse-json", summary: responseBody }
           : responseBody,
-        clientResponse: buildErrorBody(HTTP_STATUS.BAD_GATEWAY, malformedMessage),
+        clientResponse: malformedClientBody,
         claudeCacheMeta: claudePromptCacheLogMeta,
         claudeCacheUsageMeta: cacheUsageLogMeta,
         cacheSource: "upstream",
       });
       persistFailureUsage(HTTP_STATUS.BAD_GATEWAY, "malformed_translated_response");
       trackPendingRequest(model, provider, pendingConnId, false);
-      return createErrorResult(HTTP_STATUS.BAD_GATEWAY, malformedMessage);
+      return createErrorResult(
+        HTTP_STATUS.BAD_GATEWAY,
+        malformedMessage,
+        null,
+        malformed.code,
+        malformed.type
+      );
     }
 
     // ── Phase 9.1: Cache store (non-streaming, temp=0) ──
diff --git a/open-sse/handlers/chatCore/cavemanOutputAnalytics.ts b/open-sse/handlers/chatCore/cavemanOutputAnalytics.ts
index fa3a778d9cd..18307628ce6 100644
--- a/open-sse/handlers/chatCore/cavemanOutputAnalytics.ts
+++ b/open-sse/handlers/chatCore/cavemanOutputAnalytics.ts
@@ -5,11 +5,11 @@
  * Extracted from handleChatCore's request-setup compression path: when only the caveman output
  * mode was applied (no upstream compression run recorded a row), persist a single analytics row so
  * output-caveman runs still surface in compression analytics. Best-effort — returns the write
- * promise (the caller assigns it to compressionAnalyticsWritePromise) and swallows its own errors.
- * Behaviour is byte-identical to the previous inline block.
+ * promise so the caller can finish persistence before dispatch; errors remain non-fatal but are
+ * logged at warning level.
  */
 
-type LoggerLike = { debug?: (...args: unknown[]) => void } | null | undefined;
+type LoggerLike = { warn?: (...args: unknown[]) => void } | null | undefined;
 
 export function writeCavemanOutputAnalytics(args: {
   comboName: string | null | undefined;
@@ -37,7 +37,7 @@ export function writeCavemanOutputAnalytics(args: {
         output_mode: args.cavemanOutputModeIntensity,
       });
     } catch (err) {
-      args.log?.debug?.(
+      args.log?.warn?.(
         "COMPRESSION",
         "Caveman output analytics write skipped: " +
           (err instanceof Error ? err.message : String(err))
diff --git a/open-sse/handlers/chatCore/compressionAnalyticsWrite.ts b/open-sse/handlers/chatCore/compressionAnalyticsWrite.ts
index 21b21b33c02..ec68aff23cc 100644
--- a/open-sse/handlers/chatCore/compressionAnalyticsWrite.ts
+++ b/open-sse/handlers/chatCore/compressionAnalyticsWrite.ts
@@ -4,15 +4,21 @@
  *
  * Extracted from handleChatCore's request-setup compression path: persist the per-run compression
  * analytics row (cost saved, RTK raw-output pointers) plus the per-engine breakdown of a stacked
- * run. Returns the write promise (the caller assigns it to compressionAnalyticsWritePromise) and
- * swallows its own errors — best-effort, off the hot path, never throws into a request. Behaviour
- * is byte-identical to the previous inline block. Split into small builders so each stays under the
- * complexity cap.
+ * run. Returns the write promise so the caller can finish persistence before dispatching the
+ * upstream request. Errors remain best-effort and never throw into a request, but they are logged
+ * at warning level so a broken analytics path is observable. Split into small builders so each
+ * stays under the complexity cap.
  */
 
 import { type CompressionStats } from "../../services/compression/stats.ts";
 
-type LoggerLike = { debug?: (...args: unknown[]) => void } | null | undefined;
+type LoggerLike =
+  | {
+      debug?: (...args: unknown[]) => void;
+      warn?: (...args: unknown[]) => void;
+    }
+  | null
+  | undefined;
 
 type WriteOpts = {
   stats: CompressionStats;
@@ -30,6 +36,12 @@ type WriteOpts = {
 
 type RtkPointer = { id?: string | null; bytes?: number | null };
 
+type CalculateCost = typeof import("@/lib/usage/costCalculator").calculateCost;
+
+type WriteDependencies = {
+  calculateCost?: CalculateCost;
+};
+
 function buildRtkPointerFields(rtkPointers: RtkPointer[]) {
   return {
     rtk_raw_output_pointer: rtkPointers[0]?.id ?? null,
@@ -109,7 +121,7 @@ export function writeCompressionSkip(opts: WriteOpts, skipReason: string): Promi
         skip_reason: skipReason,
       });
     } catch (err) {
-      opts.log?.debug?.(
+      opts.log?.warn?.(
         "COMPRESSION",
         "Compression skip-analytics write skipped: " +
           (err instanceof Error ? err.message : String(err))
@@ -118,29 +130,42 @@ export function writeCompressionSkip(opts: WriteOpts, skipReason: string): Promi
   })();
 }
 
-export function writeCompressionAnalytics(opts: WriteOpts): Promise<void> {
+export function writeCompressionAnalytics(
+  opts: WriteOpts,
+  dependencies: WriteDependencies = {}
+): Promise<void> {
   return (async () => {
     try {
-      const { insertCompressionAnalyticsRow, insertCompressionEngineBreakdown } = await import(
-        "@/lib/db/compressionAnalytics"
-      );
-      const { calculateCost } = await import("@/lib/usage/costCalculator");
+      const { insertCompressionAnalyticsRow, insertCompressionEngineBreakdown } =
+        await import("@/lib/db/compressionAnalytics");
       const { stats } = opts;
       const tokensSaved = Math.max(0, stats.originalTokens - stats.compressedTokens);
       const rtkPointers = (stats.rtkRawOutputPointers ?? []) as RtkPointer[];
-      const estimatedUsdSaved = await calculateCost(
-        opts.provider ?? "",
-        opts.effectiveModel ?? "",
-        { input: tokensSaved },
-        { serviceTier: opts.effectiveServiceTier }
+      let estimatedUsdSaved = 0;
+      try {
+        const calculateCost =
+          dependencies.calculateCost ?? (await import("@/lib/usage/costCalculator")).calculateCost;
+        estimatedUsdSaved = await calculateCost(
+          opts.provider ?? "",
+          opts.effectiveModel ?? "",
+          { input: tokensSaved },
+          { serviceTier: opts.effectiveServiceTier }
+        );
+      } catch (err) {
+        opts.log?.debug?.(
+          "COMPRESSION",
+          "Compression cost estimate skipped: " + (err instanceof Error ? err.message : String(err))
+        );
+      }
+      insertCompressionAnalyticsRow(
+        buildAnalyticsRow(opts, tokensSaved, rtkPointers, estimatedUsdSaved)
       );
-      insertCompressionAnalyticsRow(buildAnalyticsRow(opts, tokensSaved, rtkPointers, estimatedUsdSaved));
       const breakdownRows = buildEngineBreakdownRows(stats, opts.skillRequestId);
       if (breakdownRows.length > 0) {
         insertCompressionEngineBreakdown(breakdownRows);
       }
     } catch (err) {
-      opts.log?.debug?.(
+      opts.log?.warn?.(
         "COMPRESSION",
         "Compression analytics write skipped: " + (err instanceof Error ? err.message : String(err))
       );
diff --git a/open-sse/handlers/chatCore/outputTokenBudget.ts b/open-sse/handlers/chatCore/outputTokenBudget.ts
new file mode 100644
index 00000000000..2d26d488c2a
--- /dev/null
+++ b/open-sse/handlers/chatCore/outputTokenBudget.ts
@@ -0,0 +1,117 @@
+export const OUTPUT_TOKEN_FIELDS = [
+  "max_tokens",
+  "max_completion_tokens",
+  "max_output_tokens",
+] as const;
+
+export type OutputTokenBudgetResult =
+  | {
+      ok: true;
+      body: Record<string, unknown>;
+      availableOutputTokens: number;
+      adjustedFields: string[];
+    }
+  | {
+      ok: false;
+      estimatedInputTokens: number;
+      contextLimit: number;
+    };
+
+type OutputTokenAdjustment = { field: string; value?: number; remove?: boolean };
+
+function getOutputTokenAdjustment(
+  field: string,
+  value: unknown,
+  availableOutputTokens: number
+): OutputTokenAdjustment | null {
+  if (typeof value !== "number") return null;
+  if (!Number.isFinite(value) || value <= 0) return { field, remove: true };
+
+  const capped = Math.min(Math.floor(value), availableOutputTokens);
+  return capped === value ? null : { field, value: capped };
+}
+
+function hasTranslatorOutputTokenLimit(body: Record<string, unknown>): boolean {
+  return ["max_tokens", "max_completion_tokens"].some((field) => {
+    const value = body[field];
+    return typeof value === "number" && Number.isFinite(value) && value > 0;
+  });
+}
+
+function adjustOutputTokenFields(
+  body: Record<string, unknown>,
+  availableOutputTokens: number
+): Pick<Extract<OutputTokenBudgetResult, { ok: true }>, "body" | "adjustedFields"> {
+  const adjustments = OUTPUT_TOKEN_FIELDS.map((field) =>
+    getOutputTokenAdjustment(field, body[field], availableOutputTokens)
+  ).filter((adjustment): adjustment is OutputTokenAdjustment => adjustment !== null);
+  if (adjustments.length === 0) return { body, adjustedFields: [] };
+
+  const nextBody = { ...body };
+  for (const adjustment of adjustments) {
+    if (adjustment.remove) delete nextBody[adjustment.field];
+    else nextBody[adjustment.field] = adjustment.value;
+  }
+
+  return { body: nextBody, adjustedFields: adjustments.map(({ field }) => field) };
+}
+
+/**
+ * Enforce the target model's context budget immediately before translation.
+ *
+ * Compression and combo selection are best-effort: a request may still be too
+ * large for a concrete target, and some OpenAI-compatible gateways derive an
+ * internal max_tokens value by subtracting the prompt from the context window.
+ * Reject that target locally instead of allowing the derived value to become
+ * negative upstream. Positive client limits are capped to the remaining room;
+ * invalid numeric limits are removed.
+ */
+export function enforceOutputTokenBudget(
+  body: Record<string, unknown> | null | undefined,
+  estimatedInputTokens: number,
+  contextLimit: number,
+  defaultOutputTokens = 0
+): OutputTokenBudgetResult {
+  const normalizedInputTokens = Math.max(0, Math.ceil(estimatedInputTokens));
+  const normalizedContextLimit = Math.max(1, Math.floor(contextLimit));
+  const normalizedDefaultOutputTokens = Math.max(0, Math.floor(defaultOutputTokens));
+  const availableOutputTokens = normalizedContextLimit - normalizedInputTokens;
+
+  if (availableOutputTokens < 1) {
+    return {
+      ok: false,
+      estimatedInputTokens: normalizedInputTokens,
+      contextLimit: normalizedContextLimit,
+    };
+  }
+
+  if (!body) {
+    if (normalizedDefaultOutputTokens > availableOutputTokens) {
+      return {
+        ok: false,
+        estimatedInputTokens: normalizedInputTokens,
+        contextLimit: normalizedContextLimit,
+      };
+    }
+    return {
+      ok: true,
+      body: {},
+      availableOutputTokens,
+      adjustedFields: [],
+    };
+  }
+
+  if (
+    normalizedDefaultOutputTokens > availableOutputTokens &&
+    !hasTranslatorOutputTokenLimit(body)
+  ) {
+    return {
+      ok: false,
+      estimatedInputTokens: normalizedInputTokens,
+      contextLimit: normalizedContextLimit,
+    };
+  }
+
+  const adjusted = adjustOutputTokenFields(body, availableOutputTokens);
+  return { ok: true, ...adjusted, availableOutputTokens };
+}
diff --git a/open-sse/handlers/chatCore/streamingPipeline.ts b/open-sse/handlers/chatCore/streamingPipeline.ts
index d67803918b1..2a6a7c00bbb 100644
--- a/open-sse/handlers/chatCore/streamingPipeline.ts
+++ b/open-sse/handlers/chatCore/streamingPipeline.ts
@@ -23,6 +23,17 @@ import { createPiiSseTransform as defaultPiiSse } from "@/lib/streamingPiiTransf
 import { isFeatureFlagEnabled as defaultFeatureFlag } from "@/shared/utils/featureFlags";
 import { OMNIROUTE_RESPONSE_HEADERS } from "@/shared/constants/headers";
 import { SSE_HEARTBEAT_INTERVAL_MS } from "../../config/constants.ts";
+/**
+ * Pipeline assembly instrumentation — performance.mark() along the SSE hot path.
+ * Marks are visible to Node.js perf_hooks consumers and DevTools' Performance
+ * panel when NODE_OPTIONS=--enable-node-performance-clinician or similar.
+ *
+ * Each call to assembleStreamingPipeline creates one measure record:
+ *   "omni-pipeline" — wall-clock duration of the full transform chain assembly.
+ */
+const PIPELINE_START = "omni-pipeline-start";
+const PIPELINE_END = "omni-pipeline-end";
+const PIPELINE_MEASURE = "omni-pipeline";
 
 type HeadersLike = Headers | Record<string, unknown> | null | undefined;
 
@@ -61,6 +72,10 @@ export function assembleStreamingPipeline(
   },
   deps: StreamingPipelineDeps = DEFAULT_DEPS
 ) {
+  performance.clearMarks(PIPELINE_START);
+  performance.clearMarks(PIPELINE_END);
+  performance.clearMeasures(PIPELINE_MEASURE);
+  performance.mark(PIPELINE_START);
   // ── Phase 9.3: Progress tracking (opt-in) ──
   const progressEnabled = deps.wantsProgress(args.clientRawRequestHeaders);
   let finalStream;
@@ -77,7 +92,9 @@ export function assembleStreamingPipeline(
   }
 
   if (progressEnabled) {
-    const progressTransform = deps.createProgressTransform({ signal: args.streamController.signal });
+    const progressTransform = deps.createProgressTransform({
+      signal: args.streamController.signal,
+    });
     // Chain: provider → transform → progress → client
     finalStream = piiStream.pipeThrough(progressTransform);
     args.responseHeaders[OMNIROUTE_RESPONSE_HEADERS.progress] = "enabled";
@@ -95,5 +112,7 @@ export function assembleStreamingPipeline(
   if (args.echoModel) {
     finalStream = finalStream.pipeThrough(deps.createModelEchoTransform(args.echoModel));
   }
+  performance.mark(PIPELINE_END);
+  performance.measure(PIPELINE_MEASURE, PIPELINE_START, PIPELINE_END);
   return finalStream;
 }
diff --git a/open-sse/handlers/imageGeneration.ts b/open-sse/handlers/imageGeneration.ts
index 9e0adbca126..c0d3de0ec1f 100644
--- a/open-sse/handlers/imageGeneration.ts
+++ b/open-sse/handlers/imageGeneration.ts
@@ -54,6 +54,7 @@ import { handleHyperbolicImageGeneration } from "./imageGeneration/providers/hyp
 import { handleHuggingFaceImageGeneration } from "./imageGeneration/providers/huggingface.ts";
 import { handleComfyUIImageGeneration } from "./imageGeneration/providers/comfyUI.ts";
 import { handleImagen3ImageGeneration } from "./imageGeneration/providers/imagen3.ts";
+import { handleGoogleImagenGeneration } from "./imageGeneration/providers/googleImagen.ts";
 import { handleIdeogramImageGeneration } from "./imageGeneration/providers/ideogram.ts";
 import { handleHaiperImageGeneration } from "./imageGeneration/providers/haiper.ts";
 import { handleLeonardoImageGeneration } from "./imageGeneration/providers/leonardo.ts";
@@ -375,6 +376,17 @@ export async function handleImageGeneration({
     });
   }
 
+  if (providerConfig.format === "google-imagen") {
+    return handleGoogleImagenGeneration({
+      model,
+      provider,
+      providerConfig,
+      body,
+      credentials,
+      log,
+    });
+  }
+
   if (providerConfig.format === "hyperbolic") {
     return handleHyperbolicImageGeneration({
       model,
diff --git a/open-sse/handlers/imageGeneration/providers/googleImagen.ts b/open-sse/handlers/imageGeneration/providers/googleImagen.ts
new file mode 100644
index 00000000000..b4b2ea67d86
--- /dev/null
+++ b/open-sse/handlers/imageGeneration/providers/googleImagen.ts
@@ -0,0 +1,147 @@
+// Google AI Studio (Gemini API) Imagen image generation.
+//
+// Unlike the antigravity "gemini-image" format (which wraps generateContent in a
+// Cloud Code envelope), the Imagen family on generativelanguage.googleapis.com uses
+// the dedicated ":predict" endpoint with an instances/parameters body and returns
+// base64 image bytes under `predictions[].bytesBase64Encoded`.
+//
+// Docs: https://ai.google.dev/gemini-api/docs/imagen  (Imagen requires a billing-
+// enabled Google project; free-tier keys get 403 / quota 0.)
+
+import { saveCallLog } from "@/lib/usageDb";
+import { mapImageSize } from "../../../translator/image/sizeMapper.ts";
+import { sanitizeErrorMessage } from "../../../utils/error.ts";
+
+// Only the Imagen family routes through :predict. Other gemini image models
+// (gemini-*-flash-image / nano-banana) use generateContent and belong on the chat
+// route, so they must not be dispatched here.
+export function isImagenModel(model) {
+  return /^imagen-/i.test(String(model || ""));
+}
+
+/**
+ * Build the Imagen :predict request body from an OpenAI-style image request.
+ * Pure — no I/O — so it can be unit-tested without live credentials.
+ */
+export function buildImagenPredictBody(body) {
+  const prompt = typeof body?.prompt === "string" ? body.prompt : String(body?.prompt ?? "");
+  const n = Number(body?.n);
+  const sampleCount = Number.isFinite(n) && n > 0 ? Math.min(Math.floor(n), 4) : 1;
+  return {
+    instances: [{ prompt }],
+    parameters: {
+      sampleCount,
+      aspectRatio: mapImageSize(body?.aspect_ratio || body?.size),
+    },
+  };
+}
+
+/**
+ * Normalize an Imagen :predict response into the OpenAI image-generation shape
+ * ({ created, data: [{ b64_json, revised_prompt }] }). Pure — unit-testable.
+ */
+export function parseImagenPredictResponse(data, prompt) {
+  const predictions = Array.isArray(data?.predictions) ? data.predictions : [];
+  const images = [];
+  for (const p of predictions) {
+    const b64 = p?.bytesBase64Encoded ?? p?.b64_json ?? p?.image ?? null;
+    if (typeof b64 === "string" && b64.length > 0) {
+      images.push({ b64_json: b64, revised_prompt: prompt });
+    }
+  }
+  return { created: Math.floor(Date.now() / 1000), data: images };
+}
+
+export async function handleGoogleImagenGeneration({
+  model,
+  provider,
+  providerConfig,
+  body,
+  credentials,
+  log,
+}) {
+  const startTime = Date.now();
+  const token = credentials?.apiKey || credentials?.accessToken || "";
+  const prompt = typeof body.prompt === "string" ? body.prompt : String(body.prompt ?? "");
+
+  if (!isImagenModel(model)) {
+    return {
+      success: false,
+      status: 400,
+      error: `Model ${model} is not an Imagen model. Gemini flash-image models route through /v1/chat/completions, not /v1/images/generations.`,
+    };
+  }
+
+  const upstreamBody = buildImagenPredictBody(body);
+  // baseUrl is https://generativelanguage.googleapis.com/v1beta/models
+  const url = `${providerConfig.baseUrl.replace(/\/$/, "")}/${model}:predict`;
+
+  if (log) {
+    log.info(
+      "IMAGE",
+      `${provider}/${model} (google-imagen) | prompt: "${prompt.slice(0, 60)}..." | aspectRatio: ${upstreamBody.parameters.aspectRatio}`
+    );
+  }
+
+  try {
+    const response = await fetch(url, {
+      method: "POST",
+      headers: {
+        "Content-Type": "application/json",
+        // Key travels in the header, never the URL, so it stays out of logs.
+        "x-goog-api-key": token,
+      },
+      body: JSON.stringify(upstreamBody),
+    });
+
+    if (!response.ok) {
+      const errorText = await response.text();
+      const safeError = sanitizeErrorMessage(errorText);
+      if (log) log.error("IMAGE", `${provider} error ${response.status}: ${safeError.slice(0, 200)}`);
+
+      saveCallLog({
+        method: "POST",
+        path: "/v1/images/generations",
+        status: response.status,
+        model: `${provider}/${model}`,
+        provider,
+        duration: Date.now() - startTime,
+        error: safeError.slice(0, 500),
+      }).catch(() => {});
+
+      return { success: false, status: response.status, error: safeError };
+    }
+
+    const data = await response.json();
+    const normalized = parseImagenPredictResponse(data, prompt);
+
+    saveCallLog({
+      method: "POST",
+      path: "/v1/images/generations",
+      status: 200,
+      model: `${provider}/${model}`,
+      provider,
+      duration: Date.now() - startTime,
+      responseBody: { images_count: normalized.data.length },
+    }).catch(() => {});
+
+    return { success: true, data: normalized };
+  } catch (err) {
+    const errMsg = err instanceof Error ? err.message : String(err);
+    if (log) log.error("IMAGE", `${provider} fetch error: ${errMsg}`);
+    saveCallLog({
+      method: "POST",
+      path: "/v1/images/generations",
+      status: 502,
+      model: `${provider}/${model}`,
+      provider,
+      duration: Date.now() - startTime,
+      error: errMsg,
+    }).catch(() => {});
+    return {
+      success: false,
+      status: 502,
+      error: `Image provider error: ${sanitizeErrorMessage(errMsg)}`,
+    };
+  }
+}
diff --git a/open-sse/mcp-server/README.md b/open-sse/mcp-server/README.md
index df32ebc7f28..1ef5f8aa5f5 100644
--- a/open-sse/mcp-server/README.md
+++ b/open-sse/mcp-server/README.md
@@ -1,8 +1,8 @@
 # OmniRoute MCP Server
 
-> **Model Context Protocol server** that exposes OmniRoute's gateway intelligence as **37 tools** for AI agents.
+> **Model Context Protocol server** that exposes OmniRoute's gateway intelligence as **104 tools** for AI agents.
 >
-> **Source of truth for the full tool catalog and REST surface:** [`docs/frameworks/MCP-SERVER.md`](../../docs/MCP-SERVER.md). This README focuses on architecture, configuration, and integration examples; the catalog below is a summary subset.
+> **Source of truth for the full tool catalog and REST surface:** [`docs/frameworks/MCP-SERVER.md`](../../docs/frameworks/MCP-SERVER.md). This README focuses on architecture, configuration, and integration examples; the catalog below is a summary subset.
 
 The MCP Server allows any AI agent (Claude Desktop, Cursor, VS Code Copilot, custom agents) to **monitor, control, and optimize** the OmniRoute AI gateway programmatically.
 
@@ -20,7 +20,7 @@ The MCP Server allows any AI agent (Claude Desktop, Cursor, VS Code Copilot, cus
 ┌──────────────────────────────────────────────────────────────────┐
 │                      OmniRoute MCP Server                        │
 │  ┌──────────────┐  ┌─────────────────┐  ┌────────────────────┐  │
-│  │ Scope        │  │  37 MCP Tools   │  │   Audit Logger     │  │
+│  │ Scope        │  │ 104 MCP Tools   │  │   Audit Logger     │  │
 │  │ Enforcement  │──│ (core + memory  │──│   (SHA-256/SQLite) │  │
 │  │              │  │  + skills + …)  │  │                    │  │
 │  └──────────────┘  └────────┬────────┘  └────────────────────┘  │
@@ -157,6 +157,16 @@ omniroute --mcp
 | 25  | `omniroute_set_compression_engine`  | `write:compression` | Set Caveman, RTK, or stacked compression mode and pipeline                   |
 | 26  | `omniroute_list_compression_combos` | `read:compression`  | List named compression combos and routing assignments                        |
 | 27  | `omniroute_compression_combo_stats` | `read:compression`  | Read analytics grouped by compression combo and engine                       |
+| 28  | `omniroute_ccr_store`               | `write:compression` | Store content in the caller-isolated in-memory CCR store                     |
+| 29  | `omniroute_ccr_retrieve`            | `read:compression`  | Retrieve full or ranged caller-owned CCR content                             |
+| 30  | `omniroute_ccr_inspect`             | `read:compression`  | Inspect CCR metadata without returning content                               |
+| 31  | `omniroute_ccr_list`                | `read:compression`  | List paginated caller-owned CCR metadata                                     |
+| 32  | `omniroute_ccr_delete`              | `write:compression` | Delete a caller-owned CCR block                                              |
+| 33  | `omniroute_ccr_stats`               | `read:compression`  | Report caller usage, bounded-store limits, and lifecycle counters            |
+
+CCR storage is bounded and in-memory only: 2 MiB per block, 16 MiB per principal, 64 MiB global,
+with a 24-hour default TTL. Full MCP retrieval is capped at 256 KiB; larger blocks use ranged or
+grep retrieval. All lifecycle operations are isolated by the authenticated caller principal.
 
 MCP listable metadata descriptions are compressed at registration/list time when description
 compression is enabled. `omniroute_compression_status` exposes those savings separately as
diff --git a/open-sse/mcp-server/__tests__/toolSearch.catalog.test.ts b/open-sse/mcp-server/__tests__/toolSearch.catalog.test.ts
index e0bfa82f345..b5a982e548a 100644
--- a/open-sse/mcp-server/__tests__/toolSearch.catalog.test.ts
+++ b/open-sse/mcp-server/__tests__/toolSearch.catalog.test.ts
@@ -18,10 +18,9 @@ describe("getAllToolDefinitions", () => {
     const names = all.map((t) => t.name);
     expect(new Set(names).size).toBe(names.length);
   });
-  it("includes compressionTools-only entries (omniroute_ccr_retrieve, not in MCP_TOOLS)", () => {
-    // Regression: compressionTools carries omniroute_ccr_retrieve, which is absent from
-    // MCP_TOOLS — if the collection is dropped from the catalog, tool_search can never
-    // surface it. Guards against the catalog omission caught in core review.
-    expect(all.find((t) => t.name === "omniroute_ccr_retrieve")).toBeTruthy();
+  it("includes every canonical CCR lifecycle tool", () => {
+    for (const name of ["store", "retrieve", "inspect", "list", "delete", "stats"]) {
+      expect(all.find((tool) => tool.name === `omniroute_ccr_${name}`)).toBeTruthy();
+    }
   });
 });
diff --git a/open-sse/mcp-server/schemas/ccrTools.ts b/open-sse/mcp-server/schemas/ccrTools.ts
new file mode 100644
index 00000000000..18630c0b3ca
--- /dev/null
+++ b/open-sse/mcp-server/schemas/ccrTools.ts
@@ -0,0 +1,202 @@
+import { z } from "zod";
+
+import type { McpToolDefinition } from "./toolDefinition.ts";
+
+const ccrHash = z
+  .string()
+  .regex(/^[a-f0-9]{24}$/i)
+  .describe("24-hex content hash from a CCR marker or ccr:// URI");
+
+export const ccrEntryMetadataOutput = z.object({
+  hash: ccrHash,
+  bytes: z.number().int().nonnegative(),
+  chars: z.number().int().nonnegative(),
+  lines: z.number().int().nonnegative(),
+  contentType: z.string(),
+  source: z.enum(["compression", "mcp", "ionizer", "session-dedup"]),
+  createdAt: z.number().int().nonnegative(),
+  lastAccessedAt: z.number().int().nonnegative(),
+  expiresAt: z.number().int().nonnegative(),
+  retrievalCount: z.number().int().nonnegative(),
+});
+
+export const ccrReferenceOutput = z.object({
+  hash: ccrHash,
+  uri: z.string().startsWith("ccr://"),
+  marker: z.string(),
+});
+
+export const ccrStoreInput = z.object({
+  content: z
+    .string()
+    .min(1)
+    .refine((content) => Buffer.byteLength(content, "utf8") <= 2 * 1024 * 1024, {
+      message: "Content exceeds the 2 MiB UTF-8 CCR block limit",
+    })
+    .describe("Verbatim content to keep in the in-memory CCR store (maximum 2 MiB UTF-8)"),
+  contentType: z.string().trim().min(1).max(128).optional(),
+  ttlSeconds: z
+    .number()
+    .int()
+    .min(60)
+    .max(7 * 24 * 60 * 60)
+    .optional(),
+});
+
+export const ccrStoreOutput = z.union([
+  z.object({
+    stored: z.literal(true),
+    reference: ccrReferenceOutput,
+    metadata: ccrEntryMetadataOutput,
+  }),
+  z.object({
+    stored: z.literal(false),
+    reason: z.enum(["block_too_large", "principal_budget_exceeded", "global_budget_exceeded"]),
+  }),
+]);
+
+export const ccrStoreTool: McpToolDefinition<typeof ccrStoreInput, typeof ccrStoreOutput> = {
+  name: "omniroute_ccr_store",
+  description:
+    "Store verbatim content in the caller-isolated in-memory CCR store and return a ccr:// reference plus the compatible CCR marker. Entries expire automatically and are not persisted across restarts.",
+  inputSchema: ccrStoreInput,
+  outputSchema: ccrStoreOutput,
+  scopes: ["write:compression"],
+  auditLevel: "basic",
+  phase: 2,
+  sourceEndpoints: [],
+};
+
+export const ccrRetrieveInput = z.object({
+  hash: ccrHash,
+  mode: z.enum(["full", "head", "tail", "lines", "grep", "stats"]).optional(),
+  n: z.number().int().positive().max(10_000).optional(),
+  start: z.number().int().positive().optional(),
+  end: z.number().int().positive().optional(),
+  pattern: z.string().max(512).optional(),
+  unique: z.boolean().optional(),
+});
+
+export const ccrRetrieveOutput = z.union([
+  z.object({
+    found: z.literal(false),
+    error: z.string(),
+  }),
+  z.object({
+    found: z.literal(true),
+    metadata: ccrEntryMetadataOutput,
+    content: z.string().optional(),
+    tooLargeForFull: z.boolean().optional(),
+    suggestedModes: z.array(z.enum(["head", "tail", "lines", "grep", "stats"])).optional(),
+    error: z.string().optional(),
+  }),
+]);
+
+export const ccrRetrieveTool: McpToolDefinition<typeof ccrRetrieveInput, typeof ccrRetrieveOutput> =
+  {
+    name: "omniroute_ccr_retrieve",
+    description:
+      "Retrieve caller-owned CCR content by hash. Full MCP responses are capped at 256 KiB; use head, tail, lines, grep, or stats for larger blocks.",
+    inputSchema: ccrRetrieveInput,
+    outputSchema: ccrRetrieveOutput,
+    scopes: ["read:compression"],
+    auditLevel: "basic",
+    phase: 2,
+    sourceEndpoints: ["/api/compression/retrieve"],
+  };
+
+export const ccrInspectInput = z.object({ hash: ccrHash });
+export const ccrInspectOutput = z.union([
+  z.object({ found: z.literal(false) }),
+  z.object({
+    found: z.literal(true),
+    reference: ccrReferenceOutput,
+    metadata: ccrEntryMetadataOutput,
+  }),
+]);
+export const ccrInspectTool: McpToolDefinition<typeof ccrInspectInput, typeof ccrInspectOutput> = {
+  name: "omniroute_ccr_inspect",
+  description: "Inspect metadata for a caller-owned CCR block without returning its content.",
+  inputSchema: ccrInspectInput,
+  outputSchema: ccrInspectOutput,
+  scopes: ["read:compression"],
+  auditLevel: "basic",
+  phase: 2,
+  sourceEndpoints: [],
+};
+
+export const ccrListInput = z.object({
+  offset: z.number().int().nonnegative().optional(),
+  limit: z.number().int().min(1).max(100).optional(),
+});
+export const ccrListOutput = z.object({
+  entries: z.array(z.object({ reference: ccrReferenceOutput, metadata: ccrEntryMetadataOutput })),
+  total: z.number().int().nonnegative(),
+  offset: z.number().int().nonnegative(),
+  limit: z.number().int().positive(),
+  hasMore: z.boolean(),
+});
+export const ccrListTool: McpToolDefinition<typeof ccrListInput, typeof ccrListOutput> = {
+  name: "omniroute_ccr_list",
+  description: "List paginated metadata for CCR blocks owned by the current caller.",
+  inputSchema: ccrListInput,
+  outputSchema: ccrListOutput,
+  scopes: ["read:compression"],
+  auditLevel: "basic",
+  phase: 2,
+  sourceEndpoints: [],
+};
+
+export const ccrDeleteInput = z.object({ hash: ccrHash });
+export const ccrDeleteOutput = z.object({ deleted: z.boolean() });
+export const ccrDeleteTool: McpToolDefinition<typeof ccrDeleteInput, typeof ccrDeleteOutput> = {
+  name: "omniroute_ccr_delete",
+  description: "Delete a caller-owned block from the in-memory CCR store.",
+  inputSchema: ccrDeleteInput,
+  outputSchema: ccrDeleteOutput,
+  scopes: ["write:compression"],
+  auditLevel: "basic",
+  phase: 2,
+  sourceEndpoints: [],
+};
+
+export const ccrStatsInput = z.object({});
+export const ccrStatsOutput = z.object({
+  storage: z.literal("memory"),
+  entries: z.number().int().nonnegative(),
+  bytes: z.number().int().nonnegative(),
+  limits: z.object({
+    maxEntries: z.number().int().positive(),
+    maxBlockBytes: z.number().int().positive(),
+    maxPrincipalBytes: z.number().int().positive(),
+    maxGlobalBytes: z.number().int().positive(),
+    defaultTtlSeconds: z.number().int().positive(),
+    maxTtlSeconds: z.number().int().positive(),
+    maxMcpFullBytes: z.number().int().positive(),
+  }),
+  lifecycle: z.object({
+    expiredEvictions: z.number().int().nonnegative(),
+    capacityEvictions: z.number().int().nonnegative(),
+    rejectedStores: z.number().int().nonnegative(),
+  }),
+});
+export const ccrStatsTool: McpToolDefinition<typeof ccrStatsInput, typeof ccrStatsOutput> = {
+  name: "omniroute_ccr_stats",
+  description:
+    "Return caller-scoped CCR entry and byte usage, lifecycle counters, and in-memory store limits.",
+  inputSchema: ccrStatsInput,
+  outputSchema: ccrStatsOutput,
+  scopes: ["read:compression"],
+  auditLevel: "basic",
+  phase: 2,
+  sourceEndpoints: [],
+};
+
+export const CCR_MCP_TOOLS = [
+  ccrStoreTool,
+  ccrRetrieveTool,
+  ccrInspectTool,
+  ccrListTool,
+  ccrDeleteTool,
+  ccrStatsTool,
+] as const;
diff --git a/open-sse/mcp-server/schemas/index.ts b/open-sse/mcp-server/schemas/index.ts
index e233dd03baf..fe9df69ff2d 100644
--- a/open-sse/mcp-server/schemas/index.ts
+++ b/open-sse/mcp-server/schemas/index.ts
@@ -69,6 +69,26 @@ export {
   cacheFlushInput,
   cacheFlushOutput,
   cacheFlushTool,
+  ccrEntryMetadataOutput,
+  ccrReferenceOutput,
+  ccrStoreInput,
+  ccrStoreOutput,
+  ccrStoreTool,
+  ccrRetrieveInput,
+  ccrRetrieveOutput,
+  ccrRetrieveTool,
+  ccrInspectInput,
+  ccrInspectOutput,
+  ccrInspectTool,
+  ccrListInput,
+  ccrListOutput,
+  ccrListTool,
+  ccrDeleteInput,
+  ccrDeleteOutput,
+  ccrDeleteTool,
+  ccrStatsInput,
+  ccrStatsOutput,
+  ccrStatsTool,
 } from "./tools.ts";
 
 // A2A schemas
diff --git a/open-sse/mcp-server/schemas/tools.ts b/open-sse/mcp-server/schemas/tools.ts
index c6fc9908225..bf77b115779 100644
--- a/open-sse/mcp-server/schemas/tools.ts
+++ b/open-sse/mcp-server/schemas/tools.ts
@@ -12,6 +12,7 @@
 import { z } from "zod";
 import { toolSearchTool } from "./toolSearch.ts";
 import { pickFastestModelTool } from "./pickFastestModel.ts";
+import { CCR_MCP_TOOLS } from "./ccrTools.ts";
 import {
   AUTO_ROUTING_STRATEGY_VALUES,
   ROUTING_STRATEGY_VALUES,
@@ -24,6 +25,7 @@ import {
 export type { AuditLevel, McpToolDefinition } from "./toolDefinition.ts";
 import type { McpToolDefinition } from "./toolDefinition.ts";
 export { pickFastestModelInput, pickFastestModelOutput } from "./pickFastestModel.ts";
+export * from "./ccrTools.ts";
 
 // ============ Phase 1: Essential Tools (8) ============
 
@@ -1462,6 +1464,7 @@ export const MCP_TOOLS = [
   setCompressionEngineTool,
   listCompressionCombosTool,
   compressionComboStatsTool,
+  ...CCR_MCP_TOOLS,
   oneproxyFetchTool,
   oneproxyRotateTool,
   oneproxyStatsTool,
diff --git a/open-sse/mcp-server/server.ts b/open-sse/mcp-server/server.ts
index 4d8b7be2de9..477fb9d816e 100644
--- a/open-sse/mcp-server/server.ts
+++ b/open-sse/mcp-server/server.ts
@@ -110,6 +110,7 @@ const TOTAL_MCP_TOOL_COUNT = countUniqueMcpTools({
   pluginTools,
   notionTools,
   obsidianTools,
+  compressionTools,
 });
 
 type JsonRecord = Record<string, unknown>;
diff --git a/open-sse/mcp-server/toolSearch/catalog.ts b/open-sse/mcp-server/toolSearch/catalog.ts
index bd597a4bff5..5cbd39f986b 100644
--- a/open-sse/mcp-server/toolSearch/catalog.ts
+++ b/open-sse/mcp-server/toolSearch/catalog.ts
@@ -76,9 +76,8 @@ export function getAllToolDefinitions(): ToolCatalogEntry[] {
     pluginTools,
     notionTools,
     obsidianTools,
-    // compressionTools holds omniroute_ccr_retrieve, which is NOT in MCP_TOOLS — without it
-    // a `tool_search("compression")` would miss that tool. The other 5 overlap MCP_TOOLS and
-    // are resolved by the dedup-by-name below (first wins).
+    // Keep the concrete handler collection in the catalog as a parity guard. Canonical CCR
+    // definitions now live in MCP_TOOLS too; deduplication below keeps each name visible once.
     compressionTools,
   ];
 
diff --git a/open-sse/mcp-server/tools/compressionTools.ts b/open-sse/mcp-server/tools/compressionTools.ts
index 4d1df0f4598..d9e39ffd722 100644
--- a/open-sse/mcp-server/tools/compressionTools.ts
+++ b/open-sse/mcp-server/tools/compressionTools.ts
@@ -4,6 +4,7 @@
  * Tools:
  *   1. omniroute_compression_status   — Get compression config, analytics, and cache stats
  *   2. omniroute_compression_configure — Update compression settings
+ *   3. CCR lifecycle tools             — Store, retrieve, inspect, list, delete, and stats
  */
 
 import { logToolCall } from "../audit.ts";
@@ -241,8 +242,23 @@ import {
   setCompressionEngineInput,
   listCompressionCombosInput,
   compressionComboStatsInput,
+  ccrStoreInput,
+  ccrRetrieveInput,
+  ccrInspectInput,
+  ccrListInput,
+  ccrDeleteInput,
+  ccrStatsInput,
 } from "../schemas/tools.ts";
-import { handleCcrRetrieve } from "../../services/compression/engines/ccr/index.ts";
+import {
+  MAX_CCR_MCP_FULL_BYTES,
+  buildCcrReference,
+  deleteCcrBlock,
+  getCcrStoreStats,
+  handleCcrRetrieve,
+  inspectCcrBlock,
+  listCcrBlocks,
+  tryStoreBlock,
+} from "../../services/compression/engines/ccr/index.ts";
 import {
   listRtkCommandSamples,
   discoverRepeatedNoise,
@@ -252,22 +268,161 @@ import {
 import { resolveCallerScopeContext } from "../scopeEnforcement.ts";
 import { resolveMcpCallerApiKeyId } from "../mcpCallerIdentity.ts";
 
-const ccrRetrieveInput = z.object({
-  hash: z
-    .string()
-    .min(6)
-    .max(64)
-    .describe("24-hex content hash from a [CCR retrieve hash=<hash>] marker"),
-  mode: z
-    .enum(["full", "head", "tail", "lines", "grep", "stats"])
-    .optional()
-    .describe("Retrieval mode: full (default) | head | tail | lines | grep | stats"),
-  n: z.number().int().positive().max(10000).optional().describe("head/tail: number of lines"),
-  start: z.number().int().positive().optional().describe("lines: 1-indexed inclusive start"),
-  end: z.number().int().positive().optional().describe("lines: 1-indexed inclusive end"),
-  pattern: z.string().max(512).optional().describe("grep: regex (validated safe; ReDoS-rejected)"),
-  unique: z.boolean().optional().describe("grep: dedupe matching lines"),
-});
+async function resolveCcrPrincipal(
+  extra: McpToolExtraLike | undefined,
+  scopes: readonly string[]
+): Promise<string | undefined> {
+  const apiKeyPrincipal = await resolveMcpCallerApiKeyId();
+  if (apiKeyPrincipal) return apiKeyPrincipal;
+  const { callerId } = resolveCallerScopeContext(extra, scopes);
+  return callerId === "anonymous" ? undefined : callerId;
+}
+
+export function buildCcrStoreAuditInput(args: z.infer<typeof ccrStoreInput>) {
+  return {
+    bytes: Buffer.byteLength(args.content, "utf8"),
+    contentType: args.contentType,
+    ttlSeconds: args.ttlSeconds,
+  };
+}
+
+export async function handleCcrStoreTool(
+  args: z.infer<typeof ccrStoreInput>,
+  extra?: McpToolExtraLike
+) {
+  const start = Date.now();
+  const principal = await resolveCcrPrincipal(extra, ["write:compression"]);
+  const result = tryStoreBlock(args.content, principal, {
+    contentType: args.contentType,
+    source: "mcp",
+    ttlSeconds: args.ttlSeconds,
+  });
+  const auditInput = buildCcrStoreAuditInput(args);
+  if (!result.stored) {
+    const output = { stored: false as const, reason: result.reason };
+    await logToolCall(
+      "omniroute_ccr_store",
+      auditInput,
+      output,
+      Date.now() - start,
+      false,
+      result.reason
+    );
+    return output;
+  }
+  const output = {
+    stored: true as const,
+    reference: buildCcrReference(result.hash, result.metadata.chars),
+    metadata: result.metadata,
+  };
+  await logToolCall("omniroute_ccr_store", auditInput, output, Date.now() - start, true);
+  return output;
+}
+
+export async function handleCcrRetrieveTool(
+  args: z.infer<typeof ccrRetrieveInput>,
+  extra?: McpToolExtraLike
+) {
+  const start = Date.now();
+  const principal = await resolveCcrPrincipal(extra, ["read:compression"]);
+  const metadata = inspectCcrBlock(args.hash, principal);
+  if (!metadata) {
+    const output = { found: false as const, error: "CCR block not found or expired" };
+    await logToolCall(
+      "omniroute_ccr_retrieve",
+      args,
+      output,
+      Date.now() - start,
+      false,
+      "NOT_FOUND"
+    );
+    return output;
+  }
+  if ((!args.mode || args.mode === "full") && metadata.bytes > MAX_CCR_MCP_FULL_BYTES) {
+    const output = {
+      found: true as const,
+      tooLargeForFull: true as const,
+      metadata,
+      suggestedModes: ["head", "tail", "lines", "grep", "stats"] as const,
+    };
+    await logToolCall("omniroute_ccr_retrieve", args, output, Date.now() - start, true);
+    return output;
+  }
+  const queried = handleCcrRetrieve(args, principal);
+  const refreshedMetadata = inspectCcrBlock(args.hash, principal) ?? metadata;
+  const output =
+    "content" in queried
+      ? { found: true as const, metadata: refreshedMetadata, content: queried.content }
+      : { found: true as const, metadata: refreshedMetadata, error: queried.error };
+  await logToolCall(
+    "omniroute_ccr_retrieve",
+    args,
+    {
+      ...output,
+      ...(typeof output.content === "string"
+        ? { content: `[${Buffer.byteLength(output.content, "utf8")} bytes]` }
+        : {}),
+    },
+    Date.now() - start,
+    !("error" in output),
+    "error" in output ? "INVALID_QUERY" : undefined
+  );
+  return output;
+}
+
+export async function handleCcrInspectTool(
+  args: z.infer<typeof ccrInspectInput>,
+  extra?: McpToolExtraLike
+) {
+  const start = Date.now();
+  const principal = await resolveCcrPrincipal(extra, ["read:compression"]);
+  const metadata = inspectCcrBlock(args.hash, principal);
+  const output = metadata
+    ? { found: true as const, reference: buildCcrReference(args.hash, metadata.chars), metadata }
+    : { found: false as const };
+  await logToolCall("omniroute_ccr_inspect", args, output, Date.now() - start, Boolean(metadata));
+  return output;
+}
+
+export async function handleCcrListTool(
+  args: z.infer<typeof ccrListInput>,
+  extra?: McpToolExtraLike
+) {
+  const start = Date.now();
+  const principal = await resolveCcrPrincipal(extra, ["read:compression"]);
+  const result = listCcrBlocks(principal, args);
+  const output = {
+    ...result,
+    entries: result.entries.map((metadata) => ({
+      reference: buildCcrReference(metadata.hash, metadata.chars),
+      metadata,
+    })),
+  };
+  await logToolCall("omniroute_ccr_list", args, output, Date.now() - start, true);
+  return output;
+}
+
+export async function handleCcrDeleteTool(
+  args: z.infer<typeof ccrDeleteInput>,
+  extra?: McpToolExtraLike
+) {
+  const start = Date.now();
+  const principal = await resolveCcrPrincipal(extra, ["write:compression"]);
+  const output = { deleted: deleteCcrBlock(args.hash, principal) };
+  await logToolCall("omniroute_ccr_delete", args, output, Date.now() - start, true);
+  return output;
+}
+
+export async function handleCcrStatsTool(
+  args: z.infer<typeof ccrStatsInput>,
+  extra?: McpToolExtraLike
+) {
+  const start = Date.now();
+  const principal = await resolveCcrPrincipal(extra, ["read:compression"]);
+  const output = getCcrStoreStats(principal);
+  await logToolCall("omniroute_ccr_stats", args, output, Date.now() - start, true);
+  return output;
+}
 
 export async function handleSetCompressionEngine(
   args: z.infer<typeof setCompressionEngineInput>
@@ -323,12 +478,24 @@ export async function handleCompressionComboStats(
 // T07 — RTK learn/discover exposed via MCP (read-only; suggestions only). Mines the opt-in
 // raw-output sample store, exactly like the /api/context/rtk/{discover,learn} routes.
 const rtkDiscoverInput = z.object({
-  limit: z.number().int().positive().max(2000).optional().describe("Max samples to scan (default 500)"),
+  limit: z
+    .number()
+    .int()
+    .positive()
+    .max(2000)
+    .optional()
+    .describe("Max samples to scan (default 500)"),
 });
 
 const rtkLearnInput = z.object({
   command: z.string().min(1).max(500).describe("The command to learn an RTK filter draft for"),
-  limit: z.number().int().positive().max(2000).optional().describe("Max samples to scan (default 500)"),
+  limit: z
+    .number()
+    .int()
+    .positive()
+    .max(2000)
+    .optional()
+    .describe("Max samples to scan (default 500)"),
 });
 
 function resolveSampleLimit(limit?: number): number {
@@ -401,6 +568,14 @@ export const compressionTools = {
     handler: (args: z.infer<typeof compressionComboStatsInput>) =>
       handleCompressionComboStats(args),
   },
+  omniroute_ccr_store: {
+    name: "omniroute_ccr_store",
+    description:
+      "Store verbatim content in the caller-isolated in-memory CCR store and return a ccr:// reference plus the compatible CCR marker. Entries expire automatically and are not persisted across restarts.",
+    scopes: ["write:compression"],
+    inputSchema: ccrStoreInput,
+    handler: handleCcrStoreTool,
+  },
   omniroute_ccr_retrieve: {
     name: "omniroute_ccr_retrieve",
     description:
@@ -411,22 +586,36 @@ export const compressionTools = {
       "Scope: read:compression. Always available (sticky-on).",
     scopes: ["read:compression"],
     inputSchema: ccrRetrieveInput,
-    handler: async (args: z.infer<typeof ccrRetrieveInput>, extra?: McpToolExtraLike) => {
-      // Retrieve must use the SAME principal the CCR store used at compression time:
-      // `String(apiKeyInfo.id)` (chatCore → getApiKeyMetadata(rawKey)). On MCP HTTP
-      // transports the raw key lives in httpAuthContext (not in extra.authInfo, since
-      // OmniRoute auth is API-key not OAuth-clientId) — resolve it to the same key id
-      // so the block is found. Without this the caller resolved to "anonymous" and the
-      // store-key never matched (#5649). Cross-tenant IDOR stays closed: a different
-      // key → different id → miss; no key → undefined → anonymous bucket only.
-      const apiKeyPrincipal = await resolveMcpCallerApiKeyId();
-      if (apiKeyPrincipal) {
-        return handleCcrRetrieve(args, apiKeyPrincipal);
-      }
-      // Fallback (unchanged): OAuth clientId / session scope context, then anonymous.
-      const { callerId } = resolveCallerScopeContext(extra, ["read:compression"]);
-      return handleCcrRetrieve(args, callerId === "anonymous" ? undefined : callerId);
-    },
+    handler: handleCcrRetrieveTool,
+  },
+  omniroute_ccr_inspect: {
+    name: "omniroute_ccr_inspect",
+    description: "Inspect metadata for a caller-owned CCR block without returning its content.",
+    scopes: ["read:compression"],
+    inputSchema: ccrInspectInput,
+    handler: handleCcrInspectTool,
+  },
+  omniroute_ccr_list: {
+    name: "omniroute_ccr_list",
+    description: "List paginated metadata for CCR blocks owned by the current caller.",
+    scopes: ["read:compression"],
+    inputSchema: ccrListInput,
+    handler: handleCcrListTool,
+  },
+  omniroute_ccr_delete: {
+    name: "omniroute_ccr_delete",
+    description: "Delete a caller-owned block from the in-memory CCR store.",
+    scopes: ["write:compression"],
+    inputSchema: ccrDeleteInput,
+    handler: handleCcrDeleteTool,
+  },
+  omniroute_ccr_stats: {
+    name: "omniroute_ccr_stats",
+    description:
+      "Return caller-scoped CCR entry and byte usage, lifecycle counters, and in-memory store limits.",
+    scopes: ["read:compression"],
+    inputSchema: ccrStatsInput,
+    handler: handleCcrStatsTool,
   },
   omniroute_rtk_discover: {
     name: "omniroute_rtk_discover",
diff --git a/open-sse/package.json b/open-sse/package.json
index c53d7d28fdd..6f38dfd26e4 100644
--- a/open-sse/package.json
+++ b/open-sse/package.json
@@ -12,6 +12,7 @@
   },
   "dependencies": {
     "@toon-format/toon": "^2.3.0",
-    "safe-regex": "^2.1.1"
+    "safe-regex": "^2.1.1",
+    "smol-toml": "1.6.1"
   }
 }
diff --git a/open-sse/services/accountFallback.ts b/open-sse/services/accountFallback.ts
index 10347ec0f3f..bd1dbabe661 100644
--- a/open-sse/services/accountFallback.ts
+++ b/open-sse/services/accountFallback.ts
@@ -44,6 +44,8 @@ import {
   buildSessionQuotaFallback,
 } from "./quotaTextCooldowns.ts";
 import { parseDayGranularityResetMs, shouldPreserveQuotaSignals } from "./quotaResetParsing.ts";
+import { evictLockoutOverflow } from "./accountFallback/lockoutEviction.ts";
+export { MODEL_LOCKOUT_EVICTION_CAP } from "./accountFallback/lockoutEviction.ts";
 
 export type ProviderProfile = {
   baseCooldownMs: number;
@@ -69,7 +71,7 @@ export type ProviderProfile = {
 };
 type JsonRecord = Record<string, unknown>;
 type RateLimitReasonValue = (typeof RateLimitReason)[keyof typeof RateLimitReason];
-type ModelLockoutEntry = {
+export type ModelLockoutEntry = {
   reason: string;
   until: number;
   lockedAt: number;
@@ -77,7 +79,7 @@ type ModelLockoutEntry = {
   lastFailureAt: number;
   resetAfterMs: number;
 };
-type ModelFailureState = {
+export type ModelFailureState = {
   failureCount: number;
   lastFailureAt: number;
   resetAfterMs: number;
@@ -467,6 +469,7 @@ function ensureCleanupTimer() {
       const now = Date.now();
       for (const key of modelLockouts.keys()) cleanupModelLockKey(key, now);
       for (const key of modelFailureState.keys()) cleanupModelLockKey(key, now);
+      evictModelLockoutOverflow();
     }, 15_000);
     if (typeof _cleanupTimer === "object" && "unref" in _cleanupTimer) {
       (_cleanupTimer as { unref?: () => void }).unref?.(); // Don't prevent process exit (Node.js only)
@@ -476,6 +479,14 @@ function ensureCleanupTimer() {
   }
 }
 
+/** @internal exported for testing only (both accessors below). */
+export function evictModelLockoutOverflow(): void {
+  evictLockoutOverflow(modelLockouts, modelFailureState);
+}
+export function getModelLockoutSize(): number {
+  return modelLockouts.size;
+}
+
 /**
  * Lock a specific model on a specific account
  * @param {string} provider
diff --git a/open-sse/services/accountFallback/lockoutEviction.ts b/open-sse/services/accountFallback/lockoutEviction.ts
new file mode 100644
index 00000000000..83199830a2b
--- /dev/null
+++ b/open-sse/services/accountFallback/lockoutEviction.ts
@@ -0,0 +1,53 @@
+/**
+ * accountFallback/lockoutEviction.ts — model-lockout map eviction (cap enforcement).
+ *
+ * Extracted from services/accountFallback.ts (file-size gate, #6923): the bounded-growth
+ * eviction sweep for the in-memory `modelLockouts` / `modelFailureState` maps. Pure w.r.t.
+ * module state — operates only on the maps passed in by the caller — so it is independently
+ * testable and reusable outside accountFallback.ts, which wraps evictLockoutOverflow() with
+ * its own private map instances and re-exports MODEL_LOCKOUT_EVICTION_CAP.
+ */
+
+import type { ModelLockoutEntry, ModelFailureState } from "../accountFallback.ts";
+
+// Cap prevents unbounded growth under sustained load. Entries beyond this limit
+// are evicted (oldest first, in insertion order) during the periodic cleanup.
+export const MODEL_LOCKOUT_EVICTION_CAP = 1000;
+
+/**
+ * Evict oldest (insertion-order) entries once a map exceeds the cap — but NEVER a
+ * still-active (until > now) lockout: cleanupModelLockKey() has already run on every
+ * key this tick, so anything active left here is a real, in-progress cooldown, and
+ * dropping it would wrongly let routing resume to it. If a map is still over cap
+ * purely from active entries, the cap is a soft bound in that rare case rather than
+ * a correctness trade-off.
+ */
+export function evictLockoutOverflow(
+  modelLockouts: Map<string, ModelLockoutEntry>,
+  modelFailureState: Map<string, ModelFailureState>,
+  cap: number = MODEL_LOCKOUT_EVICTION_CAP
+): void {
+  if (modelLockouts.size > cap) {
+    const overflow = modelLockouts.size - cap;
+    const now = Date.now();
+    // Only expired entries are eviction candidates (oldest-first, up to the
+    // overflow count) — active ones never appear in this list at all.
+    const evictableKeys = [...modelLockouts.entries()]
+      .filter(([, entry]) => entry.until <= now)
+      .slice(0, overflow)
+      .map(([key]) => key);
+    for (const key of evictableKeys) {
+      modelLockouts.delete(key);
+      modelFailureState.delete(key);
+    }
+  }
+  if (modelFailureState.size > cap) {
+    const overflow = modelFailureState.size - cap;
+    let evicted = 0;
+    for (const key of modelFailureState.keys()) {
+      if (evicted >= overflow) break;
+      if (!modelLockouts.has(key)) modelFailureState.delete(key);
+      evicted++;
+    }
+  }
+}
diff --git a/open-sse/services/claudeCodeCompatible.ts b/open-sse/services/claudeCodeCompatible.ts
index 01b021579fc..a33e4cd09ed 100644
--- a/open-sse/services/claudeCodeCompatible.ts
+++ b/open-sse/services/claudeCodeCompatible.ts
@@ -56,6 +56,7 @@ const CLAUDE_CODE_COMPATIBLE_DEFAULT_SYSTEM_BLOCKS = [
 const CONTEXT_1M_SUPPORTED_MODELS = [
   "claude-fable-5",
   "claude-sonnet-5",
+  "claude-sonnet-4-6",
   "claude-opus-4-8",
   "claude-opus-4-7",
   "claude-opus-4-6",
diff --git a/open-sse/services/combo.ts b/open-sse/services/combo.ts
index 3eb11d3183f..375e47fb9b3 100644
--- a/open-sse/services/combo.ts
+++ b/open-sse/services/combo.ts
@@ -67,6 +67,7 @@ import { estimateTokens } from "./contextManager.ts";
 import { getSessionConnection } from "./sessionManager.ts";
 import {
   applySessionStickiness,
+  normalizeStickinessMessages,
   recordStickyBinding,
   clearStickyBinding,
   peekStickyConnectionId,
@@ -137,6 +138,8 @@ import {
   clampComboDepth,
   shouldSkipForPredictedTtft,
   shouldRecordProviderBreakerFailure,
+  isRequestScopedUpstreamFailure,
+  shouldSkipConnDisable,
   resolveDelayMs,
   comboModelNotFoundResponse,
   isStreamReadinessFailureErrorBody,
@@ -162,6 +165,7 @@ import {
   resolveWeightedTargets,
   resolveWeightedStepGroups,
 } from "./combo/comboStructure.ts";
+import { getKnownContextOverflow } from "./combo/knownContextOverflow.ts";
 import {
   QUOTA_SOFT_DEPRIORITIZE_FACTOR,
   setCandidateQuotaSoftPenalty,
@@ -199,10 +203,21 @@ export { QUOTA_SOFT_DEPRIORITIZE_FACTOR, setCandidateQuotaSoftPenalty };
 export { scoreAutoTargets, expandAutoComboCandidatePool };
 export type { SingleModelTarget, ResolvedComboTarget };
 export { validateResponseQuality };
-export { clampComboDepth, shouldSkipForPredictedTtft, shouldRecordProviderBreakerFailure };
+export {
+  clampComboDepth,
+  shouldSkipForPredictedTtft,
+  shouldRecordProviderBreakerFailure,
+  isRequestScopedUpstreamFailure,
+  shouldSkipConnDisable,
+};
 export { resolveShadowTargets, scheduleShadowRouting };
 export { preScreenTargets };
-export { resolveComboRuntimeUnits, resolveComboTargets, filterTargetsByRequestCompatibility };
+export {
+  resolveComboRuntimeUnits,
+  resolveComboTargets,
+  filterTargetsByRequestCompatibility,
+  getKnownContextOverflow,
+};
 export {
   getComboFromData,
   getComboModelsFromData,
@@ -1151,6 +1166,31 @@ export async function handleComboChat({
 
   orderedTargets = await applyRequestTagRouting(orderedTargets, body, log);
 
+  const knownContextOverflow = getKnownContextOverflow(orderedTargets, body);
+  if (knownContextOverflow) {
+    const { requiredContextTokens, maxKnownContextTokens } = knownContextOverflow;
+    log.warn(
+      "COMBO",
+      `Request context exceeds every known target limit (${requiredContextTokens} > ${maxKnownContextTokens} tokens)`
+    );
+    return errorResponseWithComboDiagnostics(
+      400,
+      `Request requires approximately ${requiredContextTokens} tokens, but the largest known context limit in this combo is ${maxKnownContextTokens} tokens. Reduce or compact the request context.`,
+      {
+        poolSize: orderedTargets.length,
+        attempted: 0,
+        excluded: orderedTargets.map((target) => ({
+          provider: target.provider,
+          model: target.modelStr,
+          reason: "context_window",
+        })),
+        attemptOrder: [],
+        terminalReason: "context_length_exceeded",
+      },
+      { code: "context_length_exceeded", type: "invalid_request_error" }
+    );
+  }
+
   if (strategy === "weighted") {
     log.info(
       "COMBO",
@@ -1244,7 +1284,9 @@ export async function handleComboChat({
     ? ({ targets: orderedTargets, messageHash: null, stuck: false } as const)
     : await applySessionStickiness(
         orderedTargets,
-        body.messages as Array<{ role?: string; content?: unknown }>
+        // #7270: normalize both wire shapes (.messages / Responses-API .input) so the
+        // stickiness key is derivable on the /v1/responses surface, not just Chat Completions.
+        normalizeStickinessMessages(body as { messages?: unknown; input?: unknown })
       );
   orderedTargets = _sticky.targets;
   orderedTargets = orderTargetsByEvalScores(orderedTargets, config.evalRouting, log);
@@ -2064,6 +2106,7 @@ export async function handleComboChat({
                       : undefined,
                 }
               : undefined;
+          const requestScopedFailure = isRequestScopedUpstreamFailure(structuredError);
           const fallbackResult = checkFallbackError(
             result.status,
             errorText,
@@ -2176,6 +2219,7 @@ export async function handleComboChat({
               status: result.status,
               sameProviderNext,
               skipProviderBreaker: fallbackResult.skipProviderBreaker,
+              requestScopedFailure,
             })
           ) {
             recordProviderFailure(provider, log, targetWithConnection.connectionId, profile);
@@ -2201,7 +2245,7 @@ export async function handleComboChat({
             // once the model is cooling down, retrying it would waste an upstream
             // call and extend the cooldown via exponential backoff.
             let lockoutRecorded = false;
-            if (provider && rawModel && retry === 0) {
+            if (provider && rawModel && retry === 0 && !requestScopedFailure) {
               const mlSettings = resolveModelLockoutSettings(settings);
               if (mlSettings.enabled && mlSettings.errorCodes.includes(result.status)) {
                 recordModelLockoutFailure(
@@ -2244,7 +2288,7 @@ export async function handleComboChat({
           if (i > 0) fallbackCount++;
           // Wire combo failures into the resilience dashboard (model-level lockout)
           // alongside the provider-level cooldown below — they govern different scopes.
-          if (provider && rawModel) {
+          if (provider && rawModel && !requestScopedFailure) {
             const mlSettings = resolveModelLockoutSettings(settings);
             if (mlSettings.enabled && mlSettings.errorCodes.includes(result.status)) {
               recordModelLockoutFailure(
@@ -2273,6 +2317,7 @@ export async function handleComboChat({
             resilienceSettings.providerCooldown.enabled &&
             provider &&
             provider !== "unknown" &&
+            !requestScopedFailure &&
             !(result.status === 500 && hasPerModelQuota(provider, rawModel))
           ) {
             recordProviderCooldown(
@@ -2545,6 +2590,25 @@ async function handleRoundRobinCombo({
   );
   const tagFilteredTargets = await applyRequestTagRouting(orderedTargets, body, log);
   const evalRankedTargets = orderTargetsByEvalScores(tagFilteredTargets, config.evalRouting, log);
+  const knownContextOverflow = getKnownContextOverflow(evalRankedTargets, body);
+  if (knownContextOverflow) {
+    return errorResponseWithComboDiagnostics(
+      400,
+      `Request requires approximately ${knownContextOverflow.requiredContextTokens} tokens, but the largest known context limit in this combo is ${knownContextOverflow.maxKnownContextTokens} tokens. Reduce or compact the request context.`,
+      {
+        poolSize: evalRankedTargets.length,
+        attempted: 0,
+        excluded: evalRankedTargets.map((target) => ({
+          provider: target.provider,
+          model: target.modelStr,
+          reason: "context_window",
+        })),
+        attemptOrder: [],
+        terminalReason: "context_length_exceeded",
+      },
+      { code: "context_length_exceeded", type: "invalid_request_error" }
+    );
+  }
   const filteredTargets = filterTargetsByRequestCompatibility(
     evalRankedTargets,
     body,
@@ -2668,7 +2732,9 @@ async function handleRoundRobinCombo({
     ? ({ targets: filteredTargets, messageHash: null, stuck: false } as const)
     : await applySessionStickiness(
         filteredTargets,
-        body?.messages as Array<{ role?: string; content?: unknown }>
+        // #7270: normalize both wire shapes (.messages / Responses-API .input) so RR
+        // stickiness engages on the /v1/responses surface, not just Chat Completions.
+        normalizeStickinessMessages(body as { messages?: unknown; input?: unknown })
       );
   let rrStartIndex = startIndex;
   if (_rrSessionSticky.stuck) {
@@ -3039,6 +3105,7 @@ async function handleRoundRobinCombo({
                     : undefined,
               }
             : undefined;
+        const requestScopedFailure = isRequestScopedUpstreamFailure(structuredError);
         const fallbackResult = checkFallbackError(
           result.status,
           errorText,
@@ -3090,6 +3157,7 @@ async function handleRoundRobinCombo({
         if (
           !isStreamReadinessFailure &&
           !isTokenLimitBreach &&
+          !requestScopedFailure &&
           TRANSIENT_FOR_SEMAPHORE.includes(result.status) &&
           cooldownMs > 0
         ) {
@@ -3132,6 +3200,7 @@ async function handleRoundRobinCombo({
           resilienceSettings.providerCooldown.enabled &&
           provider &&
           provider !== "unknown" &&
+          !requestScopedFailure &&
           !(
             result.status === 500 &&
             hasPerModelQuota(provider, parseModel(modelStr).model || modelStr)
diff --git a/open-sse/services/combo/autoConfig.ts b/open-sse/services/combo/autoConfig.ts
index 2815444f256..ed56f11989e 100644
--- a/open-sse/services/combo/autoConfig.ts
+++ b/open-sse/services/combo/autoConfig.ts
@@ -1,4 +1,5 @@
 import { DEFAULT_WEIGHTS, type ScoringWeights } from "../autoCombo/scoring.ts";
+import { getModePack } from "../autoCombo/modePacks.ts";
 import { isRecord } from "./comboData.ts";
 import { resolveResetWindowConfig, resolveSlaRoutingPolicy } from "./quotaScoring.ts";
 import type { ComboLike, ResolvedComboTarget } from "./types.ts";
@@ -34,7 +35,7 @@ export function parseAutoConfig(combo: ComboLike, eligibleTargets: ResolvedCombo
     ? autoConfigSource.candidatePool
     : [...new Set(eligibleTargets.map((target) => target.provider))];
 
-  const weights =
+  const configuredWeights =
     autoConfigSource.weights && typeof autoConfigSource.weights === "object"
       ? (autoConfigSource.weights as ScoringWeights)
       : DEFAULT_WEIGHTS;
@@ -52,6 +53,7 @@ export function parseAutoConfig(combo: ComboLike, eligibleTargets: ResolvedCombo
       : undefined;
   const modePack =
     typeof autoConfigSource.modePack === "string" ? autoConfigSource.modePack : undefined;
+  const weights = modePack ? getModePack(modePack) || configuredWeights : configuredWeights;
   const resetWindowConfig = resolveResetWindowConfig(autoConfigSource);
   const slaPolicy = resolveSlaRoutingPolicy(autoConfigSource);
 
diff --git a/open-sse/services/combo/comboPredicates.ts b/open-sse/services/combo/comboPredicates.ts
index 887709522fa..39ba927640c 100644
--- a/open-sse/services/combo/comboPredicates.ts
+++ b/open-sse/services/combo/comboPredicates.ts
@@ -8,6 +8,7 @@
 
 import { errorResponse } from "../../utils/error.ts";
 import { parseModel } from "../model.ts";
+import { isSelfInflictedUpstreamTimeout } from "../../handlers/chatCore/cooldownClassification.ts";
 import type { ResolvedComboTarget } from "./types.ts";
 
 // Status codes that should mark round-robin target semaphores as cooling down.
@@ -150,12 +151,53 @@ export function shouldRecordProviderBreakerFailure(args: {
   status: number;
   sameProviderNext: boolean;
   skipProviderBreaker?: boolean;
+  requestScopedFailure?: boolean;
 }): boolean {
   return (
     !args.isStreamReadinessFailure &&
     PROVIDER_BREAKER_FAILURE_STATUSES.has(args.status) &&
     !args.sameProviderNext &&
-    !args.skipProviderBreaker
+    !args.skipProviderBreaker &&
+    !args.requestScopedFailure
+  );
+}
+
+const REQUEST_SCOPED_UPSTREAM_ERROR_CODES = new Set([
+  "context_length_exceeded",
+  "upstream_empty_response",
+  "upstream_response_failed",
+]);
+
+/** Request/model-specific failures must not poison provider-wide resilience state. */
+export function isRequestScopedUpstreamFailure(error?: {
+  code?: string | null;
+  type?: string | null;
+}): boolean {
+  const code = typeof error?.code === "string" ? error.code.toLowerCase() : "";
+  const type = typeof error?.type === "string" ? error.type.toLowerCase() : "";
+  return REQUEST_SCOPED_UPSTREAM_ERROR_CODES.has(code) || type === "context_length_exceeded";
+}
+
+/**
+ * #7177: whether handleSingleModelChat should skip the connection-level cooldown
+ * (markAccountUnavailable) for a failed attempt — client disconnects, a 401 when the
+ * connection has extra keys to rotate through, a known request-scoped upstream failure
+ * (e.g. context overflow — not a connection health signal), or our own self-inflicted
+ * timeout all mean the connection itself is healthy and should not be cooled down.
+ */
+export function shouldSkipConnDisable(
+  result: { status: number; errorCode?: string | null; errorType?: string | null },
+  is401: boolean,
+  hasExtraKeys: boolean,
+  provider: string
+): boolean {
+  return (
+    result.status === 499 ||
+    result.errorCode === "client_disconnected" ||
+    result.errorType === "client_disconnected" ||
+    (is401 && hasExtraKeys) ||
+    isRequestScopedUpstreamFailure({ code: result.errorCode, type: result.errorType }) ||
+    isSelfInflictedUpstreamTimeout(result.status, result.errorType, provider)
   );
 }
 
diff --git a/open-sse/services/combo/comboStructure.ts b/open-sse/services/combo/comboStructure.ts
index a98c8e820ba..0b7f8cb0fa4 100644
--- a/open-sse/services/combo/comboStructure.ts
+++ b/open-sse/services/combo/comboStructure.ts
@@ -18,6 +18,7 @@ import { getResolvedModelCapabilities } from "../modelCapabilities.ts";
 import { parseModel } from "../model.ts";
 import { dedupeTargetsByExecutionKey, isRecord } from "./comboData.ts";
 import { getTargetProvider, MAX_COMBO_DEPTH } from "./comboPredicates.ts";
+import { hasEstimableContent } from "./knownContextOverflow.ts";
 import {
   normalizeModelEntry,
   orderTargetsForWeightedFallback,
@@ -409,7 +410,7 @@ export function getModelContextLimitForModelString(modelStr: string) {
   return getModelContextLimit(provider, model);
 }
 
-type RequestCompatibilityRequirements = {
+export type RequestCompatibilityRequirements = {
   requiresTools: boolean;
   requiresVision: boolean;
   requiresStructuredOutput: boolean;
@@ -438,7 +439,7 @@ function requestRequiresStructuredOutput(body: Record<string, unknown>): boolean
 function estimateRequestInputTokens(body: Record<string, unknown>): number {
   const estimatePayload: Record<string, unknown> = {};
   for (const key of ["messages", "input", "tools", "functions", "response_format"]) {
-    if (body[key] !== undefined) estimatePayload[key] = body[key];
+    if (hasEstimableContent(body[key])) estimatePayload[key] = body[key];
   }
   return Object.keys(estimatePayload).length > 0 ? estimateTokens(estimatePayload) : 0;
 }
@@ -460,7 +461,7 @@ function valueContainsImagePart(value: unknown, depth = 0): boolean {
   return Object.values(value).some((entry) => valueContainsImagePart(entry, depth + 1));
 }
 
-function deriveRequestCompatibilityRequirements(
+export function deriveRequestCompatibilityRequirements(
   body: Record<string, unknown>
 ): RequestCompatibilityRequirements {
   const estimatedInputTokens = estimateRequestInputTokens(body);
@@ -486,21 +487,57 @@ function exceedsKnownOutputLimit(
   return maxOutputTokens < requestedOutputTokens;
 }
 
-function getKnownContextLimit(capabilities: {
-  maxInputTokens?: number | null;
-  contextWindow?: number | null;
-}): number | null {
-  return capabilities.maxInputTokens ?? capabilities.contextWindow ?? null;
+/**
+ * Decide whether a target's known context limit accommodates the request.
+ *
+ * `maxInputTokens` is an **input-only** cap — the requested output reserve is
+ * already enforced separately against `maxOutputTokens` (see
+ * `exceedsKnownOutputLimit`), so it must NOT be re-counted here. Comparing
+ * `maxInputTokens` against `estimatedInputTokens + requestedOutputTokens`
+ * double-counted the output reserve and shrank the effective input allowance
+ * (#7039).
+ *
+ * `contextWindow` is the total window, so input + output must both fit.
+ *
+ * Returns `true` when the known limit accommodates the request, `false` when
+ * it is known to be too small, and `null` when no limit metadata is known.
+ */
+function evaluateContextLimit(
+  capabilities: { maxInputTokens?: number | null; contextWindow?: number | null },
+  requirements: { estimatedInputTokens: number; requiredContextTokens: number }
+): boolean | null {
+  const hasMaxInput = capabilities.maxInputTokens != null;
+  const hasContextWindow = capabilities.contextWindow != null;
+
+  // Neither limit is known — cannot judge.
+  if (!hasMaxInput && !hasContextWindow) return null;
+
+  // The input-only cap must accommodate the estimated input.
+  const inputFits = hasMaxInput
+    ? capabilities.maxInputTokens! >= requirements.estimatedInputTokens
+    : true;
+
+  // The total window must accommodate input + requested output. The output
+  // reserve is enforced separately via `maxOutputTokens`, but when a model
+  // exposes both `maxInputTokens` and `contextWindow` the two must not be
+  // checked in isolation: a request whose input fits `maxInputTokens` but whose
+  // input + output exceeds `contextWindow` must still be rejected (#7039
+  // follow-up — shared-window models where `maxInputTokens` defaults to the
+  // total window size).
+  const totalFits = hasContextWindow
+    ? capabilities.contextWindow! >= requirements.requiredContextTokens
+    : true;
+
+  return inputFits && totalFits;
 }
 
 function hasKnownCompatibleContextLimit(
   target: ResolvedComboTarget,
-  requiredContextTokens: number
+  requirements: RequestCompatibilityRequirements
 ): boolean {
-  if (requiredContextTokens <= 0) return false;
+  if (requirements.requiredContextTokens <= 0) return false;
   const capabilities = getResolvedModelCapabilities(target.modelStr);
-  const contextLimit = getKnownContextLimit(capabilities);
-  return contextLimit !== null && contextLimit >= requiredContextTokens;
+  return evaluateContextLimit(capabilities, requirements) === true;
 }
 
 function hasOnlyContextWindowFailures(reasons: string[]): boolean {
@@ -539,12 +576,8 @@ function getTargetCompatibilityFailures(
     failures.push("output_tokens");
   }
 
-  const contextLimit = getKnownContextLimit(capabilities);
-  if (
-    requirements.requiredContextTokens > 0 &&
-    contextLimit !== null &&
-    contextLimit < requirements.requiredContextTokens
-  ) {
+  const contextVerdict = evaluateContextLimit(capabilities, requirements);
+  if (requirements.requiredContextTokens > 0 && contextVerdict === false) {
     failures.push("context_window");
   }
 
@@ -584,7 +617,7 @@ export function filterTargetsByRequestCompatibility(
   );
   if (requirements.requiredContextTokens > 0 && rejectedForContextWindow) {
     const knownContextCompatible = compatible.filter((target) =>
-      hasKnownCompatibleContextLimit(target, requirements.requiredContextTokens)
+      hasKnownCompatibleContextLimit(target, requirements)
     );
 
     if (knownContextCompatible.length > 0 && knownContextCompatible.length < compatible.length) {
diff --git a/open-sse/services/combo/knownContextOverflow.ts b/open-sse/services/combo/knownContextOverflow.ts
new file mode 100644
index 00000000000..4416d1d4516
--- /dev/null
+++ b/open-sse/services/combo/knownContextOverflow.ts
@@ -0,0 +1,97 @@
+/**
+ * Known context-overflow rejection, extracted from comboStructure.ts to keep
+ * that file under the file-size cap (#7177).
+ *
+ * Fixes: routing a request to a combo whose targets all have a KNOWN (not
+ * unknown/fail-open) context window too small for the request used to be
+ * discovered only after every target was tried and failed upstream — burning
+ * retries/cooldowns on a request that could never succeed. This lets the
+ * combo dispatcher reject it up front, before exhausting providers.
+ *
+ * getKnownContextLimit/hasEstimableContent also
+ * live here (moved from comboStructure.ts, same file-size-cap motivation):
+ * they are the "how big is a target's known context window" primitives, so
+ * they belong next to the overflow check that is their main consumer.
+ * comboStructure.ts's own compatibility filter now decides fit via its
+ * evaluateContextLimit (#7052); only hasEstimableContent is imported back.
+ */
+
+import { getResolvedModelCapabilities } from "../modelCapabilities.ts";
+import { deriveRequestCompatibilityRequirements } from "./comboStructure.ts";
+import type { ResolvedComboTarget } from "./types.ts";
+
+export type KnownContextOverflow = {
+  estimatedInputTokens: number;
+  requestedOutputTokens: number;
+  requiredContextTokens: number;
+  maxKnownContextTokens: number;
+  targetCount: number;
+};
+
+// #7177: an empty array/object (e.g. a default `messages: []` some combo entrypoints inject
+// when the caller sent none) has no real content — counting it would charge a few phantom
+// "structural" tokens (JSON.stringify braces/brackets) toward the estimate, which is enough
+// to falsely trip the exact-boundary known-context-overflow check for a request that has no
+// actual input at all.
+export function hasEstimableContent(value: unknown): boolean {
+  if (value === undefined || value === null) return false;
+  if (Array.isArray(value)) return value.length > 0;
+  if (typeof value === "object") return Object.keys(value).length > 0;
+  return true;
+}
+
+// #7177: known context limit that accounts for the request's own requested
+// output tokens — a target whose input+output would together exceed
+// maxInputTokens is exactly as incompatible as one whose contextWindow is too
+// small, so both bounds go through the same min() so far the tightest wins.
+export function getKnownContextLimit(
+  capabilities: {
+    maxInputTokens?: number | null;
+    contextWindow?: number | null;
+  },
+  requestedOutputTokens = 0
+): number | null {
+  const limits: number[] = [];
+  if (capabilities.maxInputTokens != null) {
+    limits.push(capabilities.maxInputTokens + requestedOutputTokens);
+  }
+  if (capabilities.contextWindow != null) {
+    limits.push(capabilities.contextWindow);
+  }
+  return limits.length > 0 ? Math.min(...limits) : null;
+}
+
+
+/**
+ * Return a hard context-overflow decision only when every target has a known
+ * context limit and every one of those limits is too small for the request.
+ * Unknown metadata deliberately keeps the legacy fail-open behavior.
+ */
+export function getKnownContextOverflow(
+  targets: ResolvedComboTarget[],
+  body: Record<string, unknown>
+): KnownContextOverflow | null {
+  if (targets.length === 0) return null;
+  const requirements = deriveRequestCompatibilityRequirements(body);
+  if (requirements.requiredContextTokens <= 0) return null;
+
+  const limits = targets.map((target) =>
+    getKnownContextLimit(
+      getResolvedModelCapabilities(target.modelStr),
+      requirements.requestedOutputTokens
+    )
+  );
+  if (limits.some((limit) => limit === null)) return null;
+
+  const knownLimits = limits as number[];
+  const maxKnownContextTokens = Math.max(...knownLimits);
+  if (maxKnownContextTokens >= requirements.requiredContextTokens) return null;
+
+  return {
+    estimatedInputTokens: requirements.estimatedInputTokens,
+    requestedOutputTokens: requirements.requestedOutputTokens,
+    requiredContextTokens: requirements.requiredContextTokens,
+    maxKnownContextTokens,
+    targetCount: targets.length,
+  };
+}
diff --git a/open-sse/services/combo/resolveAutoStrategy.ts b/open-sse/services/combo/resolveAutoStrategy.ts
index ef6082f9af3..9d38801b35c 100644
--- a/open-sse/services/combo/resolveAutoStrategy.ts
+++ b/open-sse/services/combo/resolveAutoStrategy.ts
@@ -10,6 +10,7 @@ import {
 } from "../autoCombo/requestControls.ts";
 import { selectWithStrategy } from "../autoCombo/routerStrategy.ts";
 import { buildComplexityRoutingHint } from "../autoCombo/complexityRouter";
+import { getModePack } from "../autoCombo/modePacks.ts";
 import { recordComboIntent } from "../comboMetrics.ts";
 import { estimateTokens } from "../contextManager.ts";
 import { classifyWithConfig } from "../intentClassifier.ts";
@@ -160,7 +161,7 @@ export async function resolveAutoStrategyOrder(
   const {
     routingStrategy,
     candidatePool,
-    weights,
+    weights: configWeights,
     explorationRate,
     budgetCap: configBudgetCap,
     budgetFallback: configBudgetFallback,
@@ -180,6 +181,17 @@ export async function resolveAutoStrategyOrder(
   const budgetFallback = requestBudgetFallback ?? configBudgetFallback;
   const requestModePack = resolveRequestModePack(relayOptions?.mode);
   const modePack = requestModePack.override ? requestModePack.modePack : configModePack;
+  // #7008: `weights` must track the *effective* (post-override) modePack, not just
+  // the combo's stored one. `selectAutoProvider()` (engine.ts) already re-derives
+  // weights internally from the `modePack` it's given, so it correctly reacts to a
+  // per-request X-OmniRoute-Mode override — but `scoreAutoTargets()` (the fallback
+  // ranking below) has no such re-derivation and only ever sees whatever `weights`
+  // it's handed. Without this recompute, a request overriding e.g. `quality-first`
+  // to `ship-fast` would select its primary target under ship-fast weights but rank
+  // every fallback under the stale quality-first weights — the same
+  // select-under-one-policy/rank-under-another bug this module's original fix
+  // (parseAutoConfig honoring the combo's own stored modePack) set out to close.
+  const weights = modePack ? getModePack(modePack) || configWeights : configWeights;
   if (requestModePack.override || requestBudgetCap !== undefined || requestBudgetFallback !== undefined) {
     log.debug?.(
       "COMBO",
diff --git a/open-sse/services/combo/sessionStickiness.ts b/open-sse/services/combo/sessionStickiness.ts
index d6d6eb06dd5..50bb9c796d7 100644
--- a/open-sse/services/combo/sessionStickiness.ts
+++ b/open-sse/services/combo/sessionStickiness.ts
@@ -254,6 +254,40 @@ const stickyMap = new Map<string, StickyEntry>();
 
 // ─── Helpers ─────────────────────────────────────────────────────────────────
 
+/**
+ * #7270: Normalize a request body's user turns into a `{role, content}[]` view for
+ * stickiness-key derivation, covering both wire formats:
+ *   - Chat Completions (`/v1/chat/completions`) → turns live in `.messages`.
+ *   - OpenAI Responses API (`/v1/responses`) → turns live in `.input`, which may be a
+ *     plain string OR an array of message items; `.messages` is never populated. Array
+ *     items may themselves be bare strings (shorthand for a user message) — the same
+ *     shape `responsesInputNormalization.ts`'s `normalizeCodexResponsesInputItem`
+ *     already special-cases — so those are mapped to `{role: "user", content: item}`.
+ * Combo target ordering runs BEFORE per-target format translation, so without this
+ * the Responses-API key resolved to null and stickiness silently no-oped for the
+ * entire surface (round-robin/random/strict-random all re-ordered every turn).
+ * `.messages` takes precedence when present (Chat Completions), then `.input`.
+ * Returns null when neither carrier yields turns (fail-open, same as deriveMessageHash).
+ */
+export function normalizeStickinessMessages(
+  body: { messages?: unknown; input?: unknown } | null | undefined
+): Array<{ role?: string; content?: unknown }> | null {
+  if (!body || typeof body !== "object") return null;
+  const { messages, input } = body as { messages?: unknown; input?: unknown };
+  if (Array.isArray(messages) && messages.length > 0) {
+    return messages as Array<{ role?: string; content?: unknown }>;
+  }
+  if (typeof input === "string" && input.length > 0) {
+    return [{ role: "user", content: input }];
+  }
+  if (Array.isArray(input) && input.length > 0) {
+    return input.map((item) =>
+      typeof item === "string" ? { role: "user", content: item } : item
+    ) as Array<{ role?: string; content?: unknown }>;
+  }
+  return null;
+}
+
 /**
  * Derive a stable 16-hex-char session key from the first user message content.
  * Returns null when the message cannot be extracted (fail-open).
diff --git a/open-sse/services/combo/targetExhaustion.ts b/open-sse/services/combo/targetExhaustion.ts
index 7091b483ef1..66f7ef5cd20 100644
--- a/open-sse/services/combo/targetExhaustion.ts
+++ b/open-sse/services/combo/targetExhaustion.ts
@@ -21,7 +21,7 @@ import {
   isProviderExhaustedReason,
 } from "../accountFallback.ts";
 import { RateLimitReason } from "../../config/constants.ts";
-import { isProviderCircuitOpenResult } from "./comboPredicates.ts";
+import { isProviderCircuitOpenResult, isRequestScopedUpstreamFailure } from "./comboPredicates.ts";
 import type { ComboLogger, ResolvedComboTarget } from "./types.ts";
 
 // Connection-level failure statuses: the provider connection itself is likely bad (upstream
@@ -104,7 +104,15 @@ export function applyComboTargetExhaustion(
     if (result.status === 429 && !isTokenLimitBreach && provider && provider !== "unknown") {
       transientRateLimitedProviders.add(provider);
     }
-    markConnectionLevelExhaustion(target, { result, errorText, sets, log, tag, rawModel });
+    markConnectionLevelExhaustion(target, {
+      result,
+      errorText,
+      sets,
+      log,
+      tag,
+      rawModel,
+      structuredError,
+    });
   }
 
   return providerExhausted;
@@ -120,16 +128,17 @@ function markConnectionLevelExhaustion(
   target: ResolvedComboTarget,
   opts: Pick<
     ApplyComboTargetExhaustionOptions,
-    "result" | "errorText" | "sets" | "log" | "tag" | "rawModel"
+    "result" | "errorText" | "sets" | "log" | "tag" | "rawModel" | "structuredError"
   >
 ): void {
-  const { result, errorText, sets, log, tag, rawModel } = opts;
+  const { result, errorText, sets, log, tag, rawModel, structuredError } = opts;
   const provider = target.provider;
   if (
     !provider ||
     provider === "unknown" ||
     !CONNECTION_LEVEL_ERROR_STATUSES.includes(result.status) ||
     isProviderCircuitOpenResult(result, errorText) ||
+    isRequestScopedUpstreamFailure(structuredError) ||
     // #5085: empty-content 502 is a healthy connection returning no body — model-level, not
     // connection-level. Don't exhaust the provider; let the remaining legs (incl. same-provider)
     // be tried in-request.
diff --git a/open-sse/services/combo/targetSorters.ts b/open-sse/services/combo/targetSorters.ts
index 3432b639fe5..7ddd3e45098 100644
--- a/open-sse/services/combo/targetSorters.ts
+++ b/open-sse/services/combo/targetSorters.ts
@@ -104,41 +104,23 @@ export async function sortTargetsByCost(targets: ResolvedComboTarget[]) {
     .filter((target): target is ResolvedComboTarget => target !== null);
 }
 
-/**
- * Sort models by usage count (least-used first) for least-used strategy
- * @param {Array<string>} models - Model strings
- * @param {string} comboName - Combo name for metrics lookup
- * @returns {Array<string>} Sorted model strings
- */
-export function sortModelsByUsage(models: string[], comboName: string): string[] {
+export function sortTargetsByUsage(targets: ResolvedComboTarget[], comboName: string) {
   const metrics = getComboMetrics(comboName);
-  if (!metrics?.byModel) return models;
-
-  const withUsage = models.map((modelStr) => ({
-    modelStr,
-    requests: metrics.byModel[modelStr]?.requests ?? 0,
-  }));
+  if (!metrics) return targets;
+
+  // Key on executionKey (unique per model + account) so a combo that repeats the
+  // same modelStr across DISTINCT accounts distributes by per-account usage
+  // instead of by the shared modelStr. The old code grouped targets under
+  // modelStr and read byModel[modelStr] (which aggregates every account of that
+  // model), so all accounts collapsed into one bucket and the first account
+  // always won — exhausting it while the others stayed idle (#7015). Per-target
+  // usage lives in byTarget[executionKey]; unknown targets rank as 0.
+  const withUsage = targets.map((target) => {
+    const requests = metrics.byTarget?.[target.executionKey]?.requests ?? 0;
+    return { target, requests };
+  });
   withUsage.sort((a, b) => a.requests - b.requests);
-  return withUsage.map((e) => e.modelStr);
-}
-
-export function sortTargetsByUsage(targets: ResolvedComboTarget[], comboName: string) {
-  const orderedModels = sortModelsByUsage(
-    targets.map((target) => target.modelStr),
-    comboName
-  );
-  const byModel = new Map<string, ResolvedComboTarget[]>();
-  for (const target of targets) {
-    const queue = byModel.get(target.modelStr) || [];
-    queue.push(target);
-    byModel.set(target.modelStr, queue);
-  }
-  return orderedModels
-    .map((modelStr) => {
-      const queue = byModel.get(modelStr);
-      return queue?.shift() || null;
-    })
-    .filter((target): target is ResolvedComboTarget => target !== null);
+  return withUsage.map((e) => e.target);
 }
 
 function getP2CTargetScore(
diff --git a/open-sse/services/combo/validateQuality.ts b/open-sse/services/combo/validateQuality.ts
index 99805eb35ca..f5330ae7736 100644
--- a/open-sse/services/combo/validateQuality.ts
+++ b/open-sse/services/combo/validateQuality.ts
@@ -566,8 +566,30 @@ export async function validateResponseQuality(
   const reasoningContent = message.reasoning_content ?? message.reasoning;
   const hasReasoningContent =
     typeof reasoningContent === "string" && reasoningContent.trim().length > 0;
-  const hasContent =
-    (content !== null && content !== undefined && content !== "") || hasReasoningContent;
+  // Issue #7000: content can be a string, an array of content parts
+  // (multimodal), or null. An empty array [] or an array of empty parts
+  // must NOT count as valid content — only arrays with at least one
+  // non-empty text/image part do.
+  let hasContent: boolean;
+  if (Array.isArray(content)) {
+    hasContent = content.some(
+      (part) =>
+        !!part &&
+        typeof part === "object" &&
+        ((typeof (part as Record<string, unknown>).text === "string" &&
+          ((part as Record<string, string>).text as string).trim().length > 0) ||
+          (part as Record<string, unknown>).type === "image_url" ||
+          (part as Record<string, unknown>).type === "input_audio" ||
+          (part as Record<string, unknown>).type === "file")
+    );
+  } else {
+    hasContent =
+      (content !== null &&
+        content !== undefined &&
+        content !== "" &&
+        (typeof content !== "string" || content.trim().length > 0)) ||
+      hasReasoningContent;
+  }
   const hasToolCalls = Array.isArray(toolCalls) && toolCalls.length > 0;
 
   if (!hasContent && !hasToolCalls) {
diff --git a/open-sse/services/compression/bodyAdapter.ts b/open-sse/services/compression/bodyAdapter.ts
index 9c42fbdd6ff..af407c510de 100644
--- a/open-sse/services/compression/bodyAdapter.ts
+++ b/open-sse/services/compression/bodyAdapter.ts
@@ -14,7 +14,11 @@ type ResponsesItem = {
   [key: string]: unknown;
 };
 
-const RESPONSES_MESSAGE_TYPES = new Set(["message", "function_call_output"]);
+const RESPONSES_MESSAGE_TYPES = new Set([
+  "message",
+  "function_call_output",
+  "custom_tool_call_output",
+]);
 const COMPRESSION_INPUT_INDEX = Symbol("compressionInputIndex");
 
 // Kiro envelope path back to the original tool-result text inside
@@ -60,11 +64,47 @@ function fromChatContent(nextContent: unknown, originalContent: unknown): unknow
   return nextContent;
 }
 
+function customToolOutputToChatContent(rawOutput: unknown): unknown {
+  if (typeof rawOutput !== "string") {
+    if (isRecord(rawOutput) && typeof rawOutput.output === "string") return rawOutput.output;
+    return rawOutput;
+  }
+
+  try {
+    const parsed = JSON.parse(rawOutput) as unknown;
+    if (isRecord(parsed) && typeof parsed.output === "string") return parsed.output;
+  } catch {
+    // Plain-text custom tool output is already in the form compression engines expect.
+  }
+  return rawOutput;
+}
+
+function restoreCustomToolOutput(nextContent: unknown, originalOutput: unknown): unknown {
+  if (typeof originalOutput === "string") {
+    try {
+      const parsed = JSON.parse(originalOutput) as unknown;
+      if (isRecord(parsed) && typeof parsed.output === "string") {
+        return JSON.stringify({ ...parsed, output: nextContent });
+      }
+    } catch {
+      // Preserve the original plain-text representation below.
+    }
+  }
+  if (isRecord(originalOutput) && typeof originalOutput.output === "string") {
+    return { ...originalOutput, output: nextContent };
+  }
+  return fromChatContent(nextContent, originalOutput);
+}
+
+function responsesToolOutputField(item: ResponsesItem): "output" | "content" {
+  return item.output !== null && item.output !== undefined ? "output" : "content";
+}
+
 function responsesItemToMessage(item: ResponsesItem): MessageLike | null {
   const type = typeof item.type === "string" ? item.type : "message";
   if (!RESPONSES_MESSAGE_TYPES.has(type)) return null;
 
-  if (type === "function_call_output") {
+  if (type === "function_call_output" || type === "custom_tool_call_output") {
     const rawOutput = item.output ?? item.content;
     // OpenAI Responses shape (Codex): body.input holds Responses items. When
     // output is a JSON object (not a string or content array), serialise it so
@@ -77,7 +117,12 @@ function responsesItemToMessage(item: ResponsesItem): MessageLike | null {
       !Array.isArray(rawOutput);
     return {
       role: "tool",
-      content: isObjectOutput ? JSON.stringify(rawOutput) : toChatContent(rawOutput),
+      content:
+        type === "custom_tool_call_output"
+          ? customToolOutputToChatContent(rawOutput)
+          : isObjectOutput
+            ? JSON.stringify(rawOutput)
+            : toChatContent(rawOutput),
     };
   }
 
@@ -89,10 +134,15 @@ function responsesItemToMessage(item: ResponsesItem): MessageLike | null {
 
 function messageToResponsesItem(message: MessageLike, originalItem: ResponsesItem): ResponsesItem {
   const type = typeof originalItem.type === "string" ? originalItem.type : "message";
-  if (type === "function_call_output") {
+  if (type === "function_call_output" || type === "custom_tool_call_output") {
+    const outputField = responsesToolOutputField(originalItem);
+    const originalOutput = originalItem[outputField];
     return {
       ...originalItem,
-      output: fromChatContent(message.content, originalItem.output),
+      [outputField]:
+        type === "custom_tool_call_output"
+          ? restoreCustomToolOutput(message.content, originalOutput)
+          : fromChatContent(message.content, originalOutput),
     };
   }
 
@@ -332,7 +382,12 @@ function rewriteKiroEntry(
     let trChanged = false;
     const nextContent = content.map((part, partIdx) => {
       if (!isRecord(part) || typeof part.text !== "string") return part;
-      const key = kiroPathKey({ scope, historyIndex, toolResultIndex: trIdx, contentIndex: partIdx });
+      const key = kiroPathKey({
+        scope,
+        historyIndex,
+        toolResultIndex: trIdx,
+        contentIndex: partIdx,
+      });
       const rewritten = rewrites.get(key);
       if (rewritten === undefined || rewritten === part.text) return part;
       trChanged = true;
diff --git a/open-sse/services/compression/engines/ccr/index.ts b/open-sse/services/compression/engines/ccr/index.ts
index 3308d201edd..5aee6c282db 100644
--- a/open-sse/services/compression/engines/ccr/index.ts
+++ b/open-sse/services/compression/engines/ccr/index.ts
@@ -26,8 +26,8 @@
  *     does not affect another's (cross-tenant state drift protection).
  *
  * Memory bound:
- *   - Both `ccrStore` and `retrievalCounts` are capped at MAX_CCR_ENTRIES
- *     entries using FIFO eviction (Map insertion-order guarantees).
+ *   - Entries are capped by count, global bytes, per-principal bytes, block bytes and TTL.
+ *   - Expired and least-recently-used entries are removed before a store is rejected.
  *
  * Conservative guards:
  *   - Never touch `role: "system"`.
@@ -60,12 +60,69 @@ const RETRIEVAL_THRESHOLD = 3;
  * ramp (only the >= threshold cliff remains — the legacy binary behavior).
  */
 const RETRIEVAL_RAMP_FACTOR_DEFAULT = 2;
-/**
- * Maximum number of entries in each bounded store.
- * When inserting beyond this cap, the oldest entry (Map insertion order) is evicted.
- * 5 000 entries × ~2 KB average ≈ 10 MB upper bound for each map.
- */
+/** Maximum number of entries in the principal-scoped, LRU-ordered store. */
 export const MAX_CCR_ENTRIES = 5_000;
+export const MAX_CCR_BLOCK_BYTES = 2 * 1024 * 1024;
+export const MAX_CCR_PRINCIPAL_BYTES = 16 * 1024 * 1024;
+export const MAX_CCR_GLOBAL_BYTES = 64 * 1024 * 1024;
+export const DEFAULT_CCR_TTL_SECONDS = 24 * 60 * 60;
+export const MAX_CCR_TTL_SECONDS = 7 * 24 * 60 * 60;
+export const MAX_CCR_MCP_FULL_BYTES = 256 * 1024;
+
+export type CcrEntrySource = "compression" | "mcp" | "ionizer" | "session-dedup";
+
+export interface CcrEntryMetadata {
+  hash: string;
+  bytes: number;
+  chars: number;
+  lines: number;
+  contentType: string;
+  source: CcrEntrySource;
+  createdAt: number;
+  lastAccessedAt: number;
+  expiresAt: number;
+  retrievalCount: number;
+}
+
+type CcrEntry = Omit<CcrEntryMetadata, "retrievalCount"> & {
+  principalId: string;
+  content: string;
+};
+
+export interface StoreCcrBlockOptions {
+  contentType?: string;
+  source?: CcrEntrySource;
+  ttlSeconds?: number;
+  now?: number;
+}
+
+export type StoreCcrBlockResult =
+  | { stored: true; hash: string; metadata: CcrEntryMetadata }
+  | {
+      stored: false;
+      hash: string;
+      reason: "block_too_large" | "principal_budget_exceeded" | "global_budget_exceeded";
+    };
+
+export interface CcrStoreStats {
+  storage: "memory";
+  entries: number;
+  bytes: number;
+  limits: {
+    maxEntries: number;
+    maxBlockBytes: number;
+    maxPrincipalBytes: number;
+    maxGlobalBytes: number;
+    defaultTtlSeconds: number;
+    maxTtlSeconds: number;
+    maxMcpFullBytes: number;
+  };
+  lifecycle: {
+    expiredEvictions: number;
+    capacityEvictions: number;
+    rejectedStores: number;
+  };
+}
 
 // ─── principal-scoped, bounded content store ──────────────────────────────────
 
@@ -73,9 +130,12 @@ export const MAX_CCR_ENTRIES = 5_000;
  * Store key = `${principalId ?? "__anon__"} ${contentHash}`.
  * Using a compound key scopes data to the principal that stored it.
  */
-const ccrStore = new Map<string, string>();
-/** Retrieval counter store — same scoping as ccrStore. */
+const ccrStore = new Map<string, CcrEntry>();
 const retrievalCounts = new Map<string, number>();
+const principalBytesMap = new Map<string, number>();
+let ccrTotalBytes = 0;
+type CcrLifecycleCounters = CcrStoreStats["lifecycle"];
+const lifecycleByPrincipal = new Map<string, CcrLifecycleCounters>();
 
 /** Sentinel used when no principalId is provided. */
 const ANON = "__anon__";
@@ -84,18 +144,88 @@ function buildStoreKey(hash: string, principalId?: string): string {
   return `${principalId ?? ANON} ${hash}`;
 }
 
-/**
- * Insert a value into a bounded Map, evicting the oldest entry when over the cap.
- */
-function boundedSet<V>(map: Map<string, V>, key: string, value: V): void {
-  if (!map.has(key) && map.size >= MAX_CCR_ENTRIES) {
-    // Map preserves insertion order — the first iterator result is the oldest entry.
-    const firstKey = map.keys().next().value;
-    if (firstKey !== undefined) {
-      map.delete(firstKey);
+function readLifecycleCounters(principalId: string): CcrLifecycleCounters {
+  return (
+    lifecycleByPrincipal.get(principalId) ?? {
+      expiredEvictions: 0,
+      capacityEvictions: 0,
+      rejectedStores: 0,
     }
+  );
+}
+
+function mutableLifecycleCounters(principalId: string): CcrLifecycleCounters {
+  const existing = lifecycleByPrincipal.get(principalId);
+  if (existing) return existing;
+  const counters = { expiredEvictions: 0, capacityEvictions: 0, rejectedStores: 0 };
+  lifecycleByPrincipal.set(principalId, counters);
+  return counters;
+}
+
+function publicMetadata(entry: CcrEntry): CcrEntryMetadata {
+  const { principalId: _principalId, content: _content, ...metadata } = entry;
+  return {
+    ...metadata,
+    retrievalCount: retrievalCounts.get(buildStoreKey(entry.hash, entry.principalId)) ?? 0,
+  };
+}
+
+function setRetrievalCount(key: string, count: number): void {
+  if (!retrievalCounts.has(key) && retrievalCounts.size >= MAX_CCR_ENTRIES) {
+    const oldestKey = retrievalCounts.keys().next().value;
+    if (oldestKey !== undefined) retrievalCounts.delete(oldestKey);
   }
-  map.set(key, value);
+  retrievalCounts.delete(key);
+  retrievalCounts.set(key, count);
+}
+
+function removeEntry(key: string, reason?: "expired" | "capacity"): boolean {
+  const entry = ccrStore.get(key);
+  if (!entry) return false;
+  ccrStore.delete(key);
+  ccrTotalBytes = Math.max(0, ccrTotalBytes - entry.bytes);
+  const remainingPrincipalBytes = Math.max(
+    0,
+    (principalBytesMap.get(entry.principalId) ?? 0) - entry.bytes
+  );
+  if (remainingPrincipalBytes === 0) principalBytesMap.delete(entry.principalId);
+  else principalBytesMap.set(entry.principalId, remainingPrincipalBytes);
+  const counters = mutableLifecycleCounters(entry.principalId);
+  if (reason === "expired") counters.expiredEvictions++;
+  if (reason === "capacity") counters.capacityEvictions++;
+  return true;
+}
+
+function purgeExpired(now = Date.now()): void {
+  for (const [key, entry] of ccrStore) {
+    if (entry.expiresAt <= now) removeEntry(key, "expired");
+  }
+}
+
+function getActiveEntry(key: string, now = Date.now()): CcrEntry | null {
+  const entry = ccrStore.get(key);
+  if (!entry) return null;
+  if (entry.expiresAt <= now) {
+    removeEntry(key, "expired");
+    return null;
+  }
+  return entry;
+}
+
+function principalBytes(principalId: string): number {
+  return principalBytesMap.get(principalId) ?? 0;
+}
+
+function evictOldestMatching(predicate: (entry: CcrEntry) => boolean): boolean {
+  for (const [key, entry] of ccrStore) {
+    if (predicate(entry)) return removeEntry(key, "capacity");
+  }
+  return false;
+}
+
+function normalizeTtlSeconds(value: number | undefined): number {
+  if (!Number.isFinite(value) || value === undefined) return DEFAULT_CCR_TTL_SECONDS;
+  return Math.max(60, Math.min(MAX_CCR_TTL_SECONDS, Math.floor(value)));
 }
 
 /**
@@ -107,26 +237,114 @@ function hashContent(text: string): string {
   return crypto.createHash("sha256").update(text).digest("hex").slice(0, 24);
 }
 
+function rejectStore(
+  hash: string,
+  owner: string,
+  reason: Exclude<StoreCcrBlockResult, { stored: true }>["reason"]
+): StoreCcrBlockResult {
+  mutableLifecycleCounters(owner).rejectedStores++;
+  return { stored: false, hash, reason };
+}
+
+function enforcePrincipalBudget(owner: string, bytes: number): boolean {
+  while (
+    principalBytes(owner) + bytes > MAX_CCR_PRINCIPAL_BYTES &&
+    evictOldestMatching((entry) => entry.principalId === owner)
+  ) {
+    // Evict the owner's least-recently-used entries until this block fits.
+  }
+  return principalBytes(owner) + bytes <= MAX_CCR_PRINCIPAL_BYTES;
+}
+
+function enforceGlobalBudget(bytes: number): boolean {
+  while (
+    (ccrStore.size >= MAX_CCR_ENTRIES || ccrTotalBytes + bytes > MAX_CCR_GLOBAL_BYTES) &&
+    evictOldestMatching(() => true)
+  ) {
+    // Enforce both entry and global byte caps with LRU eviction.
+  }
+  return ccrTotalBytes + bytes <= MAX_CCR_GLOBAL_BYTES;
+}
+
 /**
  * Store a block in the CCR store under the given principal.
  * Returns the 24-hex content hash (for embedding in the marker).
  */
-export function storeBlock(text: string, principalId?: string): string {
+export function tryStoreBlock(
+  text: string,
+  principalId?: string,
+  options: StoreCcrBlockOptions = {}
+): StoreCcrBlockResult {
   const hash = hashContent(text);
+  const owner = principalId ?? ANON;
   const key = buildStoreKey(hash, principalId);
-  if (!ccrStore.has(key)) {
-    boundedSet(ccrStore, key, text);
+  const now = options.now ?? Date.now();
+  purgeExpired(now);
+
+  const existing = ccrStore.get(key);
+  if (existing) {
+    existing.lastAccessedAt = now;
+    existing.expiresAt = now + normalizeTtlSeconds(options.ttlSeconds) * 1000;
+    ccrStore.delete(key);
+    ccrStore.set(key, existing);
+    return { stored: true, hash, metadata: publicMetadata(existing) };
+  }
+
+  const bytes = Buffer.byteLength(text, "utf8");
+  if (bytes > MAX_CCR_BLOCK_BYTES) {
+    return rejectStore(hash, owner, "block_too_large");
+  }
+
+  if (!enforcePrincipalBudget(owner, bytes)) {
+    return rejectStore(hash, owner, "principal_budget_exceeded");
+  }
+
+  if (!enforceGlobalBudget(bytes)) {
+    return rejectStore(hash, owner, "global_budget_exceeded");
   }
-  return hash;
+
+  const ttlSeconds = normalizeTtlSeconds(options.ttlSeconds);
+  const entry: CcrEntry = {
+    hash,
+    principalId: owner,
+    content: text,
+    bytes,
+    chars: text.length,
+    lines: text.length === 0 ? 0 : text.split("\n").length,
+    contentType: options.contentType?.trim().slice(0, 128) || "text/plain",
+    source: options.source ?? "compression",
+    createdAt: now,
+    lastAccessedAt: now,
+    expiresAt: now + ttlSeconds * 1000,
+  };
+  ccrStore.set(key, entry);
+  ccrTotalBytes += bytes;
+  principalBytesMap.set(owner, principalBytes(owner) + bytes);
+  return { stored: true, hash, metadata: publicMetadata(entry) };
+}
+
+export function storeBlock(
+  text: string,
+  principalId?: string,
+  options: StoreCcrBlockOptions = {}
+): string {
+  const result = tryStoreBlock(text, principalId, options);
+  if (!result.stored) throw new RangeError(`CCR store rejected block: ${result.reason}`);
+  return result.hash;
 }
 
 /**
  * Retrieve the verbatim block for a given hash and principal.
  * Returns null if not found or if the principal does not match the stored key.
  */
-export function retrieveBlock(hash: string, principalId?: string): string | null {
+export function retrieveBlock(hash: string, principalId?: string, now = Date.now()): string | null {
   const key = buildStoreKey(hash, principalId);
-  return ccrStore.get(key) ?? null;
+  const entry = getActiveEntry(key, now);
+  if (!entry) return null;
+  entry.lastAccessedAt = now;
+  ccrStore.delete(key);
+  ccrStore.set(key, entry);
+  return entry.content;
 }
 
 /**
@@ -134,7 +352,7 @@ export function retrieveBlock(hash: string, principalId?: string): string | null
  */
 export function recordRetrieval(hash: string, principalId?: string): void {
   const key = buildStoreKey(hash, principalId);
-  boundedSet(retrievalCounts, key, (retrievalCounts.get(key) ?? 0) + 1);
+  setRetrievalCount(key, (retrievalCounts.get(key) ?? 0) + 1);
 }
 
 /**
@@ -182,6 +400,65 @@ export function resolveRetrievalRampFactor(env: NodeJS.ProcessEnv = process.env)
 export function resetCcrStore(): void {
   ccrStore.clear();
   retrievalCounts.clear();
+  principalBytesMap.clear();
+  ccrTotalBytes = 0;
+  lifecycleByPrincipal.clear();
+}
+
+export function inspectCcrBlock(
+  hash: string,
+  principalId?: string,
+  now = Date.now()
+): CcrEntryMetadata | null {
+  const entry = getActiveEntry(buildStoreKey(hash, principalId), now);
+  return entry ? publicMetadata(entry) : null;
+}
+
+export function listCcrBlocks(
+  principalId?: string,
+  options: { offset?: number; limit?: number; now?: number } = {}
+): { entries: CcrEntryMetadata[]; total: number; offset: number; limit: number; hasMore: boolean } {
+  purgeExpired(options.now);
+  const owner = principalId ?? ANON;
+  const offset = Math.max(0, Math.floor(options.offset ?? 0));
+  const limit = Math.max(1, Math.min(100, Math.floor(options.limit ?? 25)));
+  const all: CcrEntryMetadata[] = [];
+  for (const entry of ccrStore.values()) {
+    if (entry.principalId === owner) all.push(publicMetadata(entry));
+  }
+  all.reverse();
+  return {
+    entries: all.slice(offset, offset + limit),
+    total: all.length,
+    offset,
+    limit,
+    hasMore: offset + limit < all.length,
+  };
+}
+
+export function deleteCcrBlock(hash: string, principalId?: string, _now = Date.now()): boolean {
+  return removeEntry(buildStoreKey(hash, principalId));
+}
+
+export function getCcrStoreStats(principalId?: string, now = Date.now()): CcrStoreStats {
+  purgeExpired(now);
+  const owner = principalId ?? ANON;
+  const entries = Array.from(ccrStore.values()).filter((entry) => entry.principalId === owner);
+  return {
+    storage: "memory",
+    entries: entries.length,
+    bytes: entries.reduce((sum, entry) => sum + entry.bytes, 0),
+    limits: {
+      maxEntries: MAX_CCR_ENTRIES,
+      maxBlockBytes: MAX_CCR_BLOCK_BYTES,
+      maxPrincipalBytes: MAX_CCR_PRINCIPAL_BYTES,
+      maxGlobalBytes: MAX_CCR_GLOBAL_BYTES,
+      defaultTtlSeconds: DEFAULT_CCR_TTL_SECONDS,
+      maxTtlSeconds: MAX_CCR_TTL_SECONDS,
+      maxMcpFullBytes: MAX_CCR_MCP_FULL_BYTES,
+    },
+    lifecycle: { ...readLifecycleCounters(owner) },
+  };
 }
 
 // ─── MCP tool handler (pure function) ────────────────────────────────────────
@@ -226,10 +503,21 @@ type MessageLike = {
 /**
  * Build a CCR marker string for a block.
  */
-function buildMarker(hash: string, charCount: number): string {
+export function buildCcrMarker(hash: string, charCount: number): string {
   return `[CCR retrieve hash=${hash} chars=${charCount}]`;
 }
 
+export function buildCcrReference(
+  hash: string,
+  charCount: number
+): {
+  hash: string;
+  uri: string;
+  marker: string;
+} {
+  return { hash, uri: `ccr://${hash}`, marker: buildCcrMarker(hash, charCount) };
+}
+
 /**
  * Replace a large text block with a CCR marker if it shrinks the content.
  * Returns the new text and a flag indicating whether replacement happened.
@@ -254,14 +542,15 @@ function maybeCcrReplace(
     return { text, replaced: false, hash: null };
   }
 
-  const marker = buildMarker(hash, text.length);
+  const marker = buildCcrMarker(hash, text.length);
 
   // Only replace if it actually shrinks
   if (marker.length >= text.length) {
     return { text, replaced: false, hash: null };
   }
 
-  storeBlock(text, principalId);
+  const stored = tryStoreBlock(text, principalId, { source: "compression" });
+  if (!stored.stored) return { text, replaced: false, hash: null };
   return { text: marker, replaced: true, hash };
 }
 
diff --git a/open-sse/services/compression/engines/ionizer/sample.ts b/open-sse/services/compression/engines/ionizer/sample.ts
index 172429d2a86..34fb0f38a17 100644
--- a/open-sse/services/compression/engines/ionizer/sample.ts
+++ b/open-sse/services/compression/engines/ionizer/sample.ts
@@ -1,5 +1,5 @@
 // open-sse/services/compression/engines/ionizer/sample.ts
-import { storeBlock } from "../ccr/index.ts";
+import { tryStoreBlock } from "../ccr/index.ts";
 
 type MessageLike = { role?: string; content?: unknown; [key: string]: unknown };
 
@@ -131,8 +131,7 @@ export interface IonizerPassResult {
 
 function isPlainObjectArray(v: unknown): v is Array<Record<string, unknown>> {
   return (
-    Array.isArray(v) &&
-    v.every((el) => el !== null && typeof el === "object" && !Array.isArray(el))
+    Array.isArray(v) && v.every((el) => el !== null && typeof el === "object" && !Array.isArray(el))
   );
 }
 
@@ -169,8 +168,12 @@ export function applyIonizerPass(
       });
       if (res.keptCount >= res.totalCount) return m;
 
-      const hash = storeBlock(serialized, opts.principalId);
-      const marker = `[ionizer: kept ${res.keptCount}/${res.totalCount} rows; full → CCR retrieve hash=${hash} chars=${serialized.length}]`;
+      const stored = tryStoreBlock(serialized, opts.principalId, {
+        contentType: "application/json",
+        source: "ionizer",
+      });
+      if (!stored.stored) return m;
+      const marker = `[ionizer: kept ${res.keptCount}/${res.totalCount} rows; full → CCR retrieve hash=${stored.hash} chars=${serialized.length}]`;
       const newContent = `${JSON.stringify(res.kept)}\n${marker}`;
       if (newContent.length >= serialized.length) return m;
 
@@ -194,7 +197,9 @@ export function runIonizerPass(
   principalId?: string
 ): IonizerPassResult {
   if (stepConfig["enabled"] === false) return { messages, ionizedCount: 0 };
-  const threshold = typeof stepConfig["threshold"] === "number" ? (stepConfig["threshold"] as number) : 200;
-  const targetRows = typeof stepConfig["targetRows"] === "number" ? (stepConfig["targetRows"] as number) : 50;
+  const threshold =
+    typeof stepConfig["threshold"] === "number" ? (stepConfig["threshold"] as number) : 200;
+  const targetRows =
+    typeof stepConfig["targetRows"] === "number" ? (stepConfig["targetRows"] as number) : 50;
   return applyIonizerPass(messages, { threshold, targetRows, principalId });
 }
diff --git a/open-sse/services/compression/engines/rtk/filterLoader.ts b/open-sse/services/compression/engines/rtk/filterLoader.ts
index aa10dd8224f..184d94417f0 100644
--- a/open-sse/services/compression/engines/rtk/filterLoader.ts
+++ b/open-sse/services/compression/engines/rtk/filterLoader.ts
@@ -4,6 +4,7 @@ import os from "node:os";
 import crypto from "node:crypto";
 import { detectCommandType } from "./commandDetector.ts";
 import { validateRtkFilter, type RtkFilterDefinition } from "./filterSchema.ts";
+import { parseRtkTomlV1, RtkTomlCompatibilityError } from "./tomlCompatibility.ts";
 
 let cache: RtkFilterDefinition[] | null = null;
 let cacheKey: string | null = null;
@@ -29,6 +30,7 @@ function cachedMatchPattern(pattern: string, value: string): boolean {
 
 export interface RtkFilterLoadDiagnostic {
   source: "project" | "global" | "builtin";
+  format?: "omniroute-json" | "rtk-toml-v1";
   path?: string;
   level: "warning" | "error";
   message: string;
@@ -38,6 +40,7 @@ interface FilterSource {
   source: "project" | "global" | "builtin";
   path: string;
   trusted: boolean;
+  format: "omniroute-json" | "rtk-toml-v1";
 }
 
 interface RtkFilterLoadOptions {
@@ -95,8 +98,12 @@ function projectFiltersTrusted(
   try {
     const filtersHash = sha256(fs.readFileSync(filtersPath, "utf8"));
     const trust = JSON.parse(fs.readFileSync(trustPath, "utf8")) as Record<string, unknown>;
-    const trustedHash =
-      typeof trust.filtersSha256 === "string"
+    const isToml = filtersPath.endsWith(".toml");
+    const trustedHash = isToml
+      ? typeof trust.filtersTomlSha256 === "string"
+        ? trust.filtersTomlSha256
+        : null
+      : typeof trust.filtersSha256 === "string"
         ? trust.filtersSha256
         : typeof trust.trustedFiltersSha256 === "string"
           ? trust.trustedFiltersSha256
@@ -110,29 +117,52 @@ function projectFiltersTrusted(
 
 function collectFilterSources(options: RtkFilterLoadOptions = {}): FilterSource[] {
   const sources: FilterSource[] = [];
-  const projectPath = path.join(process.cwd(), ".rtk", "filters.json");
-  if (options.customFiltersEnabled !== false && fs.existsSync(projectPath)) {
-    const trusted = projectFiltersTrusted(projectPath, options.trustProjectFilters === true);
+  if (options.customFiltersEnabled !== false) {
+    collectProjectFilterSources(sources, options);
+    collectGlobalFilterSources(sources);
+  }
+  collectBuiltinFilterSources(sources);
+  return sources;
+}
+
+function collectProjectFilterSources(sources: FilterSource[], options: RtkFilterLoadOptions): void {
+  const projectCandidates = [
+    { path: path.join(process.cwd(), ".rtk", "filters.toml"), format: "rtk-toml-v1" as const },
+    { path: path.join(process.cwd(), ".rtk", "filters.json"), format: "omniroute-json" as const },
+  ];
+  for (const candidate of projectCandidates) {
+    if (!fs.existsSync(candidate.path)) continue;
+    const trusted = projectFiltersTrusted(candidate.path, options.trustProjectFilters === true);
     if (trusted === true) {
-      sources.push({ source: "project", path: projectPath, trusted: true });
-    } else {
-      diagnostics.push({
-        source: "project",
-        path: projectPath,
-        level: "warning",
-        message:
-          trusted === "changed"
-            ? "Project RTK filters changed after trust and were skipped"
-            : "Project RTK filters are untrusted and were skipped",
-      });
+      sources.push({ source: "project", ...candidate, trusted: true });
+      continue;
     }
+    diagnostics.push({
+      source: "project",
+      format: candidate.format,
+      path: candidate.path,
+      level: "warning",
+      message:
+        trusted === "changed"
+          ? "Project RTK filters changed after trust and were skipped"
+          : "Project RTK filters are untrusted and were skipped",
+    });
   }
+}
 
-  const globalPath = path.join(getDataDir(), "rtk", "filters.json");
-  if (options.customFiltersEnabled !== false && fs.existsSync(globalPath)) {
-    sources.push({ source: "global", path: globalPath, trusted: true });
+function collectGlobalFilterSources(sources: FilterSource[]): void {
+  const globalCandidates = [
+    { path: path.join(getDataDir(), "rtk", "filters.toml"), format: "rtk-toml-v1" as const },
+    { path: path.join(getDataDir(), "rtk", "filters.json"), format: "omniroute-json" as const },
+  ];
+  for (const candidate of globalCandidates) {
+    if (fs.existsSync(candidate.path)) {
+      sources.push({ source: "global", ...candidate, trusted: true });
+    }
   }
+}
 
+function collectBuiltinFilterSources(sources: FilterSource[]): void {
   const builtinDir = getFiltersDir();
   if (fs.existsSync(builtinDir)) {
     let builtinFiles: string[] = [];
@@ -152,25 +182,56 @@ function collectFilterSources(options: RtkFilterLoadOptions = {}): FilterSource[
         source: "builtin",
         path: path.join(builtinDir, file),
         trusted: true,
+        format: "omniroute-json",
       });
     }
   }
-
-  return sources;
 }
 
 function parseFilterFile(source: FilterSource): RtkFilterDefinition[] {
   try {
-    const parsed = JSON.parse(fs.readFileSync(source.path, "utf8"));
-    const entries = Array.isArray(parsed) ? parsed : [parsed];
-    return entries.map(validateRtkFilter);
+    const content = fs.readFileSync(source.path, "utf8");
+    const definitions =
+      source.format === "rtk-toml-v1"
+        ? (() => {
+            const result = parseRtkTomlV1(content);
+            if (!result.passed) {
+              throw new Error("one or more inline tests failed");
+            }
+            for (const warning of result.warnings) {
+              diagnostics.push({
+                source: source.source,
+                format: source.format,
+                path: source.path,
+                level: "warning",
+                message: warning,
+              });
+            }
+            return result.filters;
+          })()
+        : (() => {
+            const parsed = JSON.parse(content);
+            const entries = Array.isArray(parsed) ? parsed : [parsed];
+            return entries.map(validateRtkFilter);
+          })();
+    return definitions.map((definition) => ({
+      ...definition,
+      source: source.source,
+      sourceFormat: definition.sourceFormat ?? source.format,
+    }));
   } catch (error) {
-    const message = error instanceof Error ? error.message : String(error);
+    const message =
+      error instanceof RtkTomlCompatibilityError
+        ? error.publicMessage
+        : error instanceof Error
+          ? error.message
+          : String(error);
     if (source.source === "builtin") {
       throw new Error(`Invalid RTK filter ${path.basename(source.path)}: ${message}`);
     }
     diagnostics.push({
       source: source.source,
+      format: source.format,
       path: source.path,
       level: "warning",
       message: `Invalid custom RTK filter skipped: ${message}`,
@@ -194,7 +255,16 @@ export function loadRtkFilters(options: RtkFilterLoadOptions = {}): RtkFilterDef
     filters.push(...parseFilterFile(source));
   }
 
-  const sorted = filters.sort((a, b) => b.priority - a.priority || a.id.localeCompare(b.id));
+  const sourceRank = { project: 3, global: 2, builtin: 1 } as const;
+  const formatRank = { "rtk-toml-v1": 2, "omniroute-json": 1 } as const;
+  const sorted = filters.sort(
+    (a, b) =>
+      sourceRank[b.source ?? "builtin"] - sourceRank[a.source ?? "builtin"] ||
+      formatRank[b.sourceFormat ?? "omniroute-json"] -
+        formatRank[a.sourceFormat ?? "omniroute-json"] ||
+      b.priority - a.priority ||
+      a.id.localeCompare(b.id)
+  );
   cache = sorted;
   cacheKey = currentCacheKey;
   return sorted;
@@ -208,7 +278,14 @@ export function getRtkFilterLoadDiagnostics(): RtkFilterLoadDiagnostic[] {
 export function getRtkFilterCatalog(): Array<
   Pick<
     RtkFilterDefinition,
-    "id" | "name" | "description" | "commandTypes" | "category" | "priority"
+    | "id"
+    | "name"
+    | "description"
+    | "commandTypes"
+    | "category"
+    | "priority"
+    | "source"
+    | "sourceFormat"
   >
 > {
   return loadRtkFilters().map((filter) => ({
@@ -218,6 +295,8 @@ export function getRtkFilterCatalog(): Array<
     commandTypes: filter.commandTypes,
     category: filter.category,
     priority: filter.priority,
+    source: filter.source,
+    sourceFormat: filter.sourceFormat,
   }));
 }
 
@@ -229,17 +308,25 @@ export function matchRtkFilter(
   const detection = detectCommandType(text, command);
   const detectedCommand = detection.command ?? command ?? "";
   const filters = loadRtkFilters(options);
-  return (
-    filters.find((filter) => filter.commandTypes.includes(detection.type)) ??
-    filters.find(
-      (filter) =>
-        detectedCommand &&
-        filter.commandPatterns.some((pattern) => cachedMatchPattern(pattern, detectedCommand))
-    ) ??
-    filters.find((filter) =>
-      filter.matchPatterns.some((pattern) => cachedMatchPattern(pattern, text))
-    ) ??
-    filters.find((filter) => filter.commandTypes.includes("generic-output")) ??
-    null
-  );
+  for (const source of ["project", "global", "builtin"] as const) {
+    const scoped = filters.filter((filter) => (filter.source ?? "builtin") === source);
+    const matched =
+      scoped.find(
+        (filter) =>
+          filter.sourceFormat === "rtk-toml-v1" &&
+          detectedCommand &&
+          filter.commandPatterns.some((pattern) => cachedMatchPattern(pattern, detectedCommand))
+      ) ??
+      scoped.find((filter) => filter.commandTypes.includes(detection.type)) ??
+      scoped.find(
+        (filter) =>
+          detectedCommand &&
+          filter.commandPatterns.some((pattern) => cachedMatchPattern(pattern, detectedCommand))
+      ) ??
+      scoped.find((filter) =>
+        filter.matchPatterns.some((pattern) => cachedMatchPattern(pattern, text))
+      );
+    if (matched) return matched;
+  }
+  return filters.find((filter) => filter.commandTypes.includes("generic-output")) ?? null;
 }
diff --git a/open-sse/services/compression/engines/rtk/filterSchema.ts b/open-sse/services/compression/engines/rtk/filterSchema.ts
index 2acceb7cabe..9eb9d5b621c 100644
--- a/open-sse/services/compression/engines/rtk/filterSchema.ts
+++ b/open-sse/services/compression/engines/rtk/filterSchema.ts
@@ -137,6 +137,12 @@ export interface RtkFilterDefinition {
   maxLines: number;
   preserveHead: number;
   preserveTail: number;
+  /** Exact RTK TOML schema-v1 head/tail stages. Undefined for OmniRoute-native JSON filters. */
+  rtkTomlHeadLines?: number;
+  rtkTomlTailLines?: number;
+  rtkTomlMaxLines?: number;
+  sourceFormat?: "omniroute-json" | "rtk-toml-v1";
+  source?: "project" | "global" | "builtin";
   tests: Array<{ name: string; input: string; expected: string; command?: string }>;
 }
 
diff --git a/open-sse/services/compression/engines/rtk/lineFilter.ts b/open-sse/services/compression/engines/rtk/lineFilter.ts
index f1311858ba8..52a6a83870d 100644
--- a/open-sse/services/compression/engines/rtk/lineFilter.ts
+++ b/open-sse/services/compression/engines/rtk/lineFilter.ts
@@ -54,6 +54,41 @@ function normalizeStderrPrefix(line: string): string {
   return line.replace(/^\s*(?:stderr|err)\s*(?:\||:)\s*/i, "");
 }
 
+function applyRtkTomlLineLimits(
+  lines: string[],
+  filter: RtkFilterDefinition,
+  appliedRules: string[]
+): string[] {
+  const head = filter.rtkTomlHeadLines;
+  const tail = filter.rtkTomlTailLines;
+  const total = lines.length;
+
+  if (head !== undefined && tail !== undefined) {
+    if (total > head + tail) {
+      lines = [
+        ...lines.slice(0, head),
+        `... (${total - head - tail} lines omitted)`,
+        ...(tail > 0 ? lines.slice(-tail) : []),
+      ];
+      appliedRules.push(`${filter.id}:rtk-head-tail`);
+    }
+  } else if (head !== undefined && total > head) {
+    lines = [...lines.slice(0, head), `... (${total - head} lines omitted)`];
+    appliedRules.push(`${filter.id}:rtk-head`);
+  } else if (tail !== undefined && total > tail) {
+    lines = [`... (${total - tail} lines omitted)`, ...(tail > 0 ? lines.slice(-tail) : [])];
+    appliedRules.push(`${filter.id}:rtk-tail`);
+  }
+
+  const maxLines = filter.rtkTomlMaxLines;
+  if (maxLines !== undefined && lines.length > maxLines) {
+    const dropped = lines.length - maxLines;
+    lines = [...lines.slice(0, maxLines), `... (${dropped} lines truncated)`];
+    appliedRules.push(`${filter.id}:rtk-max-lines`);
+  }
+  return lines;
+}
+
 function truncateUnicodeSafe(line: string, maxChars: number): string {
   if (maxChars <= 0) return line;
   const chars = Array.from(line);
@@ -70,6 +105,7 @@ export function applyLineFilter(text: string, filter: RtkFilterDefinition): Line
   const appliedRules: string[] = [];
 
   let lines = text.split(/\r?\n/);
+  if (filter.sourceFormat === "rtk-toml-v1" && lines.at(-1) === "") lines.pop();
   const originalLineCount = lines.length;
 
   if (filter.stripAnsi) {
@@ -159,6 +195,18 @@ export function applyLineFilter(text: string, filter: RtkFilterDefinition): Line
     }
   }
 
+  if (filter.sourceFormat === "rtk-toml-v1") {
+    lines = applyRtkTomlLineLimits(lines, filter, appliedRules);
+    const output = lines.join("\n");
+    const finalOutput = output.trim().length === 0 && filter.onEmpty ? filter.onEmpty : output;
+    return {
+      text: finalOutput,
+      strippedLines: Math.max(0, originalLineCount - finalOutput.split(/\r?\n/).length),
+      keptByRule: keepPatterns.length > 0,
+      appliedRules,
+    };
+  }
+
   const truncated = smartTruncate(lines.join("\n"), {
     maxLines: filter.maxLines,
     preserveHead: filter.preserveHead,
diff --git a/open-sse/services/compression/engines/rtk/tomlCompatibility.ts b/open-sse/services/compression/engines/rtk/tomlCompatibility.ts
new file mode 100644
index 00000000000..1c45c045490
--- /dev/null
+++ b/open-sse/services/compression/engines/rtk/tomlCompatibility.ts
@@ -0,0 +1,334 @@
+import crypto from "node:crypto";
+import fs from "node:fs";
+import path from "node:path";
+import os from "node:os";
+import { parse as parseToml } from "smol-toml";
+import { z } from "zod";
+import { isReDoSProne, type RtkFilterDefinition } from "./filterSchema.ts";
+import { applyLineFilter } from "./lineFilter.ts";
+
+const MAX_TOML_BYTES = 1024 * 1024;
+
+const replaceRuleSchema = z
+  .object({
+    pattern: z.string().min(1),
+    replacement: z.string(),
+  })
+  .strict();
+
+const matchOutputRuleSchema = z
+  .object({
+    pattern: z.string().min(1),
+    message: z.string(),
+    unless: z.string().min(1).optional(),
+  })
+  .strict();
+
+const filterSchema = z
+  .object({
+    description: z.string().optional(),
+    match_command: z.string().min(1),
+    strip_ansi: z.boolean().optional(),
+    filter_stderr: z.boolean().optional(),
+    strip_lines_matching: z.array(z.string()).optional(),
+    keep_lines_matching: z.array(z.string()).optional(),
+    replace: z.array(replaceRuleSchema).optional(),
+    match_output: z.array(matchOutputRuleSchema).optional(),
+    truncate_lines_at: z.number().int().min(0).optional(),
+    head_lines: z.number().int().min(0).optional(),
+    tail_lines: z.number().int().min(0).optional(),
+    max_lines: z.number().int().min(0).optional(),
+    on_empty: z.string().optional(),
+  })
+  .strict();
+
+const inlineTestSchema = z
+  .object({
+    name: z.string().min(1),
+    input: z.string(),
+    expected: z.string(),
+  })
+  .strict();
+
+const fileSchema = z
+  .object({
+    schema_version: z.literal(1),
+    filters: z.record(z.string().min(1), filterSchema).default({}),
+    tests: z.record(z.string().min(1), z.array(inlineTestSchema)).default({}),
+  })
+  .strict();
+
+type ParsedFilter = z.infer<typeof filterSchema>;
+
+function arrayOrEmpty<T>(value: T[] | undefined): T[] {
+  return value ?? [];
+}
+
+function numberOrZero(value: number | undefined): number {
+  return value ?? 0;
+}
+
+export interface RtkTomlTestOutcome {
+  filterId: string;
+  testName: string;
+  passed: boolean;
+  actual: string;
+  expected: string;
+}
+
+export interface RtkTomlCompatibilityResult {
+  schemaVersion: 1;
+  sha256: string;
+  filters: RtkFilterDefinition[];
+  outcomes: RtkTomlTestOutcome[];
+  filtersWithoutTests: string[];
+  warnings: string[];
+  passed: boolean;
+}
+
+export class RtkTomlCompatibilityError extends Error {
+  readonly publicMessage: string;
+
+  constructor(message: string) {
+    super("RTK TOML schema v1 compatibility error");
+    this.name = "RtkTomlCompatibilityError";
+    this.publicMessage = message;
+  }
+}
+
+function compatibilityError(message: string): RtkTomlCompatibilityError {
+  return new RtkTomlCompatibilityError(message);
+}
+
+function categoryFor(name: string, commandPattern: string): RtkFilterDefinition["category"] {
+  const value = `${name} ${commandPattern}`.toLowerCase();
+  if (/\b(?:git|gh)\b/.test(value)) return "git";
+  if (/\b(?:test|jest|vitest|pytest|cargo test|go test|rspec|playwright)\b/.test(value)) {
+    return "test";
+  }
+  if (/\b(?:build|tsc|eslint|ruff|clippy|gradle|make|next|vite|webpack)\b/.test(value)) {
+    return "build";
+  }
+  if (/\b(?:docker|kubectl|podman|compose)\b/.test(value)) return "docker";
+  if (/\b(?:npm|pnpm|yarn|bun|pip|poetry|uv|bundle|composer)\b/.test(value)) {
+    return "package";
+  }
+  if (/\b(?:terraform|tofu|ansible|helm|pulumi)\b/.test(value)) return "infra";
+  if (/\b(?:aws|gcloud|az|cloudflare)\b/.test(value)) return "cloud";
+  if (/\b(?:ls|find|grep|rg|df|du|ps|systemctl|ssh|rsync)\b/.test(value)) return "shell";
+  return "generic";
+}
+
+function regexFields(filter: ParsedFilter): Array<{ field: string; pattern: string }> {
+  return [
+    { field: "match_command", pattern: filter.match_command },
+    ...(filter.strip_lines_matching ?? []).map((pattern) => ({
+      field: "strip_lines_matching",
+      pattern,
+    })),
+    ...(filter.keep_lines_matching ?? []).map((pattern) => ({
+      field: "keep_lines_matching",
+      pattern,
+    })),
+    ...(filter.replace ?? []).map(({ pattern }) => ({ field: "replace.pattern", pattern })),
+    ...(filter.match_output ?? []).flatMap(({ pattern, unless }) => [
+      { field: "match_output.pattern", pattern },
+      ...(unless ? [{ field: "match_output.unless", pattern: unless }] : []),
+    ]),
+  ];
+}
+
+function validateRegexes(name: string, filter: ParsedFilter): void {
+  for (const { field, pattern } of regexFields(filter)) {
+    if (isReDoSProne(pattern)) {
+      throw compatibilityError(`filter '${name}' has an unsafe regex in ${field}`);
+    }
+    try {
+      new RegExp(pattern);
+    } catch {
+      throw compatibilityError(`filter '${name}' has an invalid regex in ${field}`);
+    }
+  }
+}
+
+function toDefinition(
+  name: string,
+  filter: ParsedFilter,
+  tests: z.infer<typeof inlineTestSchema>[]
+): RtkFilterDefinition {
+  if (
+    (filter.strip_lines_matching?.length ?? 0) > 0 &&
+    (filter.keep_lines_matching?.length ?? 0) > 0
+  ) {
+    throw compatibilityError(
+      `filter '${name}' cannot combine strip_lines_matching with keep_lines_matching`
+    );
+  }
+  validateRegexes(name, filter);
+  return {
+    id: name,
+    name,
+    description: filter.description ?? "",
+    commandTypes: [],
+    commandPatterns: [filter.match_command],
+    matchPatterns: [],
+    category: categoryFor(name, filter.match_command),
+    priority: 50,
+    stripPatterns: arrayOrEmpty(filter.strip_lines_matching),
+    keepPatterns: arrayOrEmpty(filter.keep_lines_matching),
+    priorityPatterns: [],
+    collapsePatterns: [],
+    stripAnsi: filter.strip_ansi ?? false,
+    replace: arrayOrEmpty(filter.replace),
+    matchOutput: arrayOrEmpty(filter.match_output),
+    truncateLineAt: numberOrZero(filter.truncate_lines_at),
+    onEmpty: filter.on_empty ?? "",
+    filterStderr: false,
+    deduplicate: false,
+    maxLines: numberOrZero(filter.max_lines),
+    preserveHead: 0,
+    preserveTail: 0,
+    rtkTomlHeadLines: filter.head_lines,
+    rtkTomlTailLines: filter.tail_lines,
+    rtkTomlMaxLines: filter.max_lines,
+    sourceFormat: "rtk-toml-v1",
+    tests,
+  };
+}
+
+function comparable(value: string): string {
+  return value.replace(/\n+$/g, "");
+}
+
+function tomlSyntaxLocation(error: unknown): string {
+  if (typeof error !== "object" || error === null) return "";
+  const { line, column } = error as { line?: unknown; column?: unknown };
+  if (!Number.isSafeInteger(line) || !Number.isSafeInteger(column)) return "";
+  return ` (line ${line}, column ${column})`;
+}
+
+export function parseRtkTomlV1(content: string): RtkTomlCompatibilityResult {
+  if (Buffer.byteLength(content, "utf8") > MAX_TOML_BYTES) {
+    throw compatibilityError(`file exceeds the ${MAX_TOML_BYTES}-byte limit`);
+  }
+
+  let raw: unknown;
+  try {
+    raw = parseToml(content);
+  } catch (error) {
+    throw compatibilityError(`invalid TOML syntax${tomlSyntaxLocation(error)}`);
+  }
+
+  const parsed = fileSchema.safeParse(raw);
+  if (!parsed.success) {
+    const issue = parsed.error.issues[0];
+    const field = issue?.path.length ? issue.path.join(".") : "document";
+    throw compatibilityError(`${field}: ${issue?.message ?? "invalid document"}`);
+  }
+  if (Object.keys(parsed.data.filters).length === 0) {
+    throw compatibilityError("document contains no filters");
+  }
+
+  for (const testName of Object.keys(parsed.data.tests)) {
+    if (!(testName in parsed.data.filters)) {
+      throw compatibilityError(`tests reference unknown filter '${testName}'`);
+    }
+  }
+
+  const filters = Object.entries(parsed.data.filters).map(([name, filter]) =>
+    toDefinition(name, filter, parsed.data.tests[name] ?? [])
+  );
+  const outcomes = filters.flatMap((filter) =>
+    filter.tests.map((test) => {
+      const actual = comparable(applyLineFilter(test.input, filter).text);
+      const expected = comparable(test.expected);
+      return {
+        filterId: filter.id,
+        testName: test.name,
+        passed: actual === expected,
+        actual,
+        expected,
+      };
+    })
+  );
+  const filtersWithoutTests = filters
+    .filter((filter) => filter.tests.length === 0)
+    .map((filter) => filter.id);
+  const warnings = filtersWithoutTests.map(
+    (id) => `Filter '${id}' has no inline tests and should be reviewed before installation`
+  );
+  for (const [id, filter] of Object.entries(parsed.data.filters)) {
+    if (filter.filter_stderr) {
+      warnings.push(
+        `Filter '${id}': filter_stderr is accepted as a no-op because OmniRoute receives already-captured tool output`
+      );
+    }
+  }
+
+  return {
+    schemaVersion: 1,
+    sha256: crypto.createHash("sha256").update(content).digest("hex"),
+    filters,
+    outcomes,
+    filtersWithoutTests,
+    warnings,
+    passed: outcomes.every((outcome) => outcome.passed),
+  };
+}
+
+function getDataDir(): string {
+  return process.env.DATA_DIR || path.join(os.homedir(), ".omniroute");
+}
+
+export function getGlobalRtkTomlPath(): string {
+  return path.join(getDataDir(), "rtk", "filters.toml");
+}
+
+export function installGlobalRtkTomlV1(
+  content: string,
+  options: { overwrite?: boolean } = {}
+): RtkTomlCompatibilityResult & { installedPath: string; backupCreated: boolean } {
+  const result = parseRtkTomlV1(content);
+  if (!result.passed) {
+    throw compatibilityError("one or more inline tests failed");
+  }
+
+  const target = getGlobalRtkTomlPath();
+  const directory = path.dirname(target);
+  fs.mkdirSync(directory, { recursive: true, mode: 0o700 });
+  try {
+    fs.chmodSync(directory, 0o700);
+  } catch {
+    // Best effort on filesystems that do not support POSIX permissions.
+  }
+  if (fs.existsSync(target) && !options.overwrite) {
+    throw compatibilityError("filters.toml already exists; confirm overwrite to replace it");
+  }
+
+  let backupCreated = false;
+  if (fs.existsSync(target)) {
+    fs.copyFileSync(target, `${target}.bak`);
+    try {
+      fs.chmodSync(`${target}.bak`, 0o600);
+    } catch {
+      // Best effort on filesystems that do not support POSIX permissions.
+    }
+    backupCreated = true;
+  }
+  const temporary = `${target}.${process.pid}.${crypto.randomUUID()}.tmp`;
+  try {
+    fs.writeFileSync(temporary, content, { encoding: "utf8", mode: 0o600, flag: "wx" });
+    fs.renameSync(temporary, target);
+    try {
+      fs.chmodSync(target, 0o600);
+    } catch {
+      // Best effort on filesystems that do not support POSIX permissions.
+    }
+  } finally {
+    fs.rmSync(temporary, { force: true });
+  }
+
+  return { ...result, installedPath: "rtk/filters.toml", backupCreated };
+}
+
+export const RTK_TOML_MAX_BYTES = MAX_TOML_BYTES;
diff --git a/open-sse/services/compression/engines/session-dedup/fuzzy.ts b/open-sse/services/compression/engines/session-dedup/fuzzy.ts
index f61c2521c72..2f5d37f99b5 100644
--- a/open-sse/services/compression/engines/session-dedup/fuzzy.ts
+++ b/open-sse/services/compression/engines/session-dedup/fuzzy.ts
@@ -1,5 +1,5 @@
 // open-sse/services/compression/engines/session-dedup/fuzzy.ts
-import { storeBlock } from "../ccr/index.ts";
+import { buildCcrMarker, tryStoreBlock } from "../ccr/index.ts";
 
 type MessageLike = { role?: string; content?: unknown; [key: string]: unknown };
 
@@ -111,8 +111,9 @@ export function applyFuzzyPass(messages: MessageLike[], opts: FuzzyPassOptions):
 
     const replacements = new Map<number, string>();
     for (const nd of nearDups) {
-      const hash = storeBlock(nd.block.text, opts.principalId);
-      const marker = `[CCR retrieve hash=${hash} chars=${nd.block.text.length}]`;
+      const stored = tryStoreBlock(nd.block.text, opts.principalId, { source: "session-dedup" });
+      if (!stored.stored) continue;
+      const marker = buildCcrMarker(stored.hash, nd.block.text.length);
       if (marker.length < nd.block.text.length) replacements.set(nd.block.index, marker);
     }
     if (replacements.size === 0) return { messages, fuzzyCount: 0 };
@@ -138,9 +139,7 @@ export function runFuzzyPass(
   principalId?: string
 ): FuzzyPassResult {
   const raw = stepConfig["fuzzy"] as
-    | boolean
-    | { enabled?: boolean; minJaccard?: number; shingleSize?: number }
-    | undefined;
+    boolean | { enabled?: boolean; minJaccard?: number; shingleSize?: number } | undefined;
   const cfg = typeof raw === "boolean" ? { enabled: raw } : raw;
   if (!cfg?.enabled) return { messages, fuzzyCount: 0 };
   return applyFuzzyPass(messages, {
diff --git a/open-sse/services/compression/liveZone.ts b/open-sse/services/compression/liveZone.ts
new file mode 100644
index 00000000000..d0c69cd03a2
--- /dev/null
+++ b/open-sse/services/compression/liveZone.ts
@@ -0,0 +1,397 @@
+import { createHash } from "node:crypto";
+
+import { estimateCompressionTokens } from "./stats.ts";
+import type { CompressionResult, CompressionStats } from "./types.ts";
+
+export interface LiveZoneOptions {
+  principalId?: string;
+  sessionId?: string;
+  variant: unknown;
+  ttlMinutes?: number;
+}
+
+interface LiveZoneEntry {
+  rawItemDigests: string[];
+  rawStableFieldsDigest: string;
+  transformedPrefix: unknown[];
+  transformedStableFields: Record<string, unknown>;
+  stats: CompressionStats | null;
+  lastAccess: number;
+  expiresAt: number;
+  bytes: number;
+}
+
+interface LiveZoneContext {
+  field: "messages" | "input";
+  key: string;
+  rawItems: unknown[];
+  rawItemDigests: string[];
+  rawStableFieldsDigest: string;
+  ttlMs: number;
+  now: number;
+}
+
+const MAX_ENTRIES = 100;
+const MAX_ENTRY_BYTES = 2 * 1024 * 1024;
+const MAX_TOTAL_BYTES = 32 * 1024 * 1024;
+const DEFAULT_TTL_MINUTES = 5;
+const STABLE_PREFIX_FIELDS = [
+  "system",
+  "systemInstruction",
+  "system_instruction",
+  "instructions",
+  "tools",
+  "tool_choice",
+] as const;
+
+const entries = new Map<string, LiveZoneEntry>();
+let totalBytes = 0;
+
+function serialize(value: unknown): string | null {
+  try {
+    const serialized = JSON.stringify(value);
+    return typeof serialized === "string" ? serialized : null;
+  } catch {
+    return null;
+  }
+}
+
+function digest(value: unknown): string | null {
+  const serialized = serialize(value);
+  return serialized === null ? null : createHash("sha256").update(serialized).digest("hex");
+}
+
+function cloneItems(items: unknown[]): unknown[] | null {
+  try {
+    return structuredClone(items);
+  } catch {
+    const serialized = serialize(items);
+    if (serialized === null) return null;
+    try {
+      return JSON.parse(serialized) as unknown[];
+    } catch {
+      return null;
+    }
+  }
+}
+
+function cloneValue<T>(value: T): T | null {
+  try {
+    return structuredClone(value);
+  } catch {
+    const serialized = serialize(value);
+    if (serialized === null) return null;
+    try {
+      return JSON.parse(serialized) as T;
+    } catch {
+      return null;
+    }
+  }
+}
+
+function pickStableFields(body: Record<string, unknown>): Record<string, unknown> | null {
+  const fields: Record<string, unknown> = {};
+  for (const field of STABLE_PREFIX_FIELDS) {
+    if (Object.prototype.hasOwnProperty.call(body, field)) fields[field] = body[field];
+  }
+  return cloneValue(fields);
+}
+
+function sequenceField(body: Record<string, unknown>): "messages" | "input" | null {
+  if (Array.isArray(body.messages)) return "messages";
+  if (Array.isArray(body.input)) return "input";
+  return null;
+}
+
+function isToolOutputItem(value: unknown): boolean {
+  if (!value || typeof value !== "object") return false;
+  const item = value as Record<string, unknown>;
+  return (
+    item.role === "tool" ||
+    item.role === "function" ||
+    item.role === "tool_result" ||
+    item.type === "function_call_output" ||
+    item.type === "computer_call_output" ||
+    item.type === "tool_result"
+  );
+}
+
+function makeKey(options: LiveZoneOptions, field: string): string | null {
+  const principal = options.principalId?.trim();
+  const session = options.sessionId?.trim();
+  const variant = digest(options.variant);
+  if (!principal || !session || !variant) return null;
+  return `${principal}:${session}:${field}:${variant}`;
+}
+
+function deleteEntry(key: string): void {
+  const existing = entries.get(key);
+  if (!existing) return;
+  totalBytes -= existing.bytes;
+  entries.delete(key);
+}
+
+function prune(now: number): void {
+  for (const [key, entry] of entries) {
+    if (now >= entry.expiresAt) deleteEntry(key);
+  }
+  while (entries.size > MAX_ENTRIES || totalBytes > MAX_TOTAL_BYTES) {
+    const oldest = entries.keys().next().value as string | undefined;
+    if (!oldest) break;
+    deleteEntry(oldest);
+  }
+}
+
+function store(
+  key: string,
+  rawItemDigests: string[],
+  rawStableFieldsDigest: string,
+  result: CompressionResult,
+  field: "messages" | "input",
+  now: number,
+  ttlMs: number
+): void {
+  const transformedItems = result.body[field];
+  if (!Array.isArray(transformedItems)) return;
+  const transformedPrefix = cloneItems(transformedItems);
+  const transformedStableFields = pickStableFields(result.body);
+  const stats = cloneValue(result.stats);
+  if (!transformedPrefix || !transformedStableFields) return;
+  const serialized = serialize({ transformedPrefix, transformedStableFields, stats });
+  if (serialized === null) return;
+  const bytes = Buffer.byteLength(serialized, "utf8") + rawItemDigests.length * 64;
+  if (bytes > MAX_ENTRY_BYTES) return;
+
+  deleteEntry(key);
+  entries.set(key, {
+    rawItemDigests,
+    rawStableFieldsDigest,
+    transformedPrefix,
+    transformedStableFields,
+    stats,
+    lastAccess: now,
+    expiresAt: now + ttlMs,
+    bytes,
+  });
+  totalBytes += bytes;
+  prune(now);
+}
+
+function hasExactRawPrefix(rawItemDigests: string[], entry: LiveZoneEntry): boolean {
+  if (rawItemDigests.length < entry.rawItemDigests.length) return false;
+  for (let index = 0; index < entry.rawItemDigests.length; index++) {
+    if (rawItemDigests[index] !== entry.rawItemDigests[index]) return false;
+  }
+  return true;
+}
+
+function restoreStableFields(
+  body: Record<string, unknown>,
+  stableFields: Record<string, unknown>
+): Record<string, unknown> | null {
+  const restored = cloneValue(stableFields);
+  return restored ? { ...body, ...restored } : null;
+}
+
+function withLiveZoneStats(
+  body: Record<string, unknown>,
+  result: CompressionResult,
+  frozenItems: number,
+  liveItems: number
+): CompressionResult {
+  const originalTokens = estimateCompressionTokens(body);
+  const compressedTokens = estimateCompressionTokens(result.body);
+  const savingsPercent =
+    originalTokens > 0
+      ? Math.max(
+          0,
+          Math.round(((originalTokens - compressedTokens) / originalTokens) * 10000) / 100
+        )
+      : 0;
+  const base = result.stats;
+  const stats: CompressionStats = {
+    ...(base ?? {
+      techniquesUsed: [],
+      mode: "stacked",
+      timestamp: Date.now(),
+    }),
+    originalTokens,
+    compressedTokens,
+    savingsPercent,
+    techniquesUsed: [...new Set([...(base?.techniquesUsed ?? []), "live-zone-prefix-reuse"])],
+    liveZone: {
+      cacheHit: true,
+      frozenItems,
+      liveItems,
+    },
+  };
+  return {
+    ...result,
+    compressed: result.compressed || compressedTokens < originalTokens,
+    stats,
+  };
+}
+
+function hasGlobalHardBudget(variant: unknown): boolean {
+  if (!variant || typeof variant !== "object") return false;
+  const config = (variant as Record<string, unknown>).config;
+  if (!config || typeof config !== "object") return false;
+  const record = config as Record<string, unknown>;
+  return record.targetTokens != null || record.targetRatio != null;
+}
+
+function resolveLiveZoneContext(
+  body: Record<string, unknown>,
+  options: LiveZoneOptions
+): LiveZoneContext | null {
+  const field = sequenceField(body);
+  const key = field ? makeKey(options, field) : null;
+  if (!field || !key) return null;
+
+  const rawItems = body[field] as unknown[];
+  const rawItemDigests = rawItems.map(digest);
+  if (rawItemDigests.some((value) => value === null)) return null;
+  const rawStableFieldsDigest = digest(pickStableFields(body));
+  if (!rawStableFieldsDigest) return null;
+  const ttlMinutes = Math.min(60, Math.max(1, options.ttlMinutes ?? DEFAULT_TTL_MINUTES));
+  const now = Date.now();
+  return {
+    field,
+    key,
+    rawItems,
+    rawItemDigests: rawItemDigests as string[],
+    rawStableFieldsDigest,
+    ttlMs: ttlMinutes * 60_000,
+    now,
+  };
+}
+
+async function compressAndStore(
+  body: Record<string, unknown>,
+  context: LiveZoneContext,
+  compress: (body: Record<string, unknown>) => Promise<CompressionResult>
+): Promise<CompressionResult> {
+  const result = await compress(body);
+  store(
+    context.key,
+    context.rawItemDigests,
+    context.rawStableFieldsDigest,
+    result,
+    context.field,
+    context.now,
+    context.ttlMs
+  );
+  return result;
+}
+
+async function compressLiveToolOutputs(
+  body: Record<string, unknown>,
+  field: "messages" | "input",
+  liveItems: unknown[],
+  previousStats: CompressionStats | null,
+  compress: (body: Record<string, unknown>) => Promise<CompressionResult>
+): Promise<{ liveResult: CompressionResult; transformedLive: unknown[] } | null> {
+  const transformedLive = cloneItems(liveItems);
+  if (!transformedLive) return null;
+  const liveToolIndexes = liveItems.flatMap((item, index) =>
+    isToolOutputItem(item) ? [index] : []
+  );
+  if (liveToolIndexes.length === 0) {
+    return { liveResult: { body, compressed: false, stats: previousStats }, transformedLive };
+  }
+
+  const liveToolItems = liveToolIndexes.map((index) => liveItems[index]);
+  const liveResult = await compress({ ...body, [field]: liveToolItems });
+  const transformed = liveResult.body[field];
+  if (!Array.isArray(transformed) || transformed.length !== liveToolItems.length) {
+    return { liveResult: { body, compressed: false, stats: null }, transformedLive };
+  }
+  for (let index = 0; index < liveToolIndexes.length; index++) {
+    transformedLive[liveToolIndexes[index]] = transformed[index];
+  }
+  return { liveResult, transformedLive };
+}
+
+async function reuseLiveZoneEntry(
+  body: Record<string, unknown>,
+  context: LiveZoneContext,
+  previous: LiveZoneEntry,
+  compress: (body: Record<string, unknown>) => Promise<CompressionResult>
+): Promise<CompressionResult> {
+  entries.delete(context.key);
+  previous.lastAccess = context.now;
+  entries.set(context.key, previous);
+
+  const frozenItems = previous.rawItemDigests.length;
+  const liveItems = context.rawItems.slice(frozenItems);
+  const frozenPrefix = cloneItems(previous.transformedPrefix);
+  if (!frozenPrefix) return compress(body);
+  const live = await compressLiveToolOutputs(
+    body,
+    context.field,
+    liveItems,
+    previous.stats,
+    compress
+  );
+  if (!live) return compress(body);
+  const restoredBody = restoreStableFields(live.liveResult.body, previous.transformedStableFields);
+  if (!restoredBody) return compress(body);
+  const combinedBody = {
+    ...restoredBody,
+    [context.field]: [...frozenPrefix, ...live.transformedLive],
+  };
+  const combinedResult = withLiveZoneStats(
+    body,
+    { ...live.liveResult, body: combinedBody },
+    frozenItems,
+    liveItems.length
+  );
+  if (entries.get(context.key) === previous) {
+    store(
+      context.key,
+      context.rawItemDigests,
+      context.rawStableFieldsDigest,
+      combinedResult,
+      context.field,
+      Date.now(),
+      context.ttlMs
+    );
+  }
+  return combinedResult;
+}
+
+/**
+ * Reuses the byte-identical transformed prefix from the previous request in a session and runs
+ * compression only over newly appended messages/input items. Any changed prefix, missing identity,
+ * unsupported body shape, serialization failure, or oversized entry fails open to full compression.
+ */
+export async function applyLiveZoneCompression(
+  body: Record<string, unknown>,
+  options: LiveZoneOptions,
+  compress: (body: Record<string, unknown>) => Promise<CompressionResult>
+): Promise<CompressionResult> {
+  // A global hard budget needs the complete history to make correct keep/drop decisions.
+  if (hasGlobalHardBudget(options.variant)) return compress(body);
+  const context = resolveLiveZoneContext(body, options);
+  if (!context) return compress(body);
+  prune(context.now);
+  const previous = entries.get(context.key);
+
+  if (
+    !previous ||
+    previous.rawStableFieldsDigest !== context.rawStableFieldsDigest ||
+    !hasExactRawPrefix(context.rawItemDigests, previous)
+  ) {
+    return compressAndStore(body, context, compress);
+  }
+  return reuseLiveZoneEntry(body, context, previous, compress);
+}
+
+export function resetLiveZoneCache(): void {
+  entries.clear();
+  totalBytes = 0;
+}
+
+export function getLiveZoneCacheStats(): { entries: number; bytes: number } {
+  return { entries: entries.size, bytes: totalBytes };
+}
diff --git a/open-sse/services/compression/types.ts b/open-sse/services/compression/types.ts
index 9bba7973b27..ff8aedd77b5 100644
--- a/open-sse/services/compression/types.ts
+++ b/open-sse/services/compression/types.ts
@@ -131,6 +131,11 @@ export interface ContextEditingConfig {
   enabled: boolean;
 }
 
+/** Cache-aligned compression: freeze a previously transformed prefix and process only new items. */
+export interface LiveZoneConfig {
+  enabled: boolean;
+}
+
 export interface CompressionPipelineStep {
   engine: CompressionEngineId;
   intensity?: CavemanIntensity | RtkIntensity;
@@ -193,6 +198,8 @@ export interface CompressionConfig {
   ultra?: UltraConfig;
   /** Provider-delegated context editing (Claude/Anthropic only). */
   contextEditing?: ContextEditingConfig;
+  /** Opt-in cache-aligned live-zone compression (default disabled). */
+  liveZone?: LiveZoneConfig;
   /** Per-engine opt-in toggles for the config panel. */
   engines: Record<string, EngineToggle>;
   /** Active combo preset id, or null if none selected. */
@@ -294,6 +301,11 @@ export interface CompressionStats {
   }>;
   /** Present only when QuantumLock stabilized ≥1 fragment this run. */
   quantumLock?: QuantumLockStats;
+  liveZone?: {
+    cacheHit: boolean;
+    frozenItems: number;
+    liveItems: number;
+  };
 }
 
 export interface CompressionResult {
@@ -321,6 +333,7 @@ export const DEFAULT_COMPRESSION_CONFIG: CompressionConfig = {
   activeComboId: null,
   ultraEngine: "heuristic",
   ultraSlmPrewarm: false,
+  liveZone: { enabled: false },
 };
 
 export const DEFAULT_CAVEMAN_CONFIG: CavemanConfig = {
diff --git a/open-sse/services/responsesInputSanitizer.ts b/open-sse/services/responsesInputSanitizer.ts
index c1f938a38b3..3546a294628 100644
--- a/open-sse/services/responsesInputSanitizer.ts
+++ b/open-sse/services/responsesInputSanitizer.ts
@@ -92,6 +92,29 @@ function sanitizeMessageContent(record: JsonRecord): JsonRecord {
   return { ...record, content };
 }
 
+function sanitizeNestedOutputPart(part: unknown): unknown {
+  const record = toRecord(part);
+  if (!record) return part;
+
+  // `output` on replayed items is an input-side container. Its content uses
+  // input content-part types even when the enclosing item originated from an
+  // assistant/tool response. Converting an image placeholder to output_text
+  // here makes Codex reject the request with the inverse 400.
+  if (record.type === "output_text" || record.type === "refusal") {
+    const next: JsonRecord = { ...record, type: "input_text" };
+    if (typeof next.text !== "string") {
+      next.text = typeof record.refusal === "string" ? record.refusal : "";
+    }
+    delete next.annotations;
+    delete next.logprobs;
+    delete next.obfuscation;
+    delete next.refusal;
+    return next;
+  }
+
+  return sanitizeContentPart(part, "user");
+}
+
 function sanitizeOutputContent(record: JsonRecord): JsonRecord {
   if (!Array.isArray(record.output)) return record;
 
@@ -99,8 +122,7 @@ function sanitizeOutputContent(record: JsonRecord): JsonRecord {
   // Responses input. In that shape OpenAI validates `input[n].output[m].type`
   // against output content part types, so legacy Chat-style `image_url` parts
   // must be normalized here too, not only in message.content.
-  const role = record.type === "function_call_output" ? "user" : "assistant";
-  const output = record.output.map((part) => sanitizeContentPart(part, role));
+  const output = record.output.map(sanitizeNestedOutputPart);
   return { ...record, output };
 }
 
diff --git a/open-sse/translator/request/claude-to-gemini.ts b/open-sse/translator/request/claude-to-gemini.ts
index f7af11ad302..b3d987055de 100644
--- a/open-sse/translator/request/claude-to-gemini.ts
+++ b/open-sse/translator/request/claude-to-gemini.ts
@@ -188,7 +188,9 @@ export function claudeToGeminiRequest(model, body, stream, credentials = null) {
   // Priority: thinking.budget_tokens (Claude native) > output_config.effort (Claude Code).
   if (model.startsWith("gemma-4")) {
     // gemma-4 models returns - 400: Thinking budget is not supported for this model
-  } else if (body.thinking?.type === "enabled" && body.thinking.budget_tokens) {
+  } else if (body.thinking?.type === "enabled" && body.thinking.budget_tokens !== undefined) {
+    // #6813: a truthy check here dropped `budget_tokens: 0` (dynamic thinking).
+    // `undefined` (no budget specified) still falls through to the effort branch.
     result.generationConfig.thinkingConfig = {
       thinkingBudget: body.thinking.budget_tokens,
       includeThoughts: true,
diff --git a/open-sse/translator/request/openai-responses/toResponses.ts b/open-sse/translator/request/openai-responses/toResponses.ts
index bda9113d08d..48eaea8533b 100644
--- a/open-sse/translator/request/openai-responses/toResponses.ts
+++ b/open-sse/translator/request/openai-responses/toResponses.ts
@@ -49,6 +49,32 @@ function mapChatResponseFormatToResponsesText(body: JsonRecord, result: JsonReco
   result.text = { ...existingText, format };
 }
 
+// Convert a Chat-Completions content block (string or text-part array) into the
+// Responses API `input_text` part array used by message input items.
+function buildResponsesTextParts(content: unknown): unknown[] {
+  if (typeof content === "string") {
+    return [{ type: "input_text", text: content }];
+  }
+  if (Array.isArray(content)) {
+    const parts: unknown[] = [];
+    for (const partValue of content) {
+      // A bare string inside the content array is a real text instruction
+      // (e.g. a harness-injected system reminder), not a structured part.
+      // Silently dropping it lost the instruction (#6954 follow-up).
+      if (typeof partValue === "string") {
+        parts.push({ type: "input_text", text: partValue });
+        continue;
+      }
+      const part = toRecord(partValue);
+      if (part.type === "text" || typeof part.text === "string") {
+        parts.push({ type: "input_text", text: toString(part.text) });
+      }
+    }
+    return parts.length > 0 ? parts : [{ type: "input_text", text: "" }];
+  }
+  return [{ type: "input_text", text: "" }];
+}
+
 export function openaiToOpenAIResponsesRequest(
   model: unknown,
   body: unknown,
@@ -83,7 +109,18 @@ export function openaiToOpenAIResponsesRequest(
       if (!hasSystemMessage) {
         result.instructions = typeof msg.content === "string" ? msg.content : "";
         hasSystemMessage = true;
+        continue;
       }
+      // Mid-conversation system/developer turns (e.g. harness-injected reminders
+      // from Claude Code) must survive as developer-role input items. The
+      // Responses API supports the `developer` role for exactly this; mapping
+      // them to `assistant` misattributes harness instructions as model output,
+      // and silently dropping them loses them entirely (#6954).
+      input.push({
+        type: "message",
+        role: "developer",
+        content: buildResponsesTextParts(msg.content),
+      });
       continue;
     }
 
diff --git a/open-sse/utils/diagnostics.ts b/open-sse/utils/diagnostics.ts
index 176d0f7843f..39161e41373 100644
--- a/open-sse/utils/diagnostics.ts
+++ b/open-sse/utils/diagnostics.ts
@@ -284,5 +284,27 @@ export function detectMalformedNonStream(resp: unknown): MalformedReason | null
   return null;
 }
 
+export function describeMalformedNonStream(
+  resp: unknown,
+  reason: MalformedReason
+): { message: string; code: string; type: string } {
+  const body = resp && typeof resp === "object" ? (resp as Record<string, unknown>) : null;
+  if (body?.object === "response" && body.status === "failed") {
+    return {
+      message: "upstream reported a failed response without usable output",
+      code: "upstream_response_failed",
+      type: "upstream_response_error",
+    };
+  }
+  return {
+    message:
+      reason === "no_terminal"
+        ? "upstream response did not reach a terminal state"
+        : "upstream returned an empty response without usable output",
+    code: "upstream_empty_response",
+    type: "upstream_response_error",
+  };
+}
+
 // ── Test-only export ─────────────────────────────────────────────────────────
 export const __test = { describeReason };
diff --git a/open-sse/utils/sseHeartbeat.ts b/open-sse/utils/sseHeartbeat.ts
index 9a122141457..005a19b1728 100644
--- a/open-sse/utils/sseHeartbeat.ts
+++ b/open-sse/utils/sseHeartbeat.ts
@@ -61,6 +61,20 @@ type SseHeartbeatTransformOptions = {
 
 const HEARTBEAT_ENCODER = new TextEncoder();
 
+/**
+ * Whether OmniRoute may emit SSE `:` comment lines (e.g. the `: keepalive` heartbeat).
+ * Some strict OpenAI-compatible clients parse every SSE line as JSON and crash on `:` comments.
+ * Set OMNIROUTE_SSE_COMMENTS=off to suppress comment-shaped heartbeats (they become a no-op).
+ * Defaults to enabled for backward compatibility.
+ */
+export function sseCommentsEnabled(): boolean {
+  // SSR/edge safety: `process` is not defined in Workers/Deno/edge runtimes.
+  if (typeof process === "undefined") return true;
+  const v = process.env.OMNIROUTE_SSE_COMMENTS;
+  if (v === undefined || v === "") return true;
+  return v.trim().toLowerCase() !== "off";
+}
+
 export function createSseHeartbeatTransform({
   intervalMs = DEFAULT_SSE_HEARTBEAT_INTERVAL_MS,
   signal,
@@ -72,6 +86,13 @@ export function createSseHeartbeatTransform({
     return new TransformStream<Uint8Array, Uint8Array>();
   }
 
+  // Opt-out for strict OpenAI-compatible clients that JSON.parse every SSE line and
+  // crash on `:` comment heartbeats. OMNIROUTE_SSE_COMMENTS=off disables comment-shaped
+  // heartbeats (they become a no-op); valid `data:` heartbeats are unaffected.
+  if (!sseCommentsEnabled() && shape === HEARTBEAT_SHAPES.COMMENT) {
+    return new TransformStream<Uint8Array, Uint8Array>();
+  }
+
   let intervalId: ReturnType<typeof setInterval> | undefined;
 
   const stop = () => {
diff --git a/open-sse/utils/stream.ts b/open-sse/utils/stream.ts
index 974d532e9eb..e8356db21bb 100644
--- a/open-sse/utils/stream.ts
+++ b/open-sse/utils/stream.ts
@@ -640,6 +640,24 @@ export function createSSEStream(options: StreamOptions = {}) {
     dropResponsesCommentary,
   } = options;
   const signatureNamespace = connectionId;
+  // Request-body-size metric (for monitoring payload size distribution & correlation with TTFT).
+  // The size is JSON-serialised byte count; stored as a performance mark detail so monitoring
+  // tools can query performance.getEntriesByType("mark") filtered by name.
+  let bodySize = 0;
+  try {
+    bodySize = body ? Buffer.byteLength(JSON.stringify(body), "utf8") : 0;
+  } catch {
+    /* body may not be JSON-serialisable (e.g. FormData, Blob) — metric stays 0 */
+  }
+  if (bodySize > 0) {
+    // Cleared immediately: this is a fixed-name mark created on every stream, so
+    // leaving it in the global performance timeline would accumulate without bound
+    // over a long-running server's lifetime. A wired PerformanceObserver still
+    // receives the entry (delivery is queued independently of the buffer) even though
+    // clearMarks() removes it from getEntriesByName()/getEntriesByType() right after.
+    performance.mark("omni-request-body-size", { detail: bodySize });
+    performance.clearMarks("omni-request-body-size");
+  }
 
   // Drop internal commentary-phase Responses output before forwarding (#6199).
   // Explicit option wins; otherwise read the feature flag (default on). Resolved
diff --git a/package-lock.json b/package-lock.json
index a678e41868f..d5eb7bd8fd2 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -74,6 +74,7 @@
         "recharts": "^3.8.1",
         "safe-regex": "^2.1.1",
         "selfsigned": "^5.5.0",
+        "smol-toml": "1.6.1",
         "socks": "^2.8.7",
         "sql.js": "^1.14.1",
         "sqlite-vec": "^0.1.9",
@@ -103,7 +104,7 @@
         "@testing-library/jest-dom": "^6.9.1",
         "@testing-library/react": "^16.3.2",
         "@types/better-sqlite3": "^7.6.13",
-        "@types/bun": "latest",
+        "@types/bun": "*",
         "@types/node": "^26.1.0",
         "@types/react": "^19.2.15",
         "@types/react-dom": "^19.2.3",
@@ -144,7 +145,7 @@
         "wtfnode": "^0.10.1"
       },
       "engines": {
-        "node": ">=22.0.0 <23 || >=24.0.0 <27"
+        "node": ">=22.22.2 <23 || >=24.0.0 <27"
       },
       "optionalDependencies": {
         "@atjsh/llmlingua-2": "2.0.3",
@@ -213,17 +214,6 @@
         "zod": "^3.25.76 || ^4.1.8"
       }
     },
-    "node_modules/@ai-zen/node-fetch-event-source": {
-      "version": "2.1.4",
-      "resolved": "https://registry.npmjs.org/@ai-zen/node-fetch-event-source/-/node-fetch-event-source-2.1.4.tgz",
-      "integrity": "sha512-OHFwPJecr+qwlyX5CGmTvKAKPZAdZaxvx/XDqS1lx4I2ZAk9riU0XnEaRGOOAEFrdcLZ98O5yWqubwjaQc0umg==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "cross-fetch": "^4.0.0"
-      }
-    },
     "node_modules/@alcalzone/ansi-tokenize": {
       "version": "0.3.0",
       "resolved": "https://registry.npmjs.org/@alcalzone/ansi-tokenize/-/ansi-tokenize-0.3.0.tgz",
@@ -295,9 +285,9 @@
       }
     },
     "node_modules/@anthropic-ai/claude-agent-sdk": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk/-/claude-agent-sdk-0.3.195.tgz",
-      "integrity": "sha512-FVmXu9pvOMbuBKWrF8YsYQdQ/upOpv5rS8lFAnFO5jbyXT/2hN7kEPd2vd2GJpaMvNcO/KptyQUK5AxjjTz3+w==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk/-/claude-agent-sdk-0.3.201.tgz",
+      "integrity": "sha512-InT1XLmf2QpldWdtznKDWEoGJT4p+sXh24yxbeBQ++lMJCzMrI0W27MEmmmDWx0otpa+ubdHCF5YQ6oiNt7cmg==",
       "dev": true,
       "license": "SEE LICENSE IN README.md",
       "optional": true,
@@ -305,14 +295,14 @@
         "node": ">=18.0.0"
       },
       "optionalDependencies": {
-        "@anthropic-ai/claude-agent-sdk-darwin-arm64": "0.3.195",
-        "@anthropic-ai/claude-agent-sdk-darwin-x64": "0.3.195",
-        "@anthropic-ai/claude-agent-sdk-linux-arm64": "0.3.195",
-        "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": "0.3.195",
-        "@anthropic-ai/claude-agent-sdk-linux-x64": "0.3.195",
-        "@anthropic-ai/claude-agent-sdk-linux-x64-musl": "0.3.195",
-        "@anthropic-ai/claude-agent-sdk-win32-arm64": "0.3.195",
-        "@anthropic-ai/claude-agent-sdk-win32-x64": "0.3.195"
+        "@anthropic-ai/claude-agent-sdk-darwin-arm64": "0.3.201",
+        "@anthropic-ai/claude-agent-sdk-darwin-x64": "0.3.201",
+        "@anthropic-ai/claude-agent-sdk-linux-arm64": "0.3.201",
+        "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": "0.3.201",
+        "@anthropic-ai/claude-agent-sdk-linux-x64": "0.3.201",
+        "@anthropic-ai/claude-agent-sdk-linux-x64-musl": "0.3.201",
+        "@anthropic-ai/claude-agent-sdk-win32-arm64": "0.3.201",
+        "@anthropic-ai/claude-agent-sdk-win32-x64": "0.3.201"
       },
       "peerDependencies": {
         "@anthropic-ai/sdk": ">=0.93.0",
@@ -321,9 +311,9 @@
       }
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-darwin-arm64": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-darwin-arm64/-/claude-agent-sdk-darwin-arm64-0.3.195.tgz",
-      "integrity": "sha512-WIMM/8HRCLsTDHFTIwQvvE8WCA/oaMJtdQxsP7iNyfzIGwXbuOyU95V8vYIhZfaO2yaSpbBRncunq4CtR5H4ng==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-darwin-arm64/-/claude-agent-sdk-darwin-arm64-0.3.201.tgz",
+      "integrity": "sha512-8Mcb3BDyKUGfJWFFTWwt+at37lbDH3ZwVtUNPWGG1toZ75RDCJry5U4kXRvQ2xokvJQlA0E+eNp6keWe5ZH22Q==",
       "cpu": [
         "arm64"
       ],
@@ -335,9 +325,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-darwin-x64": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-darwin-x64/-/claude-agent-sdk-darwin-x64-0.3.195.tgz",
-      "integrity": "sha512-RY7DB+4LXosE0MJ+XELmakfPrDN1YX4lkk9CTDm28jGCVcESRz9kAEqbyaiC48dZcmN9V1NCLutzINGdcr1TBg==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-darwin-x64/-/claude-agent-sdk-darwin-x64-0.3.201.tgz",
+      "integrity": "sha512-TFR2bu0+ml3RHoMrtsgD0qDK5Oknw8kYGBV7qpQHn+IWmE96gnHhogG1LpJwpHtni08XkJIjfWk1DdlsUYtRkQ==",
       "cpu": [
         "x64"
       ],
@@ -349,9 +339,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-linux-arm64": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-arm64/-/claude-agent-sdk-linux-arm64-0.3.195.tgz",
-      "integrity": "sha512-JuIq5Fnz/F1snl0aqi1gcuRZqPWoPNrL9dJ0DuievCxKkO8hnEz/Mmn5Zos7x1X8HE//ZnEvmQXoEQEZXonJew==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-arm64/-/claude-agent-sdk-linux-arm64-0.3.201.tgz",
+      "integrity": "sha512-mShTo3MwF0gkN4dDw78wWJiB6aBDVRkl81cnApvoBofpdyUBYgm9Gw16CCjDTgelMKeBFqN6ErJpwjI3wbP00A==",
       "cpu": [
         "arm64"
       ],
@@ -366,9 +356,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-linux-arm64-musl": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-arm64-musl/-/claude-agent-sdk-linux-arm64-musl-0.3.195.tgz",
-      "integrity": "sha512-ZmyBA/AFzhgutcxb7dbhCm6GTjJytwNYXTxJoKE2B3A409WCYccjMqeji6vCMNxyyfylglGo5D8dVMIxW9aoug==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-arm64-musl/-/claude-agent-sdk-linux-arm64-musl-0.3.201.tgz",
+      "integrity": "sha512-EiqbpfJIpChfkn+8Uj061Qjyw0eaRcOXtdrvVuHANyj8ZErVOr8HlH6op9PSeIUa9TX0m2+tNgKPQvOGseQckA==",
       "cpu": [
         "arm64"
       ],
@@ -383,9 +373,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-linux-x64": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-x64/-/claude-agent-sdk-linux-x64-0.3.195.tgz",
-      "integrity": "sha512-s1lNi1cL93luoqsItH+fNO4KpIhdkvnVhWGGQUQ/8ftwa2gfmcIQnOg1hG8Ks+KzeD3UUQ8L9YEVHVADnFI/9A==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-x64/-/claude-agent-sdk-linux-x64-0.3.201.tgz",
+      "integrity": "sha512-jrJBrRWrSuoFKIgjyqxHqmfd6Pb3Bs5Bvakg0knXCTC4fbUXGnC9Q6u7gdDwgXohUNP6/DD+s8U7bivvvVv0dg==",
       "cpu": [
         "x64"
       ],
@@ -400,9 +390,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-linux-x64-musl": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-x64-musl/-/claude-agent-sdk-linux-x64-musl-0.3.195.tgz",
-      "integrity": "sha512-nf8Q/LauB+ZOC6QDjxNhbsvwUtYjKYnaWJLTYFwhkmsLujePnety1AtT/1ubaUoq5AM1j297DhMlYTasa79OUA==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-x64-musl/-/claude-agent-sdk-linux-x64-musl-0.3.201.tgz",
+      "integrity": "sha512-IbxnzO5UCbqbm2TnzCHkSyJorAFw2isdKdIsFCTxJJjSs3ZC+v3LC1QSUiVCx0qi+CV6w3MKx6mLI11mrvhbbQ==",
       "cpu": [
         "x64"
       ],
@@ -417,9 +407,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-win32-arm64": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-win32-arm64/-/claude-agent-sdk-win32-arm64-0.3.195.tgz",
-      "integrity": "sha512-hbkDE+xPIZzRWm+D+BKrH9uJH6USIZdDIlsyrIlGi3JFHoieYoA1vdUNyldSS9+F3ZqQtfPjr2Qy08IVB6akYA==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-win32-arm64/-/claude-agent-sdk-win32-arm64-0.3.201.tgz",
+      "integrity": "sha512-UsoytRJ/037uHpb3ATrIoe+AgwTf+PwKuFLGjddHAV/11wERJs0hlrnSmcnp43kf0PFxoSNinngme96YYASmQg==",
       "cpu": [
         "arm64"
       ],
@@ -431,9 +421,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-win32-x64": {
-      "version": "0.3.195",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-win32-x64/-/claude-agent-sdk-win32-x64-0.3.195.tgz",
-      "integrity": "sha512-av0piEB3X1Dzhpr8A+DqHVZ9y8s1jpn8enzwX0TKKUPBn5IqLTWC7wD6v66aoUgu4f+g4ThZirmDZA6shyPEZQ==",
+      "version": "0.3.201",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-win32-x64/-/claude-agent-sdk-win32-x64-0.3.201.tgz",
+      "integrity": "sha512-PhalN/0cWcqDfbx7iwoLNR2gurjTiqhBk1G6K+NRScxEcQjWuu5xKXCcdbX8ePVpT+nbEMmFEFpn2y+8V8hIdA==",
       "cpu": [
         "x64"
       ],
@@ -445,9 +435,9 @@
       ]
     },
     "node_modules/@anthropic-ai/sdk": {
-      "version": "0.106.0",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/sdk/-/sdk-0.106.0.tgz",
-      "integrity": "sha512-ufwVvYNDBj2dzOGupBCTaNzBLxqcTnGOzI4z8Wouxlt+mT3J3HuOmatgCy1VmwCHOUueqZ41ERhm0O99OUcbWA==",
+      "version": "0.110.0",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/sdk/-/sdk-0.110.0.tgz",
+      "integrity": "sha512-hOP4bNYXDFHDxxiEgzlILXrxZIYCDnhe8sry0RDRKD/QnsEpvZcQpablCdm9X/WuD/YgOiSIkkqsL1mLLlTqJw==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
@@ -611,22 +601,22 @@
       }
     },
     "node_modules/@aws-sdk/client-bedrock-runtime": {
-      "version": "3.1081.0",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/client-bedrock-runtime/-/client-bedrock-runtime-3.1081.0.tgz",
-      "integrity": "sha512-rRAGXY5qV/NCYbVA0QZHPierv3diOOiE4+1f5vedpbyvg7Phh9m/I2pRFwu0koUtMdRSGPWgoNcvMIyaDhDj7Q==",
+      "version": "3.1088.0",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/client-bedrock-runtime/-/client-bedrock-runtime-3.1088.0.tgz",
+      "integrity": "sha512-m76gdG4tYCStbh1MjWeO8w40la/Sgs+0kpO2iWaWF09vGpZ9iwwcsR//k5g7OxIfuOWdiwP4TvgoTNaEAxy0fg==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.974.29",
-        "@aws-sdk/credential-provider-node": "^3.972.64",
-        "@aws-sdk/eventstream-handler-node": "^3.972.25",
-        "@aws-sdk/middleware-eventstream": "^3.972.21",
-        "@aws-sdk/middleware-websocket": "^3.972.37",
-        "@aws-sdk/token-providers": "3.1081.0",
-        "@aws-sdk/types": "^3.973.15",
-        "@smithy/core": "^3.29.0",
-        "@smithy/fetch-http-handler": "^5.6.2",
-        "@smithy/node-http-handler": "^4.9.2",
-        "@smithy/types": "^4.15.1",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/credential-provider-node": "^3.972.69",
+        "@aws-sdk/eventstream-handler-node": "^3.972.28",
+        "@aws-sdk/middleware-eventstream": "^3.972.24",
+        "@aws-sdk/middleware-websocket": "^3.972.41",
+        "@aws-sdk/token-providers": "3.1088.0",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/fetch-http-handler": "^5.6.6",
+        "@smithy/node-http-handler": "^4.9.6",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -679,17 +669,17 @@
       }
     },
     "node_modules/@aws-sdk/core": {
-      "version": "3.975.1",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/core/-/core-3.975.1.tgz",
-      "integrity": "sha512-8qh/6EYb7hl/ZwVfQufhbMEZs1gQIc7GbdrIf4eprQJ7cv042+74nE6l3YDfyWNzb9iPXb8fRyYSHkNIk5eE6Q==",
+      "version": "3.975.3",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/core/-/core-3.975.3.tgz",
+      "integrity": "sha512-7ur3kCKuvPLqlsZ2XlvnNBVQ7KkpSu6Y6dOTwSPHLrFpTEfZM8isLBJc4cgv96WB7GifeVM436mpycwxBd2vEA==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/types": "^3.974.0",
-        "@aws-sdk/xml-builder": "^3.972.34",
+        "@aws-sdk/types": "^3.974.2",
+        "@aws-sdk/xml-builder": "^3.972.36",
         "@aws/lambda-invoke-store": "^0.3.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/signature-v4": "^5.6.3",
-        "@smithy/types": "^4.16.0",
+        "@smithy/core": "^3.29.4",
+        "@smithy/signature-v4": "^5.6.5",
+        "@smithy/types": "^4.16.1",
         "bowser": "^2.11.0",
         "tslib": "^2.6.2"
       },
@@ -698,15 +688,15 @@
       }
     },
     "node_modules/@aws-sdk/credential-provider-env": {
-      "version": "3.972.57",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-env/-/credential-provider-env-3.972.57.tgz",
-      "integrity": "sha512-1RfJaF7SW1TOnvNGU7kaYjwUf5H3sfm+synGH1bHhRlqcnxCt3szebH3dmKEyY4tuGcbQ6ffzUT89cRitBV8OQ==",
+      "version": "3.972.59",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-env/-/credential-provider-env-3.972.59.tgz",
+      "integrity": "sha512-Ny5e4Mfh3QPmiAc0AiUe+cbTXDlxkU3Rc+EpWOfyWeWEy6yp7Fa1KmfNeCc+1a8by9zQ9gtohmiQUkMPScF3ng==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -714,17 +704,17 @@
       }
     },
     "node_modules/@aws-sdk/credential-provider-http": {
-      "version": "3.972.59",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-http/-/credential-provider-http-3.972.59.tgz",
-      "integrity": "sha512-sRCkpTiFnCdQvuaRVjQ6SVoHu6i7RUpurVo1c4F81HWhPvUJ7Wdp5MNtSdX1O29CNXc8em3O5m52hCjVtAD9SA==",
+      "version": "3.972.61",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-http/-/credential-provider-http-3.972.61.tgz",
+      "integrity": "sha512-8jAjgStl5Ytq4+HF3X/9f+EmRinaRbGRRtQGktlPfBRVx73H+R1y48vIeXerQtYGFaUqkEp3fT6jP854rVO2yQ==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/fetch-http-handler": "^5.6.4",
-        "@smithy/node-http-handler": "^4.9.4",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/fetch-http-handler": "^5.6.6",
+        "@smithy/node-http-handler": "^4.9.6",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -732,23 +722,23 @@
       }
     },
     "node_modules/@aws-sdk/credential-provider-ini": {
-      "version": "3.973.1",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-ini/-/credential-provider-ini-3.973.1.tgz",
-      "integrity": "sha512-6d8H6ZAh3ZPKZ6fe1nG2OWeZEZPtt9ravoD1dezPdPtsSkJRoxGAnFSHwKT3E/Te6fHE30zRzjV6TD12rvF6yQ==",
+      "version": "3.973.3",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-ini/-/credential-provider-ini-3.973.3.tgz",
+      "integrity": "sha512-WpuqYX4gGkx++fCTSWE8+41JzkZVcrI50SH48Ml4CsG1pyuHKyMmpw/FixBHDrmjoQ553PmeCLa/fZIcst+WyA==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/credential-provider-env": "^3.972.57",
-        "@aws-sdk/credential-provider-http": "^3.972.59",
-        "@aws-sdk/credential-provider-login": "^3.972.63",
-        "@aws-sdk/credential-provider-process": "^3.972.57",
-        "@aws-sdk/credential-provider-sso": "^3.973.1",
-        "@aws-sdk/credential-provider-web-identity": "^3.972.63",
-        "@aws-sdk/nested-clients": "^3.997.31",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/credential-provider-imds": "^4.4.7",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/credential-provider-env": "^3.972.59",
+        "@aws-sdk/credential-provider-http": "^3.972.61",
+        "@aws-sdk/credential-provider-login": "^3.972.65",
+        "@aws-sdk/credential-provider-process": "^3.972.59",
+        "@aws-sdk/credential-provider-sso": "^3.973.3",
+        "@aws-sdk/credential-provider-web-identity": "^3.972.65",
+        "@aws-sdk/nested-clients": "^3.997.33",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/credential-provider-imds": "^4.4.9",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -756,16 +746,16 @@
       }
     },
     "node_modules/@aws-sdk/credential-provider-login": {
-      "version": "3.972.63",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-login/-/credential-provider-login-3.972.63.tgz",
-      "integrity": "sha512-GREWRrMj0XnNKMaVa/Mauoaui26qBEHu71WWqXbwZOu/jFQOnPZjTf7u0KtGKC8VGa6VUs9kDWGgocrKNLS9vw==",
+      "version": "3.972.65",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-login/-/credential-provider-login-3.972.65.tgz",
+      "integrity": "sha512-xr9rgjYEdmC2Tpg2lwt9o+nOEaK9Qpd+dBjzrVCuWWyQfvhO91Ezu0Hh9ts2VUxOZxmS/k5T9msa34e4R1bnrQ==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/nested-clients": "^3.997.31",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/nested-clients": "^3.997.33",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -773,21 +763,21 @@
       }
     },
     "node_modules/@aws-sdk/credential-provider-node": {
-      "version": "3.972.67",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-node/-/credential-provider-node-3.972.67.tgz",
-      "integrity": "sha512-oYlzWst56rlhhjbYnexwv5hVLYe1cW4liLObhDfxDLI4RAQzleMVHQgQgx7XsC4HKj4e3kjT8v9DId+Pi/dndw==",
+      "version": "3.972.69",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-node/-/credential-provider-node-3.972.69.tgz",
+      "integrity": "sha512-wbJGGesd0Tl18bmUcbj1xJ+e7CpuRJ6PIpMywLFuUttGy615lua87cJ0EA8pFpY/QgPuUXbnupWBtSPJ9tyZhg==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/credential-provider-env": "^3.972.57",
-        "@aws-sdk/credential-provider-http": "^3.972.59",
-        "@aws-sdk/credential-provider-ini": "^3.973.1",
-        "@aws-sdk/credential-provider-process": "^3.972.57",
-        "@aws-sdk/credential-provider-sso": "^3.973.1",
-        "@aws-sdk/credential-provider-web-identity": "^3.972.63",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/credential-provider-imds": "^4.4.7",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/credential-provider-env": "^3.972.59",
+        "@aws-sdk/credential-provider-http": "^3.972.61",
+        "@aws-sdk/credential-provider-ini": "^3.973.3",
+        "@aws-sdk/credential-provider-process": "^3.972.59",
+        "@aws-sdk/credential-provider-sso": "^3.973.3",
+        "@aws-sdk/credential-provider-web-identity": "^3.972.65",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/credential-provider-imds": "^4.4.9",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -795,15 +785,15 @@
       }
     },
     "node_modules/@aws-sdk/credential-provider-process": {
-      "version": "3.972.57",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-process/-/credential-provider-process-3.972.57.tgz",
-      "integrity": "sha512-TiVQhuU0pbhIZAUZacbPHMyzrIdiH+lnx+PMY/Pu/b93dJrq3wdZwzUJ0TPpvNxaqbHsxJvQZW3/h/beLiKq7Q==",
+      "version": "3.972.59",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-process/-/credential-provider-process-3.972.59.tgz",
+      "integrity": "sha512-DlZF2/MhLlatDdlrIy3CUCpfdbLrKx+3SMjVo+WyHnPpwzkc/M3vwAHw4OVJf7DMvO+4vfRqSCMc/E9I1auN0g==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -811,34 +801,17 @@
       }
     },
     "node_modules/@aws-sdk/credential-provider-sso": {
-      "version": "3.973.1",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-sso/-/credential-provider-sso-3.973.1.tgz",
-      "integrity": "sha512-3foTZUJ4821Ij60X7K3NJroygiZLnbBmarN+T//O2cjkISan90zElN3NBmgSlDrTQ7Gs6z/yO8V7h60QNcDZHQ==",
+      "version": "3.973.3",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-sso/-/credential-provider-sso-3.973.3.tgz",
+      "integrity": "sha512-hmdDHoy2G5Es2e8IgelNMYUuSQI6uCIAKZMJ2u2PdKDhxvbk1uWD/g4+R7R5c/tJfKEB1+KjjWiaoCr/S+ZTiQ==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/nested-clients": "^3.997.31",
-        "@aws-sdk/token-providers": "3.1083.0",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/types": "^4.16.0",
-        "tslib": "^2.6.2"
-      },
-      "engines": {
-        "node": ">=20.0.0"
-      }
-    },
-    "node_modules/@aws-sdk/credential-provider-sso/node_modules/@aws-sdk/token-providers": {
-      "version": "3.1083.0",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/token-providers/-/token-providers-3.1083.0.tgz",
-      "integrity": "sha512-s0woKnxuHrExLc5L2ArIH5BMkbonHPtt+5hSBM8oknp9M6QTuUmmAmJ2E0EdzCGONrO+8+ADPqvv6UX0nNcc7A==",
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/nested-clients": "^3.997.31",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/nested-clients": "^3.997.33",
+        "@aws-sdk/token-providers": "3.1088.0",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -846,16 +819,16 @@
       }
     },
     "node_modules/@aws-sdk/credential-provider-web-identity": {
-      "version": "3.972.63",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-web-identity/-/credential-provider-web-identity-3.972.63.tgz",
-      "integrity": "sha512-8qZLFhM69eKcS37m459ctPR05Qimycm/74OPVioe6wNZabMT54GYhwBju0+J656RkMasNSawWQu+c8CmBe3TUQ==",
+      "version": "3.972.65",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/credential-provider-web-identity/-/credential-provider-web-identity-3.972.65.tgz",
+      "integrity": "sha512-gHQb/Kt0chjk/JQDa/GJDqmAvEuVn8n7z10wK2h0LFM9TUDRkohgOO4aEF+s2sBLM0br7Cl5W6P7phgjrrJvLQ==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/nested-clients": "^3.997.31",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/nested-clients": "^3.997.33",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -863,14 +836,14 @@
       }
     },
     "node_modules/@aws-sdk/eventstream-handler-node": {
-      "version": "3.972.25",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/eventstream-handler-node/-/eventstream-handler-node-3.972.25.tgz",
-      "integrity": "sha512-df7HN1ozwMrB9+59re9PM7tSLxLAcheMWc5u/KyfCPCAWtN/vP7y7RTUZOy48uT1K9MESisVeOPPzF3O1AW01A==",
+      "version": "3.972.28",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/eventstream-handler-node/-/eventstream-handler-node-3.972.28.tgz",
+      "integrity": "sha512-XV5sEH1xH5oydNgvUH87CR8BA2SBZXltckteS17D7ZT2k2THIBiExO9TEBYcgqUU31WAIfBH2hmx+fhFMiTL4Q==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/types": "^3.973.15",
-        "@smithy/core": "^3.29.0",
-        "@smithy/types": "^4.15.1",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -878,14 +851,14 @@
       }
     },
     "node_modules/@aws-sdk/middleware-eventstream": {
-      "version": "3.972.21",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/middleware-eventstream/-/middleware-eventstream-3.972.21.tgz",
-      "integrity": "sha512-HvLgDnxBLaHi9E5K++6Vuk+1+qqn7Pmn8zrlzd+NXH3jBzwujnuzZtAR9WHPkbUGPO92FkoQWj/M1IsdxTlBmQ==",
+      "version": "3.972.24",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/middleware-eventstream/-/middleware-eventstream-3.972.24.tgz",
+      "integrity": "sha512-oykin4mDWxNOuYQ7SF1cHzgYeuFEkF4cdRwgvjFFbIklkx09qIFBiOgsORafG9sXZFO3TayMmQuAQYgADXhI8w==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/types": "^3.973.15",
-        "@smithy/core": "^3.29.0",
-        "@smithy/types": "^4.15.1",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -912,17 +885,17 @@
       }
     },
     "node_modules/@aws-sdk/middleware-websocket": {
-      "version": "3.972.37",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/middleware-websocket/-/middleware-websocket-3.972.37.tgz",
-      "integrity": "sha512-u4J2KwTe6hr0hBrcKF7vPNxoQoPdSwdhE8mEQK/ffaY/XgYQ77NRqsbxeNQDaRthQJ3D3KhOdCjrio1E+/ocng==",
+      "version": "3.972.41",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/middleware-websocket/-/middleware-websocket-3.972.41.tgz",
+      "integrity": "sha512-LSbGvvYmjc4Br9BPYI2dTLnIclmrSiQbahkP4D6nRGVEv4qsCZ8csVuKBPVEEFCVD+EEngGh8ROls6XpumtwMg==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.974.29",
-        "@aws-sdk/types": "^3.973.15",
-        "@smithy/core": "^3.29.0",
-        "@smithy/fetch-http-handler": "^5.6.2",
-        "@smithy/signature-v4": "^5.6.1",
-        "@smithy/types": "^4.15.1",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/fetch-http-handler": "^5.6.6",
+        "@smithy/signature-v4": "^5.6.5",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -930,18 +903,18 @@
       }
     },
     "node_modules/@aws-sdk/nested-clients": {
-      "version": "3.997.31",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/nested-clients/-/nested-clients-3.997.31.tgz",
-      "integrity": "sha512-BDHTpwcsZHEBNEJzOg/B1BkFYJxAXY50dau/NyVWs3d51F0WgIUGSWZot/Os+N3KpDhXeaXnz37mWffAvduREw==",
+      "version": "3.997.33",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/nested-clients/-/nested-clients-3.997.33.tgz",
+      "integrity": "sha512-dVZOroI/r3/ENvqNGgjMPul+jjlz9GddfVusgTXlVjfZj5isibOxecLkGQbRPp8XOuX+RAfjXLFgPkD1JS5xrw==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.975.1",
-        "@aws-sdk/signature-v4-multi-region": "^3.996.39",
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/core": "^3.29.2",
-        "@smithy/fetch-http-handler": "^5.6.4",
-        "@smithy/node-http-handler": "^4.9.4",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/signature-v4-multi-region": "^3.996.41",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/fetch-http-handler": "^5.6.6",
+        "@smithy/node-http-handler": "^4.9.6",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -949,14 +922,14 @@
       }
     },
     "node_modules/@aws-sdk/signature-v4-multi-region": {
-      "version": "3.996.39",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/signature-v4-multi-region/-/signature-v4-multi-region-3.996.39.tgz",
-      "integrity": "sha512-8+srXqYIF8KYMLC4FxMLEM5Ek7kUNibJu1R4m8/fUhhNYIZZz26oGtKkCr8I/HiG2fFQxBvaGgQZT4/mqRCSnA==",
+      "version": "3.996.41",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/signature-v4-multi-region/-/signature-v4-multi-region-3.996.41.tgz",
+      "integrity": "sha512-QMUytg+FQMGouc8gHS00KoYih3+N6cqmVI/pQGOIo7Nr7OpQaiXjSYOuL+vsPZ1tymY4LAQ8MYcHJmws5LRxng==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/types": "^3.974.0",
-        "@smithy/signature-v4": "^5.6.3",
-        "@smithy/types": "^4.16.0",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/signature-v4": "^5.6.5",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -964,16 +937,16 @@
       }
     },
     "node_modules/@aws-sdk/token-providers": {
-      "version": "3.1081.0",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/token-providers/-/token-providers-3.1081.0.tgz",
-      "integrity": "sha512-kduAeI6cL+zqwj3gjPh9LhuX7kBZ83msYxutavaR+UPm5K8J7iThJBvNRAsFNyWTji92CSU8dogUgvi9T0BehA==",
+      "version": "3.1088.0",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/token-providers/-/token-providers-3.1088.0.tgz",
+      "integrity": "sha512-4ObatWt2qpJg5FBk4LOOKrTQYzaqeewAtdO3r9ZO8lH9YqLtpTzLyIdy0mJ+nVdfYOnqISkKNfmzP22bNDhwyw==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@aws-sdk/core": "^3.974.29",
-        "@aws-sdk/nested-clients": "^3.997.29",
-        "@aws-sdk/types": "^3.973.15",
-        "@smithy/core": "^3.29.0",
-        "@smithy/types": "^4.15.1",
+        "@aws-sdk/core": "^3.975.3",
+        "@aws-sdk/nested-clients": "^3.997.33",
+        "@aws-sdk/types": "^3.974.2",
+        "@smithy/core": "^3.29.4",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -981,12 +954,12 @@
       }
     },
     "node_modules/@aws-sdk/types": {
-      "version": "3.974.0",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/types/-/types-3.974.0.tgz",
-      "integrity": "sha512-QIBrw90CDm4O0UaIIzkU6DrFdeJzEb2Va5EPEVpyldj6sHJxB6cshhStJuhZxk3wR3PmjJlYsjPmY1kNb+KGBg==",
+      "version": "3.974.2",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/types/-/types-3.974.2.tgz",
+      "integrity": "sha512-3W6IUtSxFbH6X7Wb7DzGCV5QiFQsd0g8bOfntpmDxQlzBoKWUMBu/JPQR0DwkE+Hpnxd6db1tXbOwdeHddG6cA==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@smithy/types": "^4.16.0",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -994,12 +967,12 @@
       }
     },
     "node_modules/@aws-sdk/xml-builder": {
-      "version": "3.972.34",
-      "resolved": "https://registry.npmjs.org/@aws-sdk/xml-builder/-/xml-builder-3.972.34.tgz",
-      "integrity": "sha512-wHhWL1y7sN3enBA8POrPpQM5jCcmu2ozyhbRei4c8OjVcEaEs6yLucLa/pla457ggS/ysuy7bosagz3HaJkZXA==",
+      "version": "3.972.36",
+      "resolved": "https://registry.npmjs.org/@aws-sdk/xml-builder/-/xml-builder-3.972.36.tgz",
+      "integrity": "sha512-RdGmS1GLrtaTOLE1ElSluMldNrpk9Emq6uYs8SS8iHlu5xTAmM9rRkM91o48+rIRryBtyO9t+uLYCoMG6jVMVA==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@smithy/types": "^4.16.0",
+        "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
       "engines": {
@@ -3071,9 +3044,9 @@
       }
     },
     "node_modules/@eslint/eslintrc": {
-      "version": "3.3.5",
-      "resolved": "https://registry.npmjs.org/@eslint/eslintrc/-/eslintrc-3.3.5.tgz",
-      "integrity": "sha512-4IlJx0X0qftVsN5E+/vGujTRIFtwuLbNsVUe7TO6zYPDR1O6nFwvwhIKEKSrl6dZchmYBITazxKoUYOjdtjlRg==",
+      "version": "3.3.6",
+      "resolved": "https://registry.npmjs.org/@eslint/eslintrc/-/eslintrc-3.3.6.tgz",
+      "integrity": "sha512-l2Ul9PrHsPCKcEY/ac7VgFj9D80C7S68sOKc618SyHDPK36s1XcFebXY0iTzUVn4Yq+YbwvSnDmCz9yxjX+QrA==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
@@ -3083,7 +3056,7 @@
         "globals": "^14.0.0",
         "ignore": "^5.2.0",
         "import-fresh": "^3.2.1",
-        "js-yaml": "^4.1.1",
+        "js-yaml": "^4.3.0",
         "minimatch": "^3.1.5",
         "strip-json-comments": "^3.1.1"
       },
@@ -3112,9 +3085,9 @@
       }
     },
     "node_modules/@eslint/eslintrc/node_modules/js-yaml": {
-      "version": "4.2.0",
-      "resolved": "https://registry.npmjs.org/js-yaml/-/js-yaml-4.2.0.tgz",
-      "integrity": "sha512-ePWsvanv0DWuDRsW8dnt+R4jQ31SCRCQ7hhNcPXZPsoBZiemuZNYGf7adZdqX2D86j6rvKp3RpCxVTSb8WQlOw==",
+      "version": "4.3.0",
+      "resolved": "https://registry.npmjs.org/js-yaml/-/js-yaml-4.3.0.tgz",
+      "integrity": "sha512-1td788aAnnZ5qs7V2QIRl1owjtYpbKt749Y3xauqQgwIIGF/xXWz1wMTEBx5O3LK3lXLVuqXPdPxj2BoFHaW9Q==",
       "dev": true,
       "funding": [
         {
@@ -3142,9 +3115,9 @@
       "license": "MIT"
     },
     "node_modules/@eslint/js": {
-      "version": "9.39.4",
-      "resolved": "https://registry.npmjs.org/@eslint/js/-/js-9.39.4.tgz",
-      "integrity": "sha512-nE7DEIchvtiFTwBw4Lfbu59PG+kCofhjsKaCWzxTpt4lfRjRMqG6uMBzKXuEcyXhOHoUp9riAm7/aWYGhXZ9cw==",
+      "version": "9.39.5",
+      "resolved": "https://registry.npmjs.org/@eslint/js/-/js-9.39.5.tgz",
+      "integrity": "sha512-QywQuszQh77pIXCsq998c8hbhSTI/azTty1Z6N53dmAudKHhy573j3yvRLsX2BSp8YpLtoCEG8E9DJe+8zUh4A==",
       "dev": true,
       "license": "MIT",
       "engines": {
@@ -3268,18 +3241,18 @@
       "license": "MIT"
     },
     "node_modules/@formatjs/icu-messageformat-parser": {
-      "version": "3.5.12",
-      "resolved": "https://registry.npmjs.org/@formatjs/icu-messageformat-parser/-/icu-messageformat-parser-3.5.12.tgz",
-      "integrity": "sha512-YyzzxVgYJ8DELmmkhn0Yr0rUj0dTJFf9Jp628K3S0ysInBWxLVDOS8i3RP91cCp4DMK4WYb4cVMhWA9i4knSJg==",
+      "version": "3.5.14",
+      "resolved": "https://registry.npmjs.org/@formatjs/icu-messageformat-parser/-/icu-messageformat-parser-3.5.14.tgz",
+      "integrity": "sha512-jDvgtoLqe3U6yzoBlToMTkWBe38qSi7LN7kFlnXzd5ig8nn+4tSlED0xtEtdYakZVZGJJY2rW1D5xS3BFZh6kA==",
       "license": "MIT",
       "dependencies": {
-        "@formatjs/icu-skeleton-parser": "2.1.10"
+        "@formatjs/icu-skeleton-parser": "2.1.11"
       }
     },
     "node_modules/@formatjs/icu-skeleton-parser": {
-      "version": "2.1.10",
-      "resolved": "https://registry.npmjs.org/@formatjs/icu-skeleton-parser/-/icu-skeleton-parser-2.1.10.tgz",
-      "integrity": "sha512-XuSva+8ZGawk8VnD5VD6UeH8KarQ/Z022zgjHDoHmlNiAewstXuuzXc0Hk5pGFSdG+nNw5bfJKXqj1ZXHn9yUA==",
+      "version": "2.1.11",
+      "resolved": "https://registry.npmjs.org/@formatjs/icu-skeleton-parser/-/icu-skeleton-parser-2.1.11.tgz",
+      "integrity": "sha512-j8cUmOJzVgkHuS0QiQ6ga76UIoLOFSAMWhs7aZJztH3aAdCOAE6vpC8KVvFB4cU10ON0y2/5oOVmPJ43s2lTwA==",
       "license": "MIT"
     },
     "node_modules/@formatjs/intl-localematcher": {
@@ -3308,9 +3281,9 @@
       }
     },
     "node_modules/@fumadocs/tailwind": {
-      "version": "0.1.0",
-      "resolved": "https://registry.npmjs.org/@fumadocs/tailwind/-/tailwind-0.1.0.tgz",
-      "integrity": "sha512-nF/DCAwOR21HZ4AkjIOv3Iqwyqywzb6pdyeMcoa+aZzirXj5ntvNZbe3jJ0v3ehhtrRfYYeXBezvjn8ZmV+fuQ==",
+      "version": "0.1.1",
+      "resolved": "https://registry.npmjs.org/@fumadocs/tailwind/-/tailwind-0.1.1.tgz",
+      "integrity": "sha512-BnPe52UxSaG8yKlHMKBxXw8h6GpK5qO55ci6+Qd5JnquTvIw6SpfbC1P+qAi82PuPWv1KZAWY8bxRk4+x9ctXw==",
       "license": "MIT",
       "peerDependencies": {
         "tailwindcss": "^4.0.0"
@@ -3575,30 +3548,6 @@
         "node": ">=20.0.0"
       }
     },
-    "node_modules/@ibm-generative-ai/node-sdk": {
-      "version": "3.2.4",
-      "resolved": "https://registry.npmjs.org/@ibm-generative-ai/node-sdk/-/node-sdk-3.2.4.tgz",
-      "integrity": "sha512-HvJSYql3lOPYZcGb23mBw0kcWLlCX+n7EDRgJQxz7gIzx9WafUuDyl1IlTCXGfxolm0EhNIub79u9v7owtks0w==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "@ai-zen/node-fetch-event-source": "^2.1.2",
-        "fetch-retry": "^5.0.6",
-        "http-status-codes": "^2.3.0",
-        "openapi-fetch": "^0.8.2",
-        "p-queue-compat": "1.0.225",
-        "yaml": "^2.3.3"
-      },
-      "peerDependencies": {
-        "@langchain/core": ">=0.1.0"
-      },
-      "peerDependenciesMeta": {
-        "@langchain/core": {
-          "optional": true
-        }
-      }
-    },
     "node_modules/@iconify/types": {
       "version": "2.0.0",
       "resolved": "https://registry.npmjs.org/@iconify/types/-/types-2.0.0.tgz",
@@ -4994,16 +4943,16 @@
       ]
     },
     "node_modules/@lobehub/icons": {
-      "version": "5.10.1",
-      "resolved": "https://registry.npmjs.org/@lobehub/icons/-/icons-5.10.1.tgz",
-      "integrity": "sha512-KMaE+YqPAXuA8gcmzBFefLa9KgCqmJy9Mg3tlGedrL2coAzCQeps+aqivjejHNMnCDTPnGb+OHvX1um2kT1lQw==",
+      "version": "5.13.0",
+      "resolved": "https://registry.npmjs.org/@lobehub/icons/-/icons-5.13.0.tgz",
+      "integrity": "sha512-iXQF8GFvlwNJMR+PaU3jCgVAn5B8F7P48Fm6aodSXP+b+HJiR266rvlMSYvCULRAB/6/rtS1WZH3npc3p3viFw==",
       "license": "MIT",
       "workspaces": [
         "packages/*"
       ],
       "dependencies": {
         "antd-style": "^4.1.0",
-        "es-toolkit": "^1.45.1",
+        "es-toolkit": "^1.49.0",
         "lucide-react": "^0.469.0",
         "polished": "^4.3.1"
       },
@@ -6514,9 +6463,9 @@
       }
     },
     "node_modules/@openai/codex": {
-      "version": "0.142.5",
-      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.142.5.tgz",
-      "integrity": "sha512-WQEpD7l3k68eIAP0aq28EdR18ENBAf8DyprzFhzNwCOQJSv4nHzpwT8Fl30IJacprko2ZCmUBZjM2u941l2yLw==",
+      "version": "0.144.4",
+      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.144.4.tgz",
+      "integrity": "sha512-DTHzYatlKq9dw55E0/HsbK4tRCEKabuJ10ybbqpsG8gVv/kvwEdg3Z4OI3cvLXKa21xkIa4lkGlZoO/HmqmFFw==",
       "dev": true,
       "license": "Apache-2.0",
       "optional": true,
@@ -6527,19 +6476,19 @@
         "node": ">=16"
       },
       "optionalDependencies": {
-        "@openai/codex-darwin-arm64": "npm:@openai/codex@0.142.5-darwin-arm64",
-        "@openai/codex-darwin-x64": "npm:@openai/codex@0.142.5-darwin-x64",
-        "@openai/codex-linux-arm64": "npm:@openai/codex@0.142.5-linux-arm64",
-        "@openai/codex-linux-x64": "npm:@openai/codex@0.142.5-linux-x64",
-        "@openai/codex-win32-arm64": "npm:@openai/codex@0.142.5-win32-arm64",
-        "@openai/codex-win32-x64": "npm:@openai/codex@0.142.5-win32-x64"
+        "@openai/codex-darwin-arm64": "npm:@openai/codex@0.144.4-darwin-arm64",
+        "@openai/codex-darwin-x64": "npm:@openai/codex@0.144.4-darwin-x64",
+        "@openai/codex-linux-arm64": "npm:@openai/codex@0.144.4-linux-arm64",
+        "@openai/codex-linux-x64": "npm:@openai/codex@0.144.4-linux-x64",
+        "@openai/codex-win32-arm64": "npm:@openai/codex@0.144.4-win32-arm64",
+        "@openai/codex-win32-x64": "npm:@openai/codex@0.144.4-win32-x64"
       }
     },
     "node_modules/@openai/codex-darwin-arm64": {
       "name": "@openai/codex",
-      "version": "0.142.5-darwin-arm64",
-      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.142.5-darwin-arm64.tgz",
-      "integrity": "sha512-l43p8xv+Z/2/b6fCUc7/FmcQZsaPB7RFizLponGwHAnFOWe3i9Vky69p+up3BUam9AetoQQUv7Mo+2KdaFEqhA==",
+      "version": "0.144.4-darwin-arm64",
+      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.144.4-darwin-arm64.tgz",
+      "integrity": "sha512-6J3g498cM2oA7vYIJhpuGJlnIi/M5JdYmjB5BZ1Of5HQ0ziIlplFSvH801oVy9J5TQFp642ODzOu/ZEokDUXsg==",
       "cpu": [
         "arm64"
       ],
@@ -6555,9 +6504,9 @@
     },
     "node_modules/@openai/codex-darwin-x64": {
       "name": "@openai/codex",
-      "version": "0.142.5-darwin-x64",
-      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.142.5-darwin-x64.tgz",
-      "integrity": "sha512-yk6A06/VmW7NFsa48OVPaj//g/zeSpd79wjuqfXZwW8ZKRYQm3+wCd3hWjPl79F3QnXvDvM2j3JMIBL3m3GXXg==",
+      "version": "0.144.4-darwin-x64",
+      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.144.4-darwin-x64.tgz",
+      "integrity": "sha512-k1HC8gdbAy+VmMbekYkhM+r+QE2Xfgd67n1VSp94tjz7aXVKoalHcDkdKNM/uUQ8o2tvbiwhHSUftJF8Sm9/Lw==",
       "cpu": [
         "x64"
       ],
@@ -6573,9 +6522,9 @@
     },
     "node_modules/@openai/codex-linux-arm64": {
       "name": "@openai/codex",
-      "version": "0.142.5-linux-arm64",
-      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.142.5-linux-arm64.tgz",
-      "integrity": "sha512-77ka5PSnm5HdxdBT99IwntCasmbqevlS0eiC0AtEb6ZXCLkim2gm0AWm+jNYy0EhbssvNK+KghayWo34HMgXeA==",
+      "version": "0.144.4-linux-arm64",
+      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.144.4-linux-arm64.tgz",
+      "integrity": "sha512-OlKx65579OwIzech9Tt3OUH9+hFZfFrCBP1hL2MudnMIoNr1+cFZjB5YIj5MWMRoBD+K5W3wdBIpQSH855b5Sg==",
       "cpu": [
         "arm64"
       ],
@@ -6591,9 +6540,9 @@
     },
     "node_modules/@openai/codex-linux-x64": {
       "name": "@openai/codex",
-      "version": "0.142.5-linux-x64",
-      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.142.5-linux-x64.tgz",
-      "integrity": "sha512-pxY+d3NgNE57Y/MApD3/TZUAygxJN6I9h3ZeDUwe67mxWjUxsuapxMRFTKSznCalYbRAeZp752+AAXmUbmguEg==",
+      "version": "0.144.4-linux-x64",
+      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.144.4-linux-x64.tgz",
+      "integrity": "sha512-2jxrmV6+/7eBNdg5uhhmOEPFu2o28eYY/ClLzWhSBHH8uo3f2KA1z9JQcVtwlbToW03nEPlEzYNYfCF1UBqsVQ==",
       "cpu": [
         "x64"
       ],
@@ -6608,14 +6557,14 @@
       }
     },
     "node_modules/@openai/codex-sdk": {
-      "version": "0.142.5",
-      "resolved": "https://registry.npmjs.org/@openai/codex-sdk/-/codex-sdk-0.142.5.tgz",
-      "integrity": "sha512-MConZ+eoBoZmkc4reezuzOgLtoI1BQBzo/nVYsSjtAIBpwKcgeEm1rfmqfUnTfFaBNHFTxBntcS7ZeQYuDPbWA==",
+      "version": "0.144.4",
+      "resolved": "https://registry.npmjs.org/@openai/codex-sdk/-/codex-sdk-0.144.4.tgz",
+      "integrity": "sha512-JtND4npLaM1jKlTJVGU3XaX7xqnRG62yusz13kPgialnfd0CBrjOgXFGSRojYcvWZSggQoEkZBVAGtoKE0y8sw==",
       "dev": true,
       "license": "Apache-2.0",
       "optional": true,
       "dependencies": {
-        "@openai/codex": "0.142.5"
+        "@openai/codex": "0.144.4"
       },
       "engines": {
         "node": ">=18"
@@ -6623,9 +6572,9 @@
     },
     "node_modules/@openai/codex-win32-arm64": {
       "name": "@openai/codex",
-      "version": "0.142.5-win32-arm64",
-      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.142.5-win32-arm64.tgz",
-      "integrity": "sha512-65BEqGbUZ7r0ayunIHdBjo5crwgbwKX/6puOcO+VCswUw/dXvDsN2IGcbXB52+bS9U5+FxP783cUHfTT6m40DQ==",
+      "version": "0.144.4-win32-arm64",
+      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.144.4-win32-arm64.tgz",
+      "integrity": "sha512-CCgfI1smFhHZTIpTuBwDJwBr/AR40RTqaFxbBWVabu0RMeYDteRuPiDfdTlktf3C43Y1q10VZXhVGYtCokDg2w==",
       "cpu": [
         "arm64"
       ],
@@ -6641,9 +6590,9 @@
     },
     "node_modules/@openai/codex-win32-x64": {
       "name": "@openai/codex",
-      "version": "0.142.5-win32-x64",
-      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.142.5-win32-x64.tgz",
-      "integrity": "sha512-a+wI4PEx9a2fg6V5ueTTDkOkr1XpEvA5RFXIbo/L2hOfzMmGtyRnbG24bCGu5Q2RSgVxSQV0aLkdb3vdYMNH9A==",
+      "version": "0.144.4-win32-x64",
+      "resolved": "https://registry.npmjs.org/@openai/codex/-/codex-0.144.4-win32-x64.tgz",
+      "integrity": "sha512-iL1ky0ERgdQJOKzom/Ms1fhpwkSmpsA9eVrzAqURFlYGS8z7JqwEgm33+nLGCsY7y25d8Xs/LJ91Oiqz3yXcUg==",
       "cpu": [
         "x64"
       ],
@@ -6679,9 +6628,9 @@
       }
     },
     "node_modules/@opentelemetry/api-logs": {
-      "version": "0.219.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/api-logs/-/api-logs-0.219.0.tgz",
-      "integrity": "sha512-FFx7YnaYJlIjqWW/AG/yAZ0L/NEY724PipXXXQLdtZPbLwBGbUMTGL1i/esI56TWfTUXxhLfpgrnWJCG8aUJyg==",
+      "version": "0.220.0",
+      "resolved": "https://registry.npmjs.org/@opentelemetry/api-logs/-/api-logs-0.220.0.tgz",
+      "integrity": "sha512-CmVa4ImJ+ynfrPMNaAXHET6Bhb44SwzmfyVJFq9ni2jgXJR/l7C6gfVFddNmHP+ZOkP9cf4f9DBe68qVLTHc9w==",
       "dev": true,
       "license": "Apache-2.0",
       "dependencies": {
@@ -6705,9 +6654,9 @@
       }
     },
     "node_modules/@opentelemetry/core": {
-      "version": "2.8.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/core/-/core-2.8.0.tgz",
-      "integrity": "sha512-hd1Lfh8p545nNz+jq1Ejfz+Mn1hyLuxYn1YzTfFNrxr8urEWMNQLPf1Th8kjOH+HxwawCrtgBp8JpBUR4ZSgww==",
+      "version": "2.9.0",
+      "resolved": "https://registry.npmjs.org/@opentelemetry/core/-/core-2.9.0.tgz",
+      "integrity": "sha512-m2nckMT80NnmjTYSPjJQObBJ+8dgkoajEOUbznL8AHZ3T3yHRk2P7gI1PhEBc1+lOnrYE9UWrWHqJDsmqjmNbw==",
       "dev": true,
       "license": "Apache-2.0",
       "dependencies": {
@@ -6721,17 +6670,17 @@
       }
     },
     "node_modules/@opentelemetry/exporter-trace-otlp-http": {
-      "version": "0.219.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/exporter-trace-otlp-http/-/exporter-trace-otlp-http-0.219.0.tgz",
-      "integrity": "sha512-9t6SvBXXBEjOBcIzgozvBbd3jWrv3Gt3ngGhl1fhdZ/zRc7oZDVOFEqbi2zlBpW9BXhgDMKv422J0DL/3iQWfw==",
+      "version": "0.220.0",
+      "resolved": "https://registry.npmjs.org/@opentelemetry/exporter-trace-otlp-http/-/exporter-trace-otlp-http-0.220.0.tgz",
+      "integrity": "sha512-/+ExB3lRkf+erv4PnoywyL7RHKITidxtUpUTS55k7OQ0dB42S7gEF1gry7swb9MSm1hYLUhJg4QQh9W8SpwwqA==",
       "dev": true,
       "license": "Apache-2.0",
       "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/otlp-exporter-base": "0.219.0",
-        "@opentelemetry/otlp-transformer": "0.219.0",
-        "@opentelemetry/resources": "2.8.0",
-        "@opentelemetry/sdk-trace-base": "2.8.0"
+        "@opentelemetry/core": "2.9.0",
+        "@opentelemetry/otlp-exporter-base": "0.220.0",
+        "@opentelemetry/otlp-transformer": "0.220.0",
+        "@opentelemetry/resources": "2.9.0",
+        "@opentelemetry/sdk-trace": "2.9.0"
       },
       "engines": {
         "node": "^18.19.0 || >=20.6.0"
@@ -6740,50 +6689,15 @@
         "@opentelemetry/api": "^1.3.0"
       }
     },
-    "node_modules/@opentelemetry/exporter-trace-otlp-http/node_modules/@opentelemetry/resources": {
-      "version": "2.8.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/resources/-/resources-2.8.0.tgz",
-      "integrity": "sha512-qmXQ27ilDbUK/vGMqwL8D4/rhn76C+sherM4wTbjlfknR8Nvfc/hCxjRJPhkzZzUsPiNg16SA31NxMabwttRjg==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.3.0 <1.10.0"
-      }
-    },
-    "node_modules/@opentelemetry/exporter-trace-otlp-http/node_modules/@opentelemetry/sdk-trace-base": {
-      "version": "2.8.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-trace-base/-/sdk-trace-base-2.8.0.tgz",
-      "integrity": "sha512-mhU4jp+vW0mGbFRd+GeXHvmfA4aDqWjBjLC3pE5XMpLs0IE2ryYb019Ts2AQrOq67gaTF25D91+fgvEHDZEnuQ==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/resources": "2.8.0",
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.3.0 <1.10.0"
-      }
-    },
     "node_modules/@opentelemetry/otlp-exporter-base": {
-      "version": "0.219.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/otlp-exporter-base/-/otlp-exporter-base-0.219.0.tgz",
-      "integrity": "sha512-zvIxQX/AZUVKDU+hCuYx+7UkiP7GRdnk1ZbFQRYzHvYp47cAWR4j3IhoPhV9KaeXEv2xdGq3IA6PnpzDmLcmSA==",
+      "version": "0.220.0",
+      "resolved": "https://registry.npmjs.org/@opentelemetry/otlp-exporter-base/-/otlp-exporter-base-0.220.0.tgz",
+      "integrity": "sha512-CXYo8UD5Mn9YbgebO2EL4wejtA+gxLmLiu6HCk2KH2BR7XhFN6/6p1UlCb23DYCjeYkndevLHuejCCN1yx4+OQ==",
       "dev": true,
       "license": "Apache-2.0",
       "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/otlp-transformer": "0.219.0"
+        "@opentelemetry/core": "2.9.0",
+        "@opentelemetry/otlp-transformer": "0.220.0"
       },
       "engines": {
         "node": "^18.19.0 || >=20.6.0"
@@ -6793,18 +6707,18 @@
       }
     },
     "node_modules/@opentelemetry/otlp-transformer": {
-      "version": "0.219.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/otlp-transformer/-/otlp-transformer-0.219.0.tgz",
-      "integrity": "sha512-aaYKAyXhw9VchKZVGOopD3Gw/kPsyrX2c6IQ0AW32mTjqmZOh5Y6Gf5OYqTNqVktAeBjmFinhyFaCwW6GYK9YQ==",
+      "version": "0.220.0",
+      "resolved": "https://registry.npmjs.org/@opentelemetry/otlp-transformer/-/otlp-transformer-0.220.0.tgz",
+      "integrity": "sha512-lXGrv7KXZ0gNH9SVNUaa6vv6phVYGvJxfXAlMbzbakiXru75f5MZl8Z7oqiMMQD77riVHJCFlQvbZs/VVN2/4A==",
       "dev": true,
       "license": "Apache-2.0",
       "dependencies": {
-        "@opentelemetry/api-logs": "0.219.0",
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/resources": "2.8.0",
-        "@opentelemetry/sdk-logs": "0.219.0",
-        "@opentelemetry/sdk-metrics": "2.8.0",
-        "@opentelemetry/sdk-trace-base": "2.8.0"
+        "@opentelemetry/api-logs": "0.220.0",
+        "@opentelemetry/core": "2.9.0",
+        "@opentelemetry/resources": "2.9.0",
+        "@opentelemetry/sdk-logs": "0.220.0",
+        "@opentelemetry/sdk-metrics": "2.9.0",
+        "@opentelemetry/sdk-trace": "2.9.0"
       },
       "engines": {
         "node": "^18.19.0 || >=20.6.0"
@@ -6813,41 +6727,6 @@
         "@opentelemetry/api": "^1.3.0"
       }
     },
-    "node_modules/@opentelemetry/otlp-transformer/node_modules/@opentelemetry/resources": {
-      "version": "2.8.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/resources/-/resources-2.8.0.tgz",
-      "integrity": "sha512-qmXQ27ilDbUK/vGMqwL8D4/rhn76C+sherM4wTbjlfknR8Nvfc/hCxjRJPhkzZzUsPiNg16SA31NxMabwttRjg==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.3.0 <1.10.0"
-      }
-    },
-    "node_modules/@opentelemetry/otlp-transformer/node_modules/@opentelemetry/sdk-trace-base": {
-      "version": "2.8.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-trace-base/-/sdk-trace-base-2.8.0.tgz",
-      "integrity": "sha512-mhU4jp+vW0mGbFRd+GeXHvmfA4aDqWjBjLC3pE5XMpLs0IE2ryYb019Ts2AQrOq67gaTF25D91+fgvEHDZEnuQ==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/resources": "2.8.0",
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.3.0 <1.10.0"
-      }
-    },
     "node_modules/@opentelemetry/resources": {
       "version": "2.9.0",
       "resolved": "https://registry.npmjs.org/@opentelemetry/resources/-/resources-2.9.0.tgz",
@@ -6865,32 +6744,16 @@
         "@opentelemetry/api": ">=1.3.0 <1.10.0"
       }
     },
-    "node_modules/@opentelemetry/resources/node_modules/@opentelemetry/core": {
-      "version": "2.9.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/core/-/core-2.9.0.tgz",
-      "integrity": "sha512-m2nckMT80NnmjTYSPjJQObBJ+8dgkoajEOUbznL8AHZ3T3yHRk2P7gI1PhEBc1+lOnrYE9UWrWHqJDsmqjmNbw==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.0.0 <1.10.0"
-      }
-    },
     "node_modules/@opentelemetry/sdk-logs": {
-      "version": "0.219.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-logs/-/sdk-logs-0.219.0.tgz",
-      "integrity": "sha512-s6lTKRakaPClvKoWHRChxnXjDMkM/TQ30ff78jN6EBGf7MI7VzANE5PU3f4z9qDUudWjvZjOLHG0rBnBKYvoXA==",
+      "version": "0.220.0",
+      "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-logs/-/sdk-logs-0.220.0.tgz",
+      "integrity": "sha512-WywcTkQtv2iNmt+6y5Kcd4rzvx9bLVsBa2Nwcmg01IUaBTkTow3W4d9KE5vNBpEDtb9tp21WcRBY/lANRrApYA==",
       "dev": true,
       "license": "Apache-2.0",
       "dependencies": {
-        "@opentelemetry/api-logs": "0.219.0",
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/resources": "2.8.0",
+        "@opentelemetry/api-logs": "0.220.0",
+        "@opentelemetry/core": "2.9.0",
+        "@opentelemetry/resources": "2.9.0",
         "@opentelemetry/semantic-conventions": "^1.29.0"
       },
       "engines": {
@@ -6900,32 +6763,15 @@
         "@opentelemetry/api": ">=1.4.0 <1.10.0"
       }
     },
-    "node_modules/@opentelemetry/sdk-logs/node_modules/@opentelemetry/resources": {
-      "version": "2.8.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/resources/-/resources-2.8.0.tgz",
-      "integrity": "sha512-qmXQ27ilDbUK/vGMqwL8D4/rhn76C+sherM4wTbjlfknR8Nvfc/hCxjRJPhkzZzUsPiNg16SA31NxMabwttRjg==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.3.0 <1.10.0"
-      }
-    },
     "node_modules/@opentelemetry/sdk-metrics": {
-      "version": "2.8.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-metrics/-/sdk-metrics-2.8.0.tgz",
-      "integrity": "sha512-UDBGaj6W0Rgy5rTTaoxs8gVGF/aGkAKyjurJv7se6wjRxJu7FoquTLT/vt54DZfo4crbprYfhX/SOK9+BPw1qg==",
+      "version": "2.9.0",
+      "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-metrics/-/sdk-metrics-2.9.0.tgz",
+      "integrity": "sha512-Xx8RGS4H5XEBl01WuCreMIpiah9cCXMbSkeuIePPdD2cUpq/vUzYmj8E/MK1OsbOc93FuAD4jfn2WOacKwLn7Q==",
       "dev": true,
       "license": "Apache-2.0",
       "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/resources": "2.8.0"
+        "@opentelemetry/core": "2.9.0",
+        "@opentelemetry/resources": "2.9.0"
       },
       "engines": {
         "node": "^18.19.0 || >=20.6.0"
@@ -6934,23 +6780,6 @@
         "@opentelemetry/api": ">=1.9.0 <1.10.0"
       }
     },
-    "node_modules/@opentelemetry/sdk-metrics/node_modules/@opentelemetry/resources": {
-      "version": "2.8.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/resources/-/resources-2.8.0.tgz",
-      "integrity": "sha512-qmXQ27ilDbUK/vGMqwL8D4/rhn76C+sherM4wTbjlfknR8Nvfc/hCxjRJPhkzZzUsPiNg16SA31NxMabwttRjg==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.3.0 <1.10.0"
-      }
-    },
     "node_modules/@opentelemetry/sdk-trace": {
       "version": "2.9.0",
       "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-trace/-/sdk-trace-2.9.0.tgz",
@@ -6988,22 +6817,6 @@
         "@opentelemetry/api": ">=1.3.0 <1.10.0"
       }
     },
-    "node_modules/@opentelemetry/sdk-trace-base/node_modules/@opentelemetry/core": {
-      "version": "2.9.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/core/-/core-2.9.0.tgz",
-      "integrity": "sha512-m2nckMT80NnmjTYSPjJQObBJ+8dgkoajEOUbznL8AHZ3T3yHRk2P7gI1PhEBc1+lOnrYE9UWrWHqJDsmqjmNbw==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.0.0 <1.10.0"
-      }
-    },
     "node_modules/@opentelemetry/sdk-trace-node": {
       "version": "2.9.0",
       "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-trace-node/-/sdk-trace-node-2.9.0.tgz",
@@ -7022,38 +6835,6 @@
         "@opentelemetry/api": ">=1.0.0 <1.10.0"
       }
     },
-    "node_modules/@opentelemetry/sdk-trace-node/node_modules/@opentelemetry/core": {
-      "version": "2.9.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/core/-/core-2.9.0.tgz",
-      "integrity": "sha512-m2nckMT80NnmjTYSPjJQObBJ+8dgkoajEOUbznL8AHZ3T3yHRk2P7gI1PhEBc1+lOnrYE9UWrWHqJDsmqjmNbw==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.0.0 <1.10.0"
-      }
-    },
-    "node_modules/@opentelemetry/sdk-trace/node_modules/@opentelemetry/core": {
-      "version": "2.9.0",
-      "resolved": "https://registry.npmjs.org/@opentelemetry/core/-/core-2.9.0.tgz",
-      "integrity": "sha512-m2nckMT80NnmjTYSPjJQObBJ+8dgkoajEOUbznL8AHZ3T3yHRk2P7gI1PhEBc1+lOnrYE9UWrWHqJDsmqjmNbw==",
-      "dev": true,
-      "license": "Apache-2.0",
-      "dependencies": {
-        "@opentelemetry/semantic-conventions": "^1.29.0"
-      },
-      "engines": {
-        "node": "^18.19.0 || >=20.6.0"
-      },
-      "peerDependencies": {
-        "@opentelemetry/api": ">=1.0.0 <1.10.0"
-      }
-    },
     "node_modules/@opentelemetry/semantic-conventions": {
       "version": "1.43.0",
       "resolved": "https://registry.npmjs.org/@opentelemetry/semantic-conventions/-/semantic-conventions-1.43.0.tgz",
@@ -10117,9 +9898,9 @@
       }
     },
     "node_modules/@smithy/core": {
-      "version": "3.29.3",
-      "resolved": "https://registry.npmjs.org/@smithy/core/-/core-3.29.3.tgz",
-      "integrity": "sha512-L+Ys6ecjk5vwPMAKHBpPKlJ3DkqwNcnfEISXBZIsVvWG/XKXfsAP8mwIYlTeLcd2ElHdesPI8OuOmJSFAPhm6A==",
+      "version": "3.29.4",
+      "resolved": "https://registry.npmjs.org/@smithy/core/-/core-3.29.4.tgz",
+      "integrity": "sha512-G1GRglAabzEhqghJMBAd54FkRS7SAFGHEwbhcI9r+O+LIMuFsLyXkLZkCoFSgAglRu8s/URVXJB0hglq3ZipIg==",
       "license": "Apache-2.0",
       "dependencies": {
         "@smithy/types": "^4.16.1",
@@ -10130,12 +9911,12 @@
       }
     },
     "node_modules/@smithy/credential-provider-imds": {
-      "version": "4.4.8",
-      "resolved": "https://registry.npmjs.org/@smithy/credential-provider-imds/-/credential-provider-imds-4.4.8.tgz",
-      "integrity": "sha512-q9J7JTiXrAhB8sDp4px97uEPT7CwKH61Co78grdNQvU8QZAdiuaSRhP0tUVf2ogy36RZTrlMU1rBmDEH+cnkiA==",
+      "version": "4.4.9",
+      "resolved": "https://registry.npmjs.org/@smithy/credential-provider-imds/-/credential-provider-imds-4.4.9.tgz",
+      "integrity": "sha512-2nfV4qRKiYeXU4zD2vvSCfg5dfp/BuhrM73vt7q9gzBhxs4rbPxXY21wo+kyI3bRmXcEGRnCLTaW8O437jzHIg==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@smithy/core": "^3.29.3",
+        "@smithy/core": "^3.29.4",
         "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
@@ -10144,12 +9925,12 @@
       }
     },
     "node_modules/@smithy/fetch-http-handler": {
-      "version": "5.6.5",
-      "resolved": "https://registry.npmjs.org/@smithy/fetch-http-handler/-/fetch-http-handler-5.6.5.tgz",
-      "integrity": "sha512-SuqeisTyPoiIPtIYru/sGxGyXzmZ+8nnFOhC+qRPglt06Ebd1yH//CDltZB2J/3WBNVhwfUaZ0EtHB3cm2X32g==",
+      "version": "5.6.6",
+      "resolved": "https://registry.npmjs.org/@smithy/fetch-http-handler/-/fetch-http-handler-5.6.6.tgz",
+      "integrity": "sha512-NHLgAlORUFZjn5ZfhYuyyKMlXA1WLYOdGxEhyNxrPpbJzoacGbl0chn1lN2KiZ8mpNVk0tV5607CSYlYs/OFgw==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@smithy/core": "^3.29.3",
+        "@smithy/core": "^3.29.4",
         "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
@@ -10158,12 +9939,12 @@
       }
     },
     "node_modules/@smithy/node-http-handler": {
-      "version": "4.9.5",
-      "resolved": "https://registry.npmjs.org/@smithy/node-http-handler/-/node-http-handler-4.9.5.tgz",
-      "integrity": "sha512-bNqdxTQTxmLbomSmlkZFz8L6B/feQ2HHzw4L2zY7Ecp2XffYAZq2uzdWDdxJHJFbEvqd+SRuluJso0P8+xPdbw==",
+      "version": "4.9.6",
+      "resolved": "https://registry.npmjs.org/@smithy/node-http-handler/-/node-http-handler-4.9.6.tgz",
+      "integrity": "sha512-odd+HYx3OLcXRSEz0ZeF3JQdSYdK8QnRgA2N87cPW7coWIbKfRk7a9VQjfeWQLqnzrDLk23KMEn46p8N7M/JFg==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@smithy/core": "^3.29.3",
+        "@smithy/core": "^3.29.4",
         "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
@@ -10172,12 +9953,12 @@
       }
     },
     "node_modules/@smithy/signature-v4": {
-      "version": "5.6.4",
-      "resolved": "https://registry.npmjs.org/@smithy/signature-v4/-/signature-v4-5.6.4.tgz",
-      "integrity": "sha512-B89bpf2t/y/wia6LZ+4JfHXYQT9PnVftsH05rgJKKIStS7r/4XSs9HOjtPoLtgcA6HCW9jVqX5DBbq7E0PAkiQ==",
+      "version": "5.6.5",
+      "resolved": "https://registry.npmjs.org/@smithy/signature-v4/-/signature-v4-5.6.5.tgz",
+      "integrity": "sha512-MO5VEhwVl0BN7xVoVeNrZfiUFoQtqxUbgl6/RwOTlMMxCSjblG8twSrVTwz3J4w9WZxd2rBfBAUXjH77agspBg==",
       "license": "Apache-2.0",
       "dependencies": {
-        "@smithy/core": "^3.29.3",
+        "@smithy/core": "^3.29.4",
         "@smithy/types": "^4.16.1",
         "tslib": "^2.6.2"
       },
@@ -11981,9 +11762,9 @@
       "license": "MIT"
     },
     "node_modules/@types/node": {
-      "version": "26.1.0",
-      "resolved": "https://registry.npmjs.org/@types/node/-/node-26.1.0.tgz",
-      "integrity": "sha512-O0A1G3xPGy4w7AgQdAQYUlQ+BKk2Oovw8eRpofyp5KdBZULnbe+WqaOVNrm705SHphCiG4XHsACrSmPu1f+Kgw==",
+      "version": "26.1.1",
+      "resolved": "https://registry.npmjs.org/@types/node/-/node-26.1.1.tgz",
+      "integrity": "sha512-nxAkRSVkN1Y0JC1W8ky/fTfkGsMmcrRsbx+3XoZE+rMOX71kLYTV7fLXpqud1GpbpP5TuffXFqfX7fH2GgZREw==",
       "devOptional": true,
       "license": "MIT",
       "dependencies": {
@@ -12155,17 +11936,17 @@
       }
     },
     "node_modules/@typescript-eslint/eslint-plugin": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/eslint-plugin/-/eslint-plugin-8.63.0.tgz",
-      "integrity": "sha512-rvwSgqT+DHpWdzfSzPatRLm02a0GlESt++9iy3hLCDY4BgkaLcl8LBi9Yh7XGFBpwcBE/K3024QuXWTpbz4FfQ==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/eslint-plugin/-/eslint-plugin-8.64.0.tgz",
+      "integrity": "sha512-CGvQPBxN3wZLu6Rz2kFUpZeoCm78xUic92ck39KPePkO1NPOwjCqdQnm5Q87tpWw9vcBvW8XLrDXjH9PWYtJ3Q==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
         "@eslint-community/regexpp": "^4.12.2",
-        "@typescript-eslint/scope-manager": "8.63.0",
-        "@typescript-eslint/type-utils": "8.63.0",
-        "@typescript-eslint/utils": "8.63.0",
-        "@typescript-eslint/visitor-keys": "8.63.0",
+        "@typescript-eslint/scope-manager": "8.64.0",
+        "@typescript-eslint/type-utils": "8.64.0",
+        "@typescript-eslint/utils": "8.64.0",
+        "@typescript-eslint/visitor-keys": "8.64.0",
         "ignore": "^7.0.5",
         "natural-compare": "^1.4.0",
         "ts-api-utils": "^2.5.0"
@@ -12178,15 +11959,15 @@
         "url": "https://opencollective.com/typescript-eslint"
       },
       "peerDependencies": {
-        "@typescript-eslint/parser": "^8.63.0",
+        "@typescript-eslint/parser": "^8.64.0",
         "eslint": "^8.57.0 || ^9.0.0 || ^10.0.0",
         "typescript": ">=4.8.4 <6.1.0"
       }
     },
     "node_modules/@typescript-eslint/eslint-plugin/node_modules/ignore": {
-      "version": "7.0.5",
-      "resolved": "https://registry.npmjs.org/ignore/-/ignore-7.0.5.tgz",
-      "integrity": "sha512-Hs59xBNfUIunMFgWAbGX5cq6893IbWg4KnrjbYwX3tx0ztorVgTDA6B2sxf8ejHJ4wz8BqGUMYlnzNBer5NvGg==",
+      "version": "7.0.6",
+      "resolved": "https://registry.npmjs.org/ignore/-/ignore-7.0.6.tgz",
+      "integrity": "sha512-BAg6QkE8W+TuQLrrw0Ugr7HegXduRuuj8/ti2kSOc+jz1dmx8/WNcjr6XGnq5YpDWxFwwaavqD0+jIUOKelTsw==",
       "dev": true,
       "license": "MIT",
       "engines": {
@@ -12194,16 +11975,16 @@
       }
     },
     "node_modules/@typescript-eslint/parser": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/parser/-/parser-8.63.0.tgz",
-      "integrity": "sha512-gwh4gvvlaVDKKxyfxMG+Gnu1u9X0OQBwyGLkbwB65dIzBKnxeRiJlNFqlI3zwVhNXJIs6qV7mlFCn/BIajlVig==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/parser/-/parser-8.64.0.tgz",
+      "integrity": "sha512-KA0OshtlcCCXmbfqyZkM5pV3/WNraJf7DkJRLpyrmwPtud57H5BDX7C3k0LPSPxpprfRL+cJDGabF10mvNCoCw==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
-        "@typescript-eslint/scope-manager": "8.63.0",
-        "@typescript-eslint/types": "8.63.0",
-        "@typescript-eslint/typescript-estree": "8.63.0",
-        "@typescript-eslint/visitor-keys": "8.63.0",
+        "@typescript-eslint/scope-manager": "8.64.0",
+        "@typescript-eslint/types": "8.64.0",
+        "@typescript-eslint/typescript-estree": "8.64.0",
+        "@typescript-eslint/visitor-keys": "8.64.0",
         "debug": "^4.4.3"
       },
       "engines": {
@@ -12219,14 +12000,14 @@
       }
     },
     "node_modules/@typescript-eslint/project-service": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/project-service/-/project-service-8.63.0.tgz",
-      "integrity": "sha512-e5dh0/UI0ok53AlZ5wRkXCB32z/f2jUZqPR/ygAw5WYaSw8j9EoJWlS7wQjr/dmOaqWjnPIn2m+HhVPCMWGZVQ==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/project-service/-/project-service-8.64.0.tgz",
+      "integrity": "sha512-tk4WpOJ6IEbGrVHaNmM0YRrwAD3exZlIK3iadQNAxh4YKk6jvUQ4ecq18n+v7+meh+cJ3j+D8nbk8sRKhlwLQg==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
-        "@typescript-eslint/tsconfig-utils": "^8.63.0",
-        "@typescript-eslint/types": "^8.63.0",
+        "@typescript-eslint/tsconfig-utils": "^8.64.0",
+        "@typescript-eslint/types": "^8.64.0",
         "debug": "^4.4.3"
       },
       "engines": {
@@ -12241,14 +12022,14 @@
       }
     },
     "node_modules/@typescript-eslint/scope-manager": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/scope-manager/-/scope-manager-8.63.0.tgz",
-      "integrity": "sha512-uUyfMWCnDSN8bCpcrY8nGP2BLkQ9Xn0GsipcONcpIDWhwhO4ZSyHvyS14U3X75mzxWxL3I2UZIrenTzdzcJO8A==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/scope-manager/-/scope-manager-8.64.0.tgz",
+      "integrity": "sha512-CXEaFdYXjSTgKhisNkwCcJwTP8Pl+fmRrEQrri4nm3vU743bALrxzLmq7fHG/7e6a5xO0lDYeURpZmBuhHk54w==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
-        "@typescript-eslint/types": "8.63.0",
-        "@typescript-eslint/visitor-keys": "8.63.0"
+        "@typescript-eslint/types": "8.64.0",
+        "@typescript-eslint/visitor-keys": "8.64.0"
       },
       "engines": {
         "node": "^18.18.0 || ^20.9.0 || >=21.1.0"
@@ -12259,9 +12040,9 @@
       }
     },
     "node_modules/@typescript-eslint/tsconfig-utils": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/tsconfig-utils/-/tsconfig-utils-8.63.0.tgz",
-      "integrity": "sha512-sUAbkulqBAsncKnbRP3+7CtQFRKicexnj7ZwNC6ddCR7EmrXvjvdCYMJbUIqMd6lwoEriZjwLo08aS5tSjVMHg==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/tsconfig-utils/-/tsconfig-utils-8.64.0.tgz",
+      "integrity": "sha512-2yo8rRNKuzbVWQp5kslhANqZ2uDAeROQHBRZNPu8JDsHmeFNj/XJJhX/FhNUWmkHHvoNsKa6+tHJiig87EzsQw==",
       "dev": true,
       "license": "MIT",
       "engines": {
@@ -12276,15 +12057,15 @@
       }
     },
     "node_modules/@typescript-eslint/type-utils": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/type-utils/-/type-utils-8.63.0.tgz",
-      "integrity": "sha512-Nzzh/OGxVCOjObjaj1CQF2RUasyYy2Jfuh+zZ3PjLzG2fYRriAiZLib9UKtO+CpQAS3YHiAS+ckZDclwqI1TPA==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/type-utils/-/type-utils-8.64.0.tgz",
+      "integrity": "sha512-XWG4Fmmv/6SvyS9nH8jWrKs6terwJvE8cyRt1CzYYqzp9OrPhCT4cMc/f7C6RZCwG+qMmiffJS1/qJP8G1URtg==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
-        "@typescript-eslint/types": "8.63.0",
-        "@typescript-eslint/typescript-estree": "8.63.0",
-        "@typescript-eslint/utils": "8.63.0",
+        "@typescript-eslint/types": "8.64.0",
+        "@typescript-eslint/typescript-estree": "8.64.0",
+        "@typescript-eslint/utils": "8.64.0",
         "debug": "^4.4.3",
         "ts-api-utils": "^2.5.0"
       },
@@ -12301,9 +12082,9 @@
       }
     },
     "node_modules/@typescript-eslint/types": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/types/-/types-8.63.0.tgz",
-      "integrity": "sha512-xyLtl9DUBBFrcJS4x2pIqGLH68/tC2uOa4Z7pUteW09D3bXnnXUom4dyPikzWgB7llmIc1zoeI3aoUdC4rPK/Q==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/types/-/types-8.64.0.tgz",
+      "integrity": "sha512-qjhfuTfLXjA4IOzXvz0rTjT01BqEiIgPoUeMwiEjnaHKJMTNo8rH5pYW1a2L/0Dnux2fPC85AeyJoWaGa8WxTA==",
       "dev": true,
       "license": "MIT",
       "engines": {
@@ -12315,16 +12096,16 @@
       }
     },
     "node_modules/@typescript-eslint/typescript-estree": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/typescript-estree/-/typescript-estree-8.63.0.tgz",
-      "integrity": "sha512-ygBkU+B7ex5UI/gKhaqexWev79uISfIv7XQCRNYO/jmD8rGLPyWLAb3KMRT6nd8Gt9bmUBi9+iX6tBdYfOY81Q==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/typescript-estree/-/typescript-estree-8.64.0.tgz",
+      "integrity": "sha512-Pztpsn1aCE1oWDvDEfUk31nngvvF7vUB5SwHFEaZIFpvw7WJtqUHHL4plBZDA9HfWJJjL13BdG0YrJInTUvoVA==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
-        "@typescript-eslint/project-service": "8.63.0",
-        "@typescript-eslint/tsconfig-utils": "8.63.0",
-        "@typescript-eslint/types": "8.63.0",
-        "@typescript-eslint/visitor-keys": "8.63.0",
+        "@typescript-eslint/project-service": "8.64.0",
+        "@typescript-eslint/tsconfig-utils": "8.64.0",
+        "@typescript-eslint/types": "8.64.0",
+        "@typescript-eslint/visitor-keys": "8.64.0",
         "debug": "^4.4.3",
         "minimatch": "^10.2.2",
         "semver": "^7.7.3",
@@ -12395,16 +12176,16 @@
       }
     },
     "node_modules/@typescript-eslint/utils": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/utils/-/utils-8.63.0.tgz",
-      "integrity": "sha512-fUKaeAvrTuQg/Tgt3nliAUSZHJM6DlCcfyEmxCvlX8kieWSStBX+5O5Fnidtc3i2JrH+9c/GL4RY2iasd/GPTA==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/utils/-/utils-8.64.0.tgz",
+      "integrity": "sha512-aJUGVB3+U0htrrCjoA8qukw8cm8fNCGAxK/tVoS70k8aeb7DETKeFozRiVFIwEeN9WJLsjaP3ph8I60tY2XZoQ==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
         "@eslint-community/eslint-utils": "^4.9.1",
-        "@typescript-eslint/scope-manager": "8.63.0",
-        "@typescript-eslint/types": "8.63.0",
-        "@typescript-eslint/typescript-estree": "8.63.0"
+        "@typescript-eslint/scope-manager": "8.64.0",
+        "@typescript-eslint/types": "8.64.0",
+        "@typescript-eslint/typescript-estree": "8.64.0"
       },
       "engines": {
         "node": "^18.18.0 || ^20.9.0 || >=21.1.0"
@@ -12419,13 +12200,13 @@
       }
     },
     "node_modules/@typescript-eslint/visitor-keys": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/@typescript-eslint/visitor-keys/-/visitor-keys-8.63.0.tgz",
-      "integrity": "sha512-UexrHGnGTpbuQHct2ExOc2ZcFbGUS9FOesCxxqdBGcpI1BxYu/LZ6U8Aq6/72XtF/qRBk9nhuGHFJIXXMhPMdw==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/visitor-keys/-/visitor-keys-8.64.0.tgz",
+      "integrity": "sha512-mrtuL8Nsn6gi2H4mo5KMTp823M+3Q19Ew/i+Zlikq20tIMm99C3Ez0dCmkWWnxut20esQvTg8aUSEhMcAOXhEw==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
-        "@typescript-eslint/types": "8.63.0",
+        "@typescript-eslint/types": "8.64.0",
         "eslint-visitor-keys": "^5.0.0"
       },
       "engines": {
@@ -13056,6 +12837,162 @@
         "js-yaml": "bin/js-yaml.js"
       }
     },
+    "node_modules/@yuku-analyzer/binding-darwin-arm64": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-darwin-arm64/-/binding-darwin-arm64-0.6.3.tgz",
+      "integrity": "sha512-1PI1tdfk0ozQ0tbEi740fYMz/3axKn+jR2nK2qBXdYZiyQKsPYW7lDockNbUY9Z5E3+nwEFjX6Pp19X4VIgrkQ==",
+      "cpu": [
+        "arm64"
+      ],
+      "optional": true,
+      "os": [
+        "darwin"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-darwin-x64": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-darwin-x64/-/binding-darwin-x64-0.6.3.tgz",
+      "integrity": "sha512-VyC+KH0gwPzXjtysXbuBop+Qn107800pQhp8YzDElnBciu/X88Uw3xEJrCJtcyoV85sPpq+g1zvAgHWHwE8PlQ==",
+      "cpu": [
+        "x64"
+      ],
+      "optional": true,
+      "os": [
+        "darwin"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-freebsd-x64": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-freebsd-x64/-/binding-freebsd-x64-0.6.3.tgz",
+      "integrity": "sha512-T5HRWQiy0e5bHaI01xn3cguopn0YCvNV4rast6p+o4ZjORLguCwM7i3GujOUXMzwQbQ/GFxI/ToDwYfI2IFFgg==",
+      "cpu": [
+        "x64"
+      ],
+      "optional": true,
+      "os": [
+        "freebsd"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-linux-arm-gnu": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-linux-arm-gnu/-/binding-linux-arm-gnu-0.6.3.tgz",
+      "integrity": "sha512-QVMkLA7vqtADSl9sKpX0oDO9X9BmVO6rzlxwU2mJ8vNoYOucOfxOVj1NLW4I3p6m4TW2lIX6zW7kT81HVRmQtw==",
+      "cpu": [
+        "arm"
+      ],
+      "libc": [
+        "glibc"
+      ],
+      "optional": true,
+      "os": [
+        "linux"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-linux-arm-musl": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-linux-arm-musl/-/binding-linux-arm-musl-0.6.3.tgz",
+      "integrity": "sha512-MdgimxnvfC4uAMDs0UsQW8wGWi6im+cptlcdQi3l+FS5ouf0tmdIo6O5HoteM6zazUtc+vBEr9g1A4J5eFFKNw==",
+      "cpu": [
+        "arm"
+      ],
+      "libc": [
+        "musl"
+      ],
+      "optional": true,
+      "os": [
+        "linux"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-linux-arm64-gnu": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-linux-arm64-gnu/-/binding-linux-arm64-gnu-0.6.3.tgz",
+      "integrity": "sha512-dRYQU8024UvbDnfU3yNDl4NAjLptjkog+Fbd7TsFvQKc9P7rMocC0CLWvsbp6GVu6mo1yATb7GQuSA0V/MBk7g==",
+      "cpu": [
+        "arm64"
+      ],
+      "libc": [
+        "glibc"
+      ],
+      "optional": true,
+      "os": [
+        "linux"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-linux-arm64-musl": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-linux-arm64-musl/-/binding-linux-arm64-musl-0.6.3.tgz",
+      "integrity": "sha512-RNj/MBlBYVamdO+Zexxj+tYQiRBPHUMHOLNpCXJ2sraVvKc+aT+HzgWwanG1TDL7RR1hfaDxpsmJpLgtDw65nQ==",
+      "cpu": [
+        "arm64"
+      ],
+      "libc": [
+        "musl"
+      ],
+      "optional": true,
+      "os": [
+        "linux"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-linux-x64-gnu": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-linux-x64-gnu/-/binding-linux-x64-gnu-0.6.3.tgz",
+      "integrity": "sha512-gUHi0GkcJOfGc+RHkqyTSpjyLRNIm0cZUSEOVdH309iXDEMqD9E0Fz3kUoxepIrBwkHDAjF3xYrp0hEGogfkUw==",
+      "cpu": [
+        "x64"
+      ],
+      "libc": [
+        "glibc"
+      ],
+      "optional": true,
+      "os": [
+        "linux"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-linux-x64-musl": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-linux-x64-musl/-/binding-linux-x64-musl-0.6.3.tgz",
+      "integrity": "sha512-7xIqcdYwyf6mSCJid9C0ZMxd6KdN3S72Ywnws55Wz3O7XWvVscZj+51DHAeqKklsD0gXey4evLQkaDtKxd8jBQ==",
+      "cpu": [
+        "x64"
+      ],
+      "libc": [
+        "musl"
+      ],
+      "optional": true,
+      "os": [
+        "linux"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-win32-arm64": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-win32-arm64/-/binding-win32-arm64-0.6.3.tgz",
+      "integrity": "sha512-tyU9RPF0reQ4Lu2JKDhsSZZIqSHJZdO3QD7vfcJQDhZAU2JBvfubpVejvf3uvAS+/2c0Ajfgj/K1JpDyAqVbQA==",
+      "cpu": [
+        "arm64"
+      ],
+      "optional": true,
+      "os": [
+        "win32"
+      ]
+    },
+    "node_modules/@yuku-analyzer/binding-win32-x64": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/@yuku-analyzer/binding-win32-x64/-/binding-win32-x64-0.6.3.tgz",
+      "integrity": "sha512-84vgw5+SNDYhTJYFkLkgobYM++1IbN8fPY4QIRVSBuLZftHvuaKQnrjQTQkXvmPc0fiebY8ITnxNjHydDrItDA==",
+      "cpu": [
+        "x64"
+      ],
+      "optional": true,
+      "os": [
+        "win32"
+      ]
+    },
+    "node_modules/@yuku-toolchain/types": {
+      "version": "0.5.43",
+      "resolved": "https://registry.npmjs.org/@yuku-toolchain/types/-/types-0.5.43.tgz",
+      "integrity": "sha512-kSpvPntnXw5+lYjO71ffBEnQ5ycQ74KGIYknh0TS4xeyCuBkOqxyJumxZkMhLBBUCLjDAbx2+Icnr3Zh4ftjpQ==",
+      "license": "MIT"
+    },
     "node_modules/a-sync-waterfall": {
       "version": "1.0.1",
       "resolved": "https://registry.npmjs.org/a-sync-waterfall/-/a-sync-waterfall-1.0.1.tgz",
@@ -15951,67 +15888,6 @@
         "node": ">=20"
       }
     },
-    "node_modules/cross-fetch": {
-      "version": "4.1.0",
-      "resolved": "https://registry.npmjs.org/cross-fetch/-/cross-fetch-4.1.0.tgz",
-      "integrity": "sha512-uKm5PU+MHTootlWEY+mZ4vvXoCn4fLQxT9dSc1sXVMSFkINTJVN8cAQROpwcKm8bJ/c7rgZVIBWzH5T78sNZZw==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "node-fetch": "^2.7.0"
-      }
-    },
-    "node_modules/cross-fetch/node_modules/node-fetch": {
-      "version": "2.7.0",
-      "resolved": "https://registry.npmjs.org/node-fetch/-/node-fetch-2.7.0.tgz",
-      "integrity": "sha512-c4FRfUm/dbcWZ7U+1Wq0AwCyFL+3nt2bEw05wfxSz+DWpWsitgmSgYmy2dQdWyKC1694ELPqMs/YzUSNozLt8A==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "whatwg-url": "^5.0.0"
-      },
-      "engines": {
-        "node": "4.x || >=6.0.0"
-      },
-      "peerDependencies": {
-        "encoding": "^0.1.0"
-      },
-      "peerDependenciesMeta": {
-        "encoding": {
-          "optional": true
-        }
-      }
-    },
-    "node_modules/cross-fetch/node_modules/tr46": {
-      "version": "0.0.3",
-      "resolved": "https://registry.npmjs.org/tr46/-/tr46-0.0.3.tgz",
-      "integrity": "sha512-N3WMsuqV66lT30CrXNbEjx4GEwlow3v6rr4mCcv6prnfwhS01rkgyFdjPNBYd9br7LpXV1+Emh01fHnq2Gdgrw==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true
-    },
-    "node_modules/cross-fetch/node_modules/webidl-conversions": {
-      "version": "3.0.1",
-      "resolved": "https://registry.npmjs.org/webidl-conversions/-/webidl-conversions-3.0.1.tgz",
-      "integrity": "sha512-2JAn3z8AR6rjK8Sm8orRC0h/bcl/DqL7tRPdGZ4I1CjdF+EaMLmYxBHyXuKL849eucPFhvBoxMsflfOb8kxaeQ==",
-      "dev": true,
-      "license": "BSD-2-Clause",
-      "optional": true
-    },
-    "node_modules/cross-fetch/node_modules/whatwg-url": {
-      "version": "5.0.0",
-      "resolved": "https://registry.npmjs.org/whatwg-url/-/whatwg-url-5.0.0.tgz",
-      "integrity": "sha512-saE57nupxk6v3HY35+jzBwYa0rKSy0XR8JSxZPwgLr7ys0IBzhGviA1/TUGJLmSVqs8pb9AnvICXEuOHLprYTw==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "tr46": "~0.0.3",
-        "webidl-conversions": "^3.0.0"
-      }
-    },
     "node_modules/cross-spawn": {
       "version": "7.0.6",
       "resolved": "https://registry.npmjs.org/cross-spawn/-/cross-spawn-7.0.6.tgz",
@@ -17953,9 +17829,9 @@
       }
     },
     "node_modules/es-toolkit": {
-      "version": "1.45.1",
-      "resolved": "https://registry.npmjs.org/es-toolkit/-/es-toolkit-1.45.1.tgz",
-      "integrity": "sha512-/jhoOj/Fx+A+IIyDNOvO3TItGmlMKhtX8ISAHKE90c4b/k1tqaqEZ+uUqfpU8DMnW5cgNJv606zS55jGvza0Xw==",
+      "version": "1.49.0",
+      "resolved": "https://registry.npmjs.org/es-toolkit/-/es-toolkit-1.49.0.tgz",
+      "integrity": "sha512-G5iZ6Pc/FNRY/soKZHC+TxGDD83rHUDXxzaWhGCX44vAv/tMs56WMusnm/KMNK+luUPsgA9U28cGr4RDlSzL2g==",
       "license": "MIT",
       "workspaces": [
         "docs",
@@ -18133,9 +18009,9 @@
       }
     },
     "node_modules/eslint": {
-      "version": "9.39.4",
-      "resolved": "https://registry.npmjs.org/eslint/-/eslint-9.39.4.tgz",
-      "integrity": "sha512-XoMjdBOwe/esVgEvLmNsD3IRHkm7fbKIUGvrleloJXUZgDHig2IPWNniv+GwjyJXzuNqVjlr5+4yVUZjycJwfQ==",
+      "version": "9.39.5",
+      "resolved": "https://registry.npmjs.org/eslint/-/eslint-9.39.5.tgz",
+      "integrity": "sha512-DgZS62aPLXKlnxILS/AYCoRvHaZeXceIzlXPkkGGzJWSow1aEk0lbTlxUSlyjC8jcaKxAdOnTDz+o1JFSBsyjw==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
@@ -18144,8 +18020,8 @@
         "@eslint/config-array": "^0.21.2",
         "@eslint/config-helpers": "^0.4.2",
         "@eslint/core": "^0.17.0",
-        "@eslint/eslintrc": "^3.3.5",
-        "@eslint/js": "9.39.4",
+        "@eslint/eslintrc": "^3.3.6",
+        "@eslint/js": "9.39.5",
         "@eslint/plugin-kit": "^0.4.1",
         "@humanfs/node": "^0.16.6",
         "@humanwhocodes/module-importer": "^1.0.1",
@@ -18486,9 +18362,9 @@
       }
     },
     "node_modules/eslint-plugin-sonarjs": {
-      "version": "4.1.0",
-      "resolved": "https://registry.npmjs.org/eslint-plugin-sonarjs/-/eslint-plugin-sonarjs-4.1.0.tgz",
-      "integrity": "sha512-rh+FlVz0yfd2RNIb6WqSkuGh0addX/Qi5scwQ5FphXDFrM6fZKcxP1+attJ78yUKcyYfiu6MTaISPpAFPzqRJw==",
+      "version": "4.2.0",
+      "resolved": "https://registry.npmjs.org/eslint-plugin-sonarjs/-/eslint-plugin-sonarjs-4.2.0.tgz",
+      "integrity": "sha512-bqADfuNtTL7VK6RU29eoiFTtaaBKIpVPuX3bOl+rBpWSBa0zIBVZlqZNZQjfP6s4iXkAJokv5IsD8OsACkwApg==",
       "dev": true,
       "license": "LGPL-3.0-only",
       "dependencies": {
@@ -18496,14 +18372,14 @@
         "builtin-modules": "^3.3.0",
         "bytes": "^3.1.2",
         "functional-red-black-tree": "^1.0.1",
-        "globals": "^17.6.0",
+        "globals": "^17.7.0",
         "jsx-ast-utils-x": "^0.1.0",
         "lodash.merge": "^4.6.2",
         "minimatch": "^10.2.5",
         "scslre": "^0.3.0",
-        "semver": "^7.8.4",
+        "semver": "^7.8.5",
         "ts-api-utils": "^2.5.0",
-        "typescript": ">=5",
+        "typescript": ">=5 <6.1.0",
         "yaml": "^2.9.0"
       },
       "peerDependencies": {
@@ -18534,9 +18410,9 @@
       }
     },
     "node_modules/eslint-plugin-sonarjs/node_modules/globals": {
-      "version": "17.6.0",
-      "resolved": "https://registry.npmjs.org/globals/-/globals-17.6.0.tgz",
-      "integrity": "sha512-sepffkT8stwnIYbsMBpoCHJuJM5l98FUF2AnE07hfvE0m/qp3R586hw4jF4uadbhvg1ooIdzuu7CsfD2jzCaNA==",
+      "version": "17.7.0",
+      "resolved": "https://registry.npmjs.org/globals/-/globals-17.7.0.tgz",
+      "integrity": "sha512-Czmyns5dUsq4seFBR/Kdydhmo8y9kC79hiSkPn0YcGtNnYWnrgt0vjrSjx9tspoDGWm2CMarffRuLjM4xUz8xg==",
       "dev": true,
       "license": "MIT",
       "engines": {
@@ -18563,9 +18439,9 @@
       }
     },
     "node_modules/eslint-plugin-sonarjs/node_modules/semver": {
-      "version": "7.8.4",
-      "resolved": "https://registry.npmjs.org/semver/-/semver-7.8.4.tgz",
-      "integrity": "sha512-rUCObTnP32Q08R2uuIrt7r9PlEonuTmtuXYcW6s5kjdlj3xbnwe+21yXptAUYcMAABLkYYTtnmzb3w3EDZfueA==",
+      "version": "7.8.5",
+      "resolved": "https://registry.npmjs.org/semver/-/semver-7.8.5.tgz",
+      "integrity": "sha512-Y7/KDsb8LjooZpwaqGyulO6DQlksgCncchHGk+sZIY4SBvUocMBEFH5Ur1fI4dV+Jvl0w6cjvucaIi40puRioA==",
       "dev": true,
       "license": "ISC",
       "bin": {
@@ -19051,9 +18927,9 @@
       "license": "MIT"
     },
     "node_modules/fast-check": {
-      "version": "4.8.0",
-      "resolved": "https://registry.npmjs.org/fast-check/-/fast-check-4.8.0.tgz",
-      "integrity": "sha512-GOJ158CUMnN6cSahsv4+ExARvIDuzzinFjkp0E9WtiBa5zcVeLozVkWaE4IzFcc+Y48Wp1EDlUZsXRyAztQcSg==",
+      "version": "4.9.0",
+      "resolved": "https://registry.npmjs.org/fast-check/-/fast-check-4.9.0.tgz",
+      "integrity": "sha512-7ms6T7SybUev/PQITciI0yLM2pOSFy5zpG8Ty7tQofcVaQUvrMXp6CBwqF6fThLCLOrfBtuHAtwq6Yu4XPCllg==",
       "dev": true,
       "funding": [
         {
@@ -19318,14 +19194,6 @@
         "node": "^12.20 || >= 14.13"
       }
     },
-    "node_modules/fetch-retry": {
-      "version": "5.0.6",
-      "resolved": "https://registry.npmjs.org/fetch-retry/-/fetch-retry-5.0.6.tgz",
-      "integrity": "sha512-3yurQZ2hD9VISAhJJP9bpYFNQrHHBXE2JxxjY5aLEcDi46RmAzJE2OC9FAde0yis5ElW0jTTzs0zfg/Cca4XqQ==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true
-    },
     "node_modules/fetch-socks": {
       "version": "1.3.3",
       "resolved": "https://registry.npmjs.org/fetch-socks/-/fetch-socks-1.3.3.tgz",
@@ -19727,9 +19595,9 @@
       }
     },
     "node_modules/fumadocs-core": {
-      "version": "16.11.1",
-      "resolved": "https://registry.npmjs.org/fumadocs-core/-/fumadocs-core-16.11.1.tgz",
-      "integrity": "sha512-tKuh1AKoVTb+f7IoAOM2cfz5djd3YhePeqA95q6mf422gEvDTeJms23OJ+icYRWZ6ryNQ5W/ZsgKEe87M5HVYg==",
+      "version": "16.11.5",
+      "resolved": "https://registry.npmjs.org/fumadocs-core/-/fumadocs-core-16.11.5.tgz",
+      "integrity": "sha512-YrHjS09+QYYKOSTGyiZbxF/VDs7ciMcjurYBGfmYqtzdj14k7Ho0HX9c6VuvG54YsYHQs5mGWemT1TXm7vDBaA==",
       "license": "MIT",
       "dependencies": {
         "@orama/orama": "^3.1.18",
@@ -19737,7 +19605,6 @@
         "github-slugger": "^2.0.0",
         "hast-util-to-estree": "^3.1.3",
         "hast-util-to-jsx-runtime": "^2.3.6",
-        "js-yaml": "^5.2.1",
         "mdast-util-mdx": "^3.0.0",
         "mdast-util-to-markdown": "^2.1.2",
         "remark": "^15.0.1",
@@ -19748,7 +19615,8 @@
         "tinyglobby": "^0.2.17",
         "unified": "^11.0.5",
         "unist-util-visit": "^5.1.0",
-        "vfile": "^6.0.3"
+        "vfile": "^6.0.3",
+        "yaml": "^2.9.0"
       },
       "peerDependencies": {
         "@mdx-js/mdx": "*",
@@ -19828,9 +19696,9 @@
       }
     },
     "node_modules/fumadocs-mdx": {
-      "version": "15.1.0",
-      "resolved": "https://registry.npmjs.org/fumadocs-mdx/-/fumadocs-mdx-15.1.0.tgz",
-      "integrity": "sha512-2nDusSlYFuNVcyB51jgY3tA3r01ALTwoURrMDNoc7cbJKZ2sac/PW+CDq6SHTArkgRMmFiKYQGfspJdjgTtPTg==",
+      "version": "15.2.0",
+      "resolved": "https://registry.npmjs.org/fumadocs-mdx/-/fumadocs-mdx-15.2.0.tgz",
+      "integrity": "sha512-+yBP8QYw5wA9LF5eVdMhwbP7KT1OF4B/YfC6PZoD2jz0amZi1B+6QHTI6XoRRSTmhWrI4cL5LU1DspW0itk+NA==",
       "license": "MIT",
       "dependencies": {
         "@mdx-js/mdx": "^3.1.1",
@@ -19839,7 +19707,7 @@
         "esbuild": "^0.28.1",
         "estree-util-value-to-estree": "^3.5.0",
         "github-slugger": "^2.0.0",
-        "js-yaml": "^5.2.1",
+        "magic-string": "^0.30.21",
         "mdast-util-mdx": "^3.0.0",
         "picocolors": "^1.1.1",
         "picomatch": "^4.0.5",
@@ -19849,6 +19717,8 @@
         "unist-util-remove-position": "^5.0.0",
         "unist-util-visit": "^5.1.0",
         "vfile": "^6.0.3",
+        "yaml": "^2.9.0",
+        "yuku-analyzer": "^0.6.3",
         "zod": "^4.4.3"
       },
       "bin": {
@@ -19901,26 +19771,26 @@
       }
     },
     "node_modules/fumadocs-ui": {
-      "version": "16.11.1",
-      "resolved": "https://registry.npmjs.org/fumadocs-ui/-/fumadocs-ui-16.11.1.tgz",
-      "integrity": "sha512-Dq819PFV4RGhAI9Wd4erSCiRlEDLVOZae+kgE5LeOKFH8mbKX49U8N17ldFOhdkC9EZpxMZdEKul77RDgFHQww==",
+      "version": "16.11.5",
+      "resolved": "https://registry.npmjs.org/fumadocs-ui/-/fumadocs-ui-16.11.5.tgz",
+      "integrity": "sha512-Eda7x2Hk7E1iIjZ4uES0xxGr25Z72efRM5kP8sbgLSLhWg8TDCyWddvKAkzXIq8bupPOuJkdZa/YVvXbCktIEA==",
       "license": "MIT",
       "dependencies": {
         "@fuma-translate/react": "^1.0.2",
-        "@fumadocs/tailwind": "0.1.0",
-        "@radix-ui/react-accordion": "^1.2.15",
-        "@radix-ui/react-collapsible": "^1.1.15",
-        "@radix-ui/react-dialog": "^1.1.18",
+        "@fumadocs/tailwind": "0.1.1",
+        "@radix-ui/react-accordion": "^1.2.16",
+        "@radix-ui/react-collapsible": "^1.1.16",
+        "@radix-ui/react-dialog": "^1.1.19",
         "@radix-ui/react-direction": "^1.1.2",
-        "@radix-ui/react-navigation-menu": "^1.2.17",
-        "@radix-ui/react-popover": "^1.1.18",
-        "@radix-ui/react-presence": "^1.1.6",
-        "@radix-ui/react-scroll-area": "^1.2.13",
+        "@radix-ui/react-navigation-menu": "^1.2.18",
+        "@radix-ui/react-popover": "^1.1.19",
+        "@radix-ui/react-presence": "^1.1.7",
+        "@radix-ui/react-scroll-area": "^1.2.14",
         "@radix-ui/react-slot": "^1.3.0",
-        "@radix-ui/react-tabs": "^1.1.16",
+        "@radix-ui/react-tabs": "^1.1.17",
         "class-variance-authority": "^0.7.1",
         "cnfast": "^0.0.8",
-        "lucide-react": "^1.23.0",
+        "lucide-react": "^1.24.0",
         "motion": "^12.42.2",
         "next-themes": "^0.4.6",
         "react-remove-scroll": "^2.7.2",
@@ -19930,18 +19800,15 @@
         "unist-util-visit": "^5.1.0"
       },
       "peerDependencies": {
-        "@takumi-rs/image-response": "*",
         "@types/mdx": "*",
         "@types/react": "*",
-        "fumadocs-core": "16.11.1",
+        "fumadocs-core": "16.11.5",
         "next": "16.x.x",
         "react": "^19.2.0",
-        "react-dom": "^19.2.0"
+        "react-dom": "^19.2.0",
+        "takumi-js": "*"
       },
       "peerDependenciesMeta": {
-        "@takumi-rs/image-response": {
-          "optional": true
-        },
         "@types/mdx": {
           "optional": true
         },
@@ -19950,6 +19817,9 @@
         },
         "next": {
           "optional": true
+        },
+        "takumi-js": {
+          "optional": true
         }
       }
     },
@@ -21267,14 +21137,6 @@
         "node": "^22.15.0 || ^24.0.0 || >=26.0.0"
       }
     },
-    "node_modules/http-status-codes": {
-      "version": "2.3.0",
-      "resolved": "https://registry.npmjs.org/http-status-codes/-/http-status-codes-2.3.0.tgz",
-      "integrity": "sha512-RJ8XvFvpPM/Dmc5SV+dC4y5PCeOhT3x1Hq0NU3rjGeg5a/CqlhZ7uudknPwZFz4aeAXDcbAyaeP7GAo9lvngtA==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true
-    },
     "node_modules/http-z": {
       "version": "8.1.1",
       "resolved": "https://registry.npmjs.org/http-z/-/http-z-8.1.1.tgz",
@@ -21753,9 +21615,9 @@
       }
     },
     "node_modules/icu-minify": {
-      "version": "4.13.1",
-      "resolved": "https://registry.npmjs.org/icu-minify/-/icu-minify-4.13.1.tgz",
-      "integrity": "sha512-nFYW2im0WJ3RUVZwabd71J8QTZRtkK1xxZBY7klg7a6KS/os17LZSj9q1VhbRSnk3S8Mv2I7F1izr/aEJ6cUsw==",
+      "version": "4.13.2",
+      "resolved": "https://registry.npmjs.org/icu-minify/-/icu-minify-4.13.2.tgz",
+      "integrity": "sha512-XhYQTEnBXBCyF6ERiwItFweoOXDUciujaGjIWCA7RhOCEPDfhVSTtuwfRq6HdVk7tKnYJh+yaurP96zGnaKsPg==",
       "funding": [
         {
           "type": "individual",
@@ -22669,19 +22531,19 @@
       }
     },
     "node_modules/intl-messageformat": {
-      "version": "11.2.9",
-      "resolved": "https://registry.npmjs.org/intl-messageformat/-/intl-messageformat-11.2.9.tgz",
-      "integrity": "sha512-cGzymZerpDhVXRKjKLgXKda9gI29TU2o88L7gwNMHp3WZVxA/0c5tX52udXbW9JklDApolvMXZG6Dhhdz5eirA==",
+      "version": "11.2.11",
+      "resolved": "https://registry.npmjs.org/intl-messageformat/-/intl-messageformat-11.2.11.tgz",
+      "integrity": "sha512-aDG5bvFRbQvRoT2Bh9FV6yV8t7o0MjEGknZ6pnin5Wt52PJwaBOHDfvz+oPEe78Pl3InQYKugBgCdXijLj6viQ==",
       "license": "BSD-3-Clause",
       "dependencies": {
-        "@formatjs/fast-memoize": "3.1.6",
-        "@formatjs/icu-messageformat-parser": "3.5.12"
+        "@formatjs/fast-memoize": "3.1.7",
+        "@formatjs/icu-messageformat-parser": "3.5.14"
       }
     },
     "node_modules/intl-messageformat/node_modules/@formatjs/fast-memoize": {
-      "version": "3.1.6",
-      "resolved": "https://registry.npmjs.org/@formatjs/fast-memoize/-/fast-memoize-3.1.6.tgz",
-      "integrity": "sha512-H5aexk1Le7T9TPmscacZ+1pR6CTa2n1wq+HDVGXhH8TzUlQQpeXzZs91dRtmFHrbeNbjPFPfQujUqm7MHgVoXQ==",
+      "version": "3.1.7",
+      "resolved": "https://registry.npmjs.org/@formatjs/fast-memoize/-/fast-memoize-3.1.7.tgz",
+      "integrity": "sha512-zXfhLpvA6T7+efdt9JLbBwZ00tT7NsBMDVnDu8rpHeNNv8KfRZAMo2gkG0k9lK/Nzc//3kJ9pImsfuJxk3KhUA==",
       "license": "MIT"
     },
     "node_modules/ioredis": {
@@ -24409,9 +24271,9 @@
       "integrity": "sha512-Ls993zuzfayK269Svk9hzpeGUKob/sIgZzyHYdjQoAdQetRKpOLj+k/QQQ/6Qi0Yz65mlROrfd+Ev+1+7dz9Kw=="
     },
     "node_modules/knip": {
-      "version": "6.25.0",
-      "resolved": "https://registry.npmjs.org/knip/-/knip-6.25.0.tgz",
-      "integrity": "sha512-Q3n41VjOOB/aqsbxb8kallAcFKrUz3b2S5fD5pTODljVpP01t+rvAgy2x3j0Cq8yEpRRHNdar1vHuqFfGuIakQ==",
+      "version": "6.27.0",
+      "resolved": "https://registry.npmjs.org/knip/-/knip-6.27.0.tgz",
+      "integrity": "sha512-CngYEYrD0n20N06FXA8n3u/0Wnnugoa+B9k14OP+iKIgkCHuzvIdsP3nfwjhByoc1WfogpxfiriMboAXFETDUw==",
       "dev": true,
       "funding": [
         {
@@ -25817,9 +25679,9 @@
       }
     },
     "node_modules/lucide-react": {
-      "version": "1.23.0",
-      "resolved": "https://registry.npmjs.org/lucide-react/-/lucide-react-1.23.0.tgz",
-      "integrity": "sha512-38BpJcD0JhFosxHApP/BYsBetLpQFRoTRzEzstM/XCc3jsAG7wqaY1lgVwxiUe3xqYE+lNxo2PkCmYwXWrwwIw==",
+      "version": "1.24.0",
+      "resolved": "https://registry.npmjs.org/lucide-react/-/lucide-react-1.24.0.tgz",
+      "integrity": "sha512-YT6mBD8lGKkg4nM39enlm94/sfJIiW0YKUT60fBy4YK8tai31ylg1VhGNWxkpSKHo9UagfnZqwIff3HTDQwXeA==",
       "license": "ISC",
       "peerDependencies": {
         "react": "^16.5.1 || ^17.0.0 || ^18.0.0 || ^19.0.0"
@@ -25829,7 +25691,6 @@
       "version": "0.30.21",
       "resolved": "https://registry.npmjs.org/magic-string/-/magic-string-0.30.21.tgz",
       "integrity": "sha512-vd2F4YUyEXKGcLHoq+TEyCjxueSeHnFxyyjNp80yg0XV4vUhnDer/lvvlqM/arB5bXQN5K2/3oinyCRyx8T2CQ==",
-      "dev": true,
       "license": "MIT",
       "dependencies": {
         "@jridgewell/sourcemap-codec": "^1.5.5"
@@ -25911,9 +25772,9 @@
       }
     },
     "node_modules/marked": {
-      "version": "18.0.5",
-      "resolved": "https://registry.npmjs.org/marked/-/marked-18.0.5.tgz",
-      "integrity": "sha512-S6GcvALHg6K4ohtu4E7x0a1AqhAjp6cV8KhLSyN9qVapnzJkusVBxZRcIU9AeYsbe6P1hKDusSbEOzGyyuce6w==",
+      "version": "18.0.6",
+      "resolved": "https://registry.npmjs.org/marked/-/marked-18.0.6.tgz",
+      "integrity": "sha512-MrV5puXBfuiy6wl6DLaq3BtIJQAJToAd5zt/ZKhRfGRAuFPALE7/4Y7jnxRQoEgK/pBgurGqLyAuRgZ2xOjr6w==",
       "license": "MIT",
       "bin": {
         "marked": "bin/marked.js"
@@ -25981,9 +25842,9 @@
       }
     },
     "node_modules/material-symbols": {
-      "version": "0.45.6",
-      "resolved": "https://registry.npmjs.org/material-symbols/-/material-symbols-0.45.6.tgz",
-      "integrity": "sha512-sPsLRMFIRETZKVrkwc5VW8lH6FT62vMuTmmQVjftFQ1UJXhK0dQX8isGDDwR72N+rN3oYATC2w9Q5oZB8lczdw==",
+      "version": "0.45.8",
+      "resolved": "https://registry.npmjs.org/material-symbols/-/material-symbols-0.45.8.tgz",
+      "integrity": "sha512-1kpL3jl+/f6W34fmENPYMyftCmBNXUannoZ8XzlMsuEEFlIxesX+Z6ZCj2qSJGxzqwJ5yE/3x1XzVPsycieG2g==",
       "license": "Apache-2.0"
     },
     "node_modules/math-intrinsics": {
@@ -28097,9 +27958,9 @@
       }
     },
     "node_modules/next-intl": {
-      "version": "4.13.1",
-      "resolved": "https://registry.npmjs.org/next-intl/-/next-intl-4.13.1.tgz",
-      "integrity": "sha512-aS8KTA+nNhSNJJBlIhxgvU135WzoObwzFwav4wTDti/Gmhxqe0fs/Q343igo8Z7HGqPB/xgmoagwySZAlHmIfA==",
+      "version": "4.13.2",
+      "resolved": "https://registry.npmjs.org/next-intl/-/next-intl-4.13.2.tgz",
+      "integrity": "sha512-iCYycEP7/PE+1ue4MWBuz1qns6ESA+SGzuXxNMNN1qiYUh2fLhJPiZvHW1zZC7zMTn44aJUglOVZxUb1+hUz6g==",
       "funding": [
         {
           "type": "individual",
@@ -28111,11 +27972,11 @@
         "@formatjs/intl-localematcher": "^0.8.1",
         "@parcel/watcher": "^2.4.1",
         "@swc/core": "^1.15.2",
-        "icu-minify": "^4.13.1",
+        "icu-minify": "^4.13.2",
         "negotiator": "^1.0.0",
-        "next-intl-swc-plugin-extractor": "^4.13.1",
+        "next-intl-swc-plugin-extractor": "^4.13.2",
         "po-parser": "^2.1.1",
-        "use-intl": "^4.13.1"
+        "use-intl": "^4.13.2"
       },
       "peerDependencies": {
         "next": "^12.0.0 || ^13.0.0 || ^14.0.0 || ^15.0.0 || ^16.0.0",
@@ -28128,9 +27989,9 @@
       }
     },
     "node_modules/next-intl-swc-plugin-extractor": {
-      "version": "4.13.1",
-      "resolved": "https://registry.npmjs.org/next-intl-swc-plugin-extractor/-/next-intl-swc-plugin-extractor-4.13.1.tgz",
-      "integrity": "sha512-RhlH2DR1ViEXzcX7G3tDXAvzNrBL2Ph54Hq/q/9oP9eXQV/okh3UQpA/lx2k9U5Ck83CZOSgH0eFTNi5U+zyXw==",
+      "version": "4.13.2",
+      "resolved": "https://registry.npmjs.org/next-intl-swc-plugin-extractor/-/next-intl-swc-plugin-extractor-4.13.2.tgz",
+      "integrity": "sha512-O30N/Y4ifzRe5Sz80jD1Qkg4VY6Zfef4SbHNNE166QkHswBf3/Kygpdd1X52sUUwTCk6uvdisO8ybfAN/VYJHQ==",
       "license": "MIT"
     },
     "node_modules/next-themes": {
@@ -28892,9 +28753,9 @@
       "license": "MIT"
     },
     "node_modules/omniglyph": {
-      "version": "1.0.2",
-      "resolved": "https://registry.npmjs.org/omniglyph/-/omniglyph-1.0.2.tgz",
-      "integrity": "sha512-GGLet99n3HVxOx3WuNPda4B0ETptX9SA8h1fnm/AYSXmvsKXE3mN11Ae2jfX8ldAOA/rVJRXHzPhNkXsRHzMsg==",
+      "version": "1.3.1",
+      "resolved": "https://registry.npmjs.org/omniglyph/-/omniglyph-1.3.1.tgz",
+      "integrity": "sha512-6QnZCoXYczjsPN2x+XpbimimjO6kCoSZUzsdSvoKjtw28U1U724VgLICBNaLX4FFs5jd7SrYNNs9Aee2iIkcoA==",
       "license": "MIT",
       "dependencies": {
         "gpt-tokenizer": "^3.4.0"
@@ -29093,25 +28954,6 @@
         }
       }
     },
-    "node_modules/openapi-fetch": {
-      "version": "0.8.2",
-      "resolved": "https://registry.npmjs.org/openapi-fetch/-/openapi-fetch-0.8.2.tgz",
-      "integrity": "sha512-4g+NLK8FmQ51RW6zLcCBOVy/lwYmFJiiT+ckYZxJWxUxH4XFhsNcX2eeqVMfVOi+mDNFja6qDXIZAz2c5J/RVw==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "openapi-typescript-helpers": "^0.0.5"
-      }
-    },
-    "node_modules/openapi-typescript-helpers": {
-      "version": "0.0.5",
-      "resolved": "https://registry.npmjs.org/openapi-typescript-helpers/-/openapi-typescript-helpers-0.0.5.tgz",
-      "integrity": "sha512-MRffg93t0hgGZbYTxg60hkRIK2sRuEOHEtCUgMuLgbCC33TMQ68AmxskzUlauzZYD47+ENeGV/ElI7qnWqrAxA==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true
-    },
     "node_modules/opener": {
       "version": "1.5.2",
       "resolved": "https://registry.npmjs.org/opener/-/opener-1.5.2.tgz",
@@ -29366,21 +29208,6 @@
         "url": "https://github.com/sponsors/sindresorhus"
       }
     },
-    "node_modules/p-queue-compat": {
-      "version": "1.0.225",
-      "resolved": "https://registry.npmjs.org/p-queue-compat/-/p-queue-compat-1.0.225.tgz",
-      "integrity": "sha512-SdfGSQSJJpD7ZR+dJEjjn9GuuBizHPLW/yarJpXnmrHRruzrq7YM8OqsikSrKeoPv+Pi1YXw9IIBSIg5WveQHA==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "eventemitter3": "5.x",
-        "p-timeout-compat": "^1.0.3"
-      },
-      "engines": {
-        "node": ">=12"
-      }
-    },
     "node_modules/p-queue/node_modules/eventemitter3": {
       "version": "4.0.7",
       "resolved": "https://registry.npmjs.org/eventemitter3/-/eventemitter3-4.0.7.tgz",
@@ -29429,17 +29256,6 @@
         "node": ">=8"
       }
     },
-    "node_modules/p-timeout-compat": {
-      "version": "1.0.8",
-      "resolved": "https://registry.npmjs.org/p-timeout-compat/-/p-timeout-compat-1.0.8.tgz",
-      "integrity": "sha512-+7LpKr1ilnWU0LbV2r+Wz4srwMcFTUysmgL824ZxJcZP3u4Hyi/D/39pbyEs4j0XXCHvbv069+LDPxlCijfVRQ==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "engines": {
-        "node": ">=12"
-      }
-    },
     "node_modules/pac-proxy-agent": {
       "version": "9.1.0",
       "resolved": "https://registry.npmjs.org/pac-proxy-agent/-/pac-proxy-agent-9.1.0.tgz",
@@ -30393,9 +30209,9 @@
       }
     },
     "node_modules/prettier": {
-      "version": "3.9.4",
-      "resolved": "https://registry.npmjs.org/prettier/-/prettier-3.9.4.tgz",
-      "integrity": "sha512-yWG/o/4oJfo036EKAfK6ACAoDOfHeRHx4tuxkfBZiauURiaSmYwlpOr5LQqKtIkRD2z1PLteme2WoxEnj4tHTg==",
+      "version": "3.9.5",
+      "resolved": "https://registry.npmjs.org/prettier/-/prettier-3.9.5.tgz",
+      "integrity": "sha512-/FVl766LpUfB5vXgCYOYa0MeV/441Ia99AeICQIQFTY/Nw0roZwULcXpku5i1/m5kt/baz+s4Zogspd839HSMg==",
       "dev": true,
       "license": "MIT",
       "bin": {
@@ -30540,9 +30356,9 @@
       }
     },
     "node_modules/promptfoo": {
-      "version": "0.121.18",
-      "resolved": "https://registry.npmjs.org/promptfoo/-/promptfoo-0.121.18.tgz",
-      "integrity": "sha512-avytaJ3Vi043Cp/LHRNstKK7PzaDso5QvPa1llMAsISfG8uC7w3mKATGlLcO8Qo6SIhmibfz2JUQvG0EuFuJ0g==",
+      "version": "0.121.19",
+      "resolved": "https://registry.npmjs.org/promptfoo/-/promptfoo-0.121.19.tgz",
+      "integrity": "sha512-5YebsCED/bmR9JktH9YNU62Tr1m3ncFMlM2tKrguI8vFFUfvqxhNzUBa3Z6huG7OvDKbi69UpamU4CLtYLDezQ==",
       "dev": true,
       "license": "MIT",
       "workspaces": [
@@ -30550,7 +30366,7 @@
         "site"
       ],
       "dependencies": {
-        "@anthropic-ai/sdk": "0.106.0",
+        "@anthropic-ai/sdk": "0.110.0",
         "@apidevtools/json-schema-ref-parser": "^15.3.1",
         "@inquirer/checkbox": "^5.1.0",
         "@inquirer/confirm": "^6.0.8",
@@ -30561,8 +30377,8 @@
         "@inquirer/select": "^5.1.0",
         "@libsql/client": "^0.17.3",
         "@opentelemetry/api": "^1.9.0",
-        "@opentelemetry/core": "2.8.0",
-        "@opentelemetry/exporter-trace-otlp-http": "^0.219.0",
+        "@opentelemetry/core": "2.9.0",
+        "@opentelemetry/exporter-trace-otlp-http": "^0.220.0",
         "@opentelemetry/resources": "^2.6.0",
         "@opentelemetry/sdk-trace-base": "^2.6.0",
         "@opentelemetry/sdk-trace-node": "^2.6.0",
@@ -30599,7 +30415,7 @@
         "http-z": "^8.1.1",
         "istextorbinary": "^9.5.0",
         "js-rouge": "^3.2.0",
-        "js-yaml": "5.2.0",
+        "js-yaml": "5.2.1",
         "json5": "^2.2.3",
         "keyv": "^5.6.0",
         "keyv-file": "^5.3.3",
@@ -30638,7 +30454,7 @@
         "node": "^20.20.0 || >=22.22.0"
       },
       "optionalDependencies": {
-        "@anthropic-ai/claude-agent-sdk": "0.3.195",
+        "@anthropic-ai/claude-agent-sdk": "0.3.201",
         "@aws-sdk/client-bedrock-agent-runtime": "^3.1045.0",
         "@aws-sdk/client-bedrock-runtime": "^3.1045.0",
         "@aws-sdk/client-s3": "^3.1003.0",
@@ -30653,10 +30469,9 @@
         "@googleapis/sheets": "^13.0.1",
         "@huggingface/transformers": "^4.0.0",
         "@ibm-cloud/watsonx-ai": "^1.7.14",
-        "@ibm-generative-ai/node-sdk": "^3.2.4",
         "@modelcontextprotocol/sdk": "^1.29.0",
         "@openai/agents": "^0.11.3",
-        "@openai/codex-sdk": "^0.142.3",
+        "@openai/codex-sdk": "^0.144.0",
         "@opencode-ai/sdk": "^1.14.33",
         "@playwright/browser-chromium": "^1.60.0",
         "@rollup/rollup-linux-x64-gnu": "^4.62.0",
@@ -30680,7 +30495,7 @@
         "playwright": "^1.60.0",
         "playwright-extra": "^4.3.6",
         "read-excel-file": "^9.0.0",
-        "sharp": "^0.35.1"
+        "sharp": "^0.35.3"
       }
     },
     "node_modules/promptfoo/node_modules/@huggingface/jinja": {
@@ -30881,29 +30696,6 @@
         "url": "https://github.com/sponsors/sindresorhus"
       }
     },
-    "node_modules/promptfoo/node_modules/js-yaml": {
-      "version": "5.2.0",
-      "resolved": "https://registry.npmjs.org/js-yaml/-/js-yaml-5.2.0.tgz",
-      "integrity": "sha512-YeLUMlvR4Ou1B119LIaM0r65JvbOBooJDc9yEu0dClb/uSC5P4FrLU8OCCz/HXWvtPoIrR0dRzABTjo1sTN9Bw==",
-      "dev": true,
-      "funding": [
-        {
-          "type": "github",
-          "url": "https://github.com/sponsors/puzrin"
-        },
-        {
-          "type": "github",
-          "url": "https://github.com/sponsors/nodeca"
-        }
-      ],
-      "license": "MIT",
-      "dependencies": {
-        "argparse": "^2.0.1"
-      },
-      "bin": {
-        "js-yaml": "bin/js-yaml.mjs"
-      }
-    },
     "node_modules/promptfoo/node_modules/keyv": {
       "version": "5.6.0",
       "resolved": "https://registry.npmjs.org/keyv/-/keyv-5.6.0.tgz",
@@ -33988,7 +33780,6 @@
       "version": "1.6.1",
       "resolved": "https://registry.npmjs.org/smol-toml/-/smol-toml-1.6.1.tgz",
       "integrity": "sha512-dWUG8F5sIIARXih1DTaQAX4SsiTXhInKf1buxdY9DIg4ZYPZK5nGM1VRIYmEbDbsHt7USo99xSLFu5Q1IqTmsg==",
-      "dev": true,
       "license": "BSD-3-Clause",
       "engines": {
         "node": ">= 18"
@@ -35622,9 +35413,9 @@
       "license": "0BSD"
     },
     "node_modules/tsx": {
-      "version": "4.23.0",
-      "resolved": "https://registry.npmjs.org/tsx/-/tsx-4.23.0.tgz",
-      "integrity": "sha512-eUdUIaCr963q2h5u3+QwvYp0+eqPvn+egeqZUm0hwERCqqx1E3kK5ehbGCvqSE5MQAULr67ww0cA3jKc3YkM1w==",
+      "version": "4.23.1",
+      "resolved": "https://registry.npmjs.org/tsx/-/tsx-4.23.1.tgz",
+      "integrity": "sha512-GQHnkIfxyx1wYCOS/wonik5MVRZU9hi1TEZmzGZSCJB1y9YgoZ8H6itNE/u4suE+yLmOzuE4E5S4TZ/ZX2wcWQ==",
       "license": "MIT",
       "dependencies": {
         "esbuild": "~0.28.0"
@@ -35956,16 +35747,16 @@
       }
     },
     "node_modules/typescript-eslint": {
-      "version": "8.63.0",
-      "resolved": "https://registry.npmjs.org/typescript-eslint/-/typescript-eslint-8.63.0.tgz",
-      "integrity": "sha512-xgwXyzG4sK9ALkBxbyGkTMMOS+imnW65iPhxCQMK83KhxyoDNW7l+IDqEf9vMdoUidHpOoS967RCq4eMiTexwQ==",
+      "version": "8.64.0",
+      "resolved": "https://registry.npmjs.org/typescript-eslint/-/typescript-eslint-8.64.0.tgz",
+      "integrity": "sha512-0qg+pDNMnqYzqH9AnNK+39tejHvsShUOUUoRUgtnTGE7QuMZhiFDnozq8nHJVq+Wae6NMLKNWLg5WmkcC/ndyQ==",
       "dev": true,
       "license": "MIT",
       "dependencies": {
-        "@typescript-eslint/eslint-plugin": "8.63.0",
-        "@typescript-eslint/parser": "8.63.0",
-        "@typescript-eslint/typescript-estree": "8.63.0",
-        "@typescript-eslint/utils": "8.63.0"
+        "@typescript-eslint/eslint-plugin": "8.64.0",
+        "@typescript-eslint/parser": "8.64.0",
+        "@typescript-eslint/typescript-estree": "8.64.0",
+        "@typescript-eslint/utils": "8.64.0"
       },
       "engines": {
         "node": "^18.18.0 || ^20.9.0 || >=21.1.0"
@@ -36415,9 +36206,9 @@
       }
     },
     "node_modules/use-intl": {
-      "version": "4.13.1",
-      "resolved": "https://registry.npmjs.org/use-intl/-/use-intl-4.13.1.tgz",
-      "integrity": "sha512-UU5C3zAC7yVg3m7rq5C8VF5J5jhAfvS19Wi9bPNCB9xB7jQYBsUcrqfdxs4Mxl9XR3x6BDB5K++iAw7/rcm3gg==",
+      "version": "4.13.2",
+      "resolved": "https://registry.npmjs.org/use-intl/-/use-intl-4.13.2.tgz",
+      "integrity": "sha512-p6/gCromeBoec+wEuOIkPaytH77RBjW94KWv8MRsNrbI4UOx8QgwNhgl9lbPOKM/b0dpanLhCYztLEH5yQUjtg==",
       "funding": [
         {
           "type": "individual",
@@ -36428,7 +36219,7 @@
       "dependencies": {
         "@formatjs/fast-memoize": "^3.1.0",
         "@schummar/icu-type-parser": "1.21.5",
-        "icu-minify": "^4.13.1",
+        "icu-minify": "^4.13.2",
         "intl-messageformat": "^11.1.0"
       },
       "peerDependencies": {
@@ -37357,9 +37148,9 @@
       }
     },
     "node_modules/ws": {
-      "version": "8.21.0",
-      "resolved": "https://registry.npmjs.org/ws/-/ws-8.21.0.tgz",
-      "integrity": "sha512-Vsp28b7DRcimFQvrqu2Wek3z1iYxDCWqHYB8Qsnk/S4RfaCQzPGPyBNuVjJV3cd6UiKtUtp6sNM77gWvzcCH+g==",
+      "version": "8.21.1",
+      "resolved": "https://registry.npmjs.org/ws/-/ws-8.21.1.tgz",
+      "integrity": "sha512-+0NTnW77fFN/DjQi6k/Sq/Yvk4Sgajw7urW8V+asjXnRgDs9gyGkdb7EzgfhA4goXsRIZKE28fzIXBHEzhuiWw==",
       "license": "MIT",
       "engines": {
         "node": ">=10.0.0"
@@ -37580,7 +37371,6 @@
       "version": "2.9.0",
       "resolved": "https://registry.npmjs.org/yaml/-/yaml-2.9.0.tgz",
       "integrity": "sha512-2AvhNX3mb8zd6Zy7INTtSpl1F15HW6Wnqj0srWlkKLcpYl/gMIMJiyuGq2KeI2YFxUPjdlB+3Lc10seMLtL4cA==",
-      "dev": true,
       "license": "ISC",
       "bin": {
         "yaml": "bin.mjs"
@@ -37733,6 +37523,28 @@
       "integrity": "sha512-0LPOt3AxKqMdFBZA3HBAt/t/8vIKq7VaQYbuA8WxCgung+p9TVyKRYdpvCb80HcdTN2NkbIKbhNwKUfm3tQywQ==",
       "license": "MIT"
     },
+    "node_modules/yuku-analyzer": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmjs.org/yuku-analyzer/-/yuku-analyzer-0.6.3.tgz",
+      "integrity": "sha512-RQ02dPtOa5d2AA3Np45EWD3EJUwZDruCrMMulPwUT/9GK1P7aKAhjaxG4Jv/1qqzMUIt0RUe3Dn2RR6d7+qTrA==",
+      "license": "MIT",
+      "dependencies": {
+        "@yuku-toolchain/types": "0.5.43"
+      },
+      "optionalDependencies": {
+        "@yuku-analyzer/binding-darwin-arm64": "0.6.3",
+        "@yuku-analyzer/binding-darwin-x64": "0.6.3",
+        "@yuku-analyzer/binding-freebsd-x64": "0.6.3",
+        "@yuku-analyzer/binding-linux-arm-gnu": "0.6.3",
+        "@yuku-analyzer/binding-linux-arm-musl": "0.6.3",
+        "@yuku-analyzer/binding-linux-arm64-gnu": "0.6.3",
+        "@yuku-analyzer/binding-linux-arm64-musl": "0.6.3",
+        "@yuku-analyzer/binding-linux-x64-gnu": "0.6.3",
+        "@yuku-analyzer/binding-linux-x64-musl": "0.6.3",
+        "@yuku-analyzer/binding-win32-arm64": "0.6.3",
+        "@yuku-analyzer/binding-win32-x64": "0.6.3"
+      }
+    },
     "node_modules/zod": {
       "version": "4.4.3",
       "resolved": "https://registry.npmjs.org/zod/-/zod-4.4.3.tgz",
@@ -37808,7 +37620,8 @@
       "version": "3.8.49",
       "dependencies": {
         "@toon-format/toon": "^2.3.0",
-        "safe-regex": "^2.1.1"
+        "safe-regex": "^2.1.1",
+        "smol-toml": "1.6.1"
       }
     }
   }
diff --git a/package.json b/package.json
index 937d70363d2..fa272ed3d6d 100644
--- a/package.json
+++ b/package.json
@@ -49,7 +49,7 @@
     "open-sse"
   ],
   "engines": {
-    "node": ">=22.0.0 <23 || >=24.0.0 <27"
+    "node": ">=22.22.2 <23 || >=24.0.0 <27"
   },
   "keywords": [
     "ai",
@@ -79,6 +79,11 @@
     "gen:provider-reference": "bun scripts/docs/gen-provider-reference.ts",
     "bench:compression": "bun scripts/compression/benchmark.ts",
     "eval:compression": "node --import tsx scripts/compression-eval/index.ts",
+    "eval:router": "node --import tsx scripts/router-eval/index.ts",
+    "eval:router:compare": "node --import tsx scripts/router-eval/compare.ts",
+    "eval:router:patch-compare": "node --import tsx scripts/router-eval/patch-compare.ts",
+    "eval:router:search": "node --import tsx scripts/router-eval/search.ts",
+    "eval:router:trends": "node --import tsx scripts/router-eval/trends.ts",
     "release:sync-changelog-i18n": "node scripts/release/sync-changelog-i18n.mjs",
     "build": "node scripts/build/build-next-isolated.mjs",
     "build:secure": "OMNIROUTE_BUILD_PROFILE=minimal node scripts/build/build-next-isolated.mjs",
@@ -118,6 +123,7 @@
     "check:docs-counts": "node scripts/check/check-docs-counts-sync.mjs",
     "check:deprecated-versions": "node scripts/check/check-deprecated-versions.mjs",
     "check:compression-budget": "bun scripts/check/check-compression-budget.ts",
+    "check:router-eval": "node --import tsx scripts/check/check-router-eval-regression.ts",
     "check:doc-links": "node scripts/check/check-doc-links.mjs",
     "check:fabricated-docs": "node scripts/check/check-fabricated-docs.mjs --strict",
     "check:docs-all": "npm run check:docs-sync && npm run check:docs-counts && npm run check:env-doc-sync && npm run check:deprecated-versions && npm run check:doc-links && npm run check:fabricated-docs",
@@ -137,6 +143,7 @@
     "check:openapi-security-tiers": "node scripts/check/check-openapi-security-tiers.mjs",
     "check:provider-consistency": "bun scripts/check/check-provider-consistency.ts",
     "check:provider-assets": "node scripts/check/check-provider-assets.mjs",
+    "check:nvidia-catalog-drift": "node --import tsx/esm scripts/check/check-nvidia-catalog-drift.ts",
     "check:fetch-targets": "node scripts/check/check-fetch-targets.mjs",
     "check:openapi-routes": "node scripts/check/check-openapi-routes.mjs",
     "check:api-docs-refs": "node scripts/check/check-api-docs-refs.mjs",
@@ -285,6 +292,7 @@
     "recharts": "^3.8.1",
     "safe-regex": "^2.1.1",
     "selfsigned": "^5.5.0",
+    "smol-toml": "1.6.1",
     "socks": "^2.8.7",
     "sql.js": "^1.14.1",
     "sqlite-vec": "^0.1.9",
diff --git a/scripts/check/check-nvidia-catalog-drift.ts b/scripts/check/check-nvidia-catalog-drift.ts
new file mode 100644
index 00000000000..bbd279e56cd
--- /dev/null
+++ b/scripts/check/check-nvidia-catalog-drift.ts
@@ -0,0 +1,108 @@
+import { FREE_MODEL_BUDGETS } from "../../open-sse/config/freeModelCatalog.data.ts";
+import reviewedLiveIds from "../../open-sse/config/nvidiaHostedModels.snapshot.json" with { type: "json" };
+
+const NVIDIA_MODELS_URL = "https://integrate.api.nvidia.com/v1/models";
+
+export interface NvidiaCatalogDrift {
+  liveCount: number;
+  reviewedLiveCount: number;
+  documentedFreeCount: number;
+  newLiveIds: string[];
+  removedLiveIds: string[];
+  documentedMissingUpstreamIds: string[];
+}
+
+function normalizeIds(ids: Iterable<string>): Set<string> {
+  return new Set(
+    [...ids].filter((id) => typeof id === "string" && id.trim().length > 0).map((id) => id.trim())
+  );
+}
+
+export function computeNvidiaCatalogDrift(
+  liveIdsInput: Iterable<string>,
+  reviewedLiveIdsInput: Iterable<string>,
+  documentedFreeIdsInput: Iterable<string>
+): NvidiaCatalogDrift {
+  const liveIds = normalizeIds(liveIdsInput);
+  const reviewedIds = normalizeIds(reviewedLiveIdsInput);
+  const documentedFreeIds = normalizeIds(documentedFreeIdsInput);
+  return {
+    liveCount: liveIds.size,
+    reviewedLiveCount: reviewedIds.size,
+    documentedFreeCount: documentedFreeIds.size,
+    newLiveIds: [...liveIds].filter((id) => !reviewedIds.has(id)).sort(),
+    removedLiveIds: [...reviewedIds].filter((id) => !liveIds.has(id)).sort(),
+    documentedMissingUpstreamIds: [...documentedFreeIds].filter((id) => !liveIds.has(id)).sort(),
+  };
+}
+
+function printIds(label: string, ids: string[]): void {
+  console.log(`${label} (${ids.length})`);
+  for (const id of ids) console.log(`  - ${id}`);
+}
+
+async function main(): Promise<void> {
+  const apiKey = process.env.NVIDIA_API_KEY?.trim();
+  if (!apiKey) {
+    console.error("NVIDIA_API_KEY is required to query the live NVIDIA NIM model catalog.");
+    process.exitCode = 2;
+    return;
+  }
+
+  let liveIds: string[] = [];
+  try {
+    const response = await fetch(NVIDIA_MODELS_URL, {
+      headers: { Authorization: `Bearer ${apiKey}`, Accept: "application/json" },
+      signal: AbortSignal.timeout(30_000),
+    });
+    if (!response.ok) {
+      console.error(`NVIDIA model catalog returned HTTP ${response.status}.`);
+      process.exitCode = 2;
+      return;
+    }
+
+    const body = (await response.json()) as { data?: Array<{ id?: unknown }> } | null;
+    liveIds = (body && Array.isArray(body.data) ? body.data : [])
+      .map((model) => model?.id)
+      .filter((id): id is string => typeof id === "string");
+  } catch (error) {
+    console.error(
+      "Failed to fetch or parse NVIDIA model catalog:",
+      error instanceof Error ? error.message : error
+    );
+    process.exitCode = 2;
+    return;
+  }
+  const documentedFreeIds = FREE_MODEL_BUDGETS.filter((model) => model.provider === "nvidia").map(
+    (model) => model.modelId
+  );
+  const drift = computeNvidiaCatalogDrift(liveIds, reviewedLiveIds, documentedFreeIds);
+
+  console.log(
+    `NVIDIA catalog: ${drift.liveCount} live model(s), ${drift.reviewedLiveCount} reviewed live model(s), ${drift.documentedFreeCount} documented free/trial model(s).`
+  );
+  printIds("New live models requiring metadata review", drift.newLiveIds);
+  printIds("Reviewed models removed from the live catalog", drift.removedLiveIds);
+  printIds(
+    "Documented free models missing from the live catalog",
+    drift.documentedMissingUpstreamIds
+  );
+
+  if (
+    drift.newLiveIds.length ||
+    drift.removedLiveIds.length ||
+    drift.documentedMissingUpstreamIds.length
+  ) {
+    console.warn(
+      "Catalog drift requires review. Availability alone does not prove that a model is free; verify NVIDIA's model page before updating FREE_MODEL_BUDGETS."
+    );
+    if (process.argv.includes("--strict")) process.exitCode = 1;
+  } else {
+    console.log("No NVIDIA catalog drift detected.");
+  }
+}
+
+const isDirectExecution = process.argv[1]?.endsWith("check-nvidia-catalog-drift.ts");
+if (isDirectExecution) {
+  await main();
+}
diff --git a/scripts/check/check-router-eval-regression.ts b/scripts/check/check-router-eval-regression.ts
new file mode 100644
index 00000000000..46d696d6b8e
--- /dev/null
+++ b/scripts/check/check-router-eval-regression.ts
@@ -0,0 +1,287 @@
+#!/usr/bin/env node
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import { spawnSync } from "node:child_process";
+import { fileURLToPath } from "node:url";
+
+type Args = {
+  baseline: string;
+  candidate: string;
+  baselinePatch?: string;
+  candidatePatch?: string;
+  output: string;
+  jsonOutput: string;
+  patchOutput: string;
+  patchJsonOutput: string;
+  artifactDir?: string;
+  runId: string;
+  maxAiqDrop: string;
+  maxCostIncrease: string;
+  maxPatchAiqDrop: string;
+  maxPatchCostIncrease: string;
+  maxPatchLatencyIncrease: string;
+  maxPatchRegressionIncrease: string;
+};
+
+type GateManifest = {
+  schemaVersion: 1;
+  kind: "router-eval-gate-run";
+  runId: string;
+  generatedAt: string;
+  command: string[];
+  thresholds: {
+    maxAiqDrop: number;
+    maxCostIncrease: number;
+    patch?: {
+      maxAiqDrop: number;
+      maxCostIncrease: number;
+      maxLatencyIncrease: number;
+      maxRegressionIncrease: number;
+    };
+  };
+  inputs: {
+    baseline: string;
+    candidate: string;
+    baselinePatch?: string;
+    candidatePatch?: string;
+  };
+  outputs: {
+    markdown: string;
+    json: string;
+    patchMarkdown?: string;
+    patchJson?: string;
+  };
+  environment: {
+    runtime: "bun" | "node";
+    platform: NodeJS.Platform;
+  };
+  result: {
+    status: number;
+  };
+};
+
+const repoRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../..");
+const isBunRuntime = "Bun" in globalThis;
+
+function runTypeScriptScript(args: string[]) {
+  return spawnSync(process.execPath, isBunRuntime ? args : ["--import", "tsx", ...args], {
+    cwd: repoRoot,
+    encoding: "utf8",
+    stdio: ["ignore", "pipe", "pipe"],
+  });
+}
+const defaultFixtureDir = path.join(repoRoot, "tests/fixtures/router-eval");
+const defaultArtifactDir = path.join(os.tmpdir(), "omniroute-router-eval");
+
+function getArgValue(name: string): string | undefined {
+  const index = process.argv.indexOf(`--${name}`);
+  if (index < 0) return undefined;
+  const value = process.argv[index + 1];
+  return value && !value.startsWith("--") ? value : undefined;
+}
+
+function readArgs(): Args {
+  const artifactDir = getArgValue("artifact-dir");
+  const runId = getArgValue("run-id") ?? new Date().toISOString().replace(/[:.]/g, "-");
+  const retainedDir = artifactDir ? path.join(path.resolve(artifactDir), runId) : undefined;
+  return {
+    baseline: getArgValue("baseline") ?? path.join(defaultFixtureDir, "baseline.ndjson"),
+    candidate: getArgValue("candidate") ?? path.join(defaultFixtureDir, "candidate.ndjson"),
+    baselinePatch: getArgValue("baseline-patch"),
+    candidatePatch: getArgValue("candidate-patch"),
+    output: getArgValue("output") ?? path.join(retainedDir ?? defaultArtifactDir, "router-eval.md"),
+    jsonOutput:
+      getArgValue("json-output") ??
+      path.join(retainedDir ?? defaultArtifactDir, "router-eval.json"),
+    patchOutput:
+      getArgValue("patch-output") ??
+      path.join(retainedDir ?? defaultArtifactDir, "patch-comparison.md"),
+    patchJsonOutput:
+      getArgValue("patch-json-output") ??
+      path.join(retainedDir ?? defaultArtifactDir, "patch-comparison.json"),
+    artifactDir,
+    runId,
+    maxAiqDrop: getArgValue("max-aiq-drop") ?? "1",
+    maxCostIncrease: getArgValue("max-cost-increase") ?? "0.05",
+    maxPatchAiqDrop: getArgValue("max-patch-aiq-drop") ?? "1",
+    maxPatchCostIncrease: getArgValue("max-patch-cost-increase") ?? "0.05",
+    maxPatchLatencyIncrease: getArgValue("max-patch-latency-increase") ?? "0.05",
+    maxPatchRegressionIncrease: getArgValue("max-patch-regression-increase") ?? "0",
+  };
+}
+
+function ensureReadable(filePath: string, label: string): void {
+  if (!fs.existsSync(filePath)) {
+    console.error(`[router-eval] ${label} missing: ${filePath}`);
+    process.exit(2);
+  }
+}
+
+function writeRetainedRun(args: Args, status: number): void {
+  if (!args.artifactDir) return;
+
+  const runDir = path.join(path.resolve(args.artifactDir), args.runId);
+  const inputDir = path.join(runDir, "inputs");
+  fs.mkdirSync(inputDir, { recursive: true });
+
+  const baselineCopy = path.join(inputDir, "baseline.ndjson");
+  const candidateCopy = path.join(inputDir, "candidate.ndjson");
+  fs.copyFileSync(args.baseline, baselineCopy);
+  fs.copyFileSync(args.candidate, candidateCopy);
+  const baselinePatchCopy = args.baselinePatch
+    ? path.join(inputDir, "baseline.patch.json")
+    : undefined;
+  const candidatePatchCopy = args.candidatePatch
+    ? path.join(inputDir, "candidate.patch.json")
+    : undefined;
+  if (args.baselinePatch && baselinePatchCopy)
+    fs.copyFileSync(args.baselinePatch, baselinePatchCopy);
+  if (args.candidatePatch && candidatePatchCopy)
+    fs.copyFileSync(args.candidatePatch, candidatePatchCopy);
+
+  const manifest: GateManifest = {
+    schemaVersion: 1,
+    kind: "router-eval-gate-run",
+    runId: args.runId,
+    generatedAt: new Date().toISOString(),
+    command: process.argv.slice(1),
+    thresholds: {
+      maxAiqDrop: Number.parseFloat(args.maxAiqDrop),
+      maxCostIncrease: Number.parseFloat(args.maxCostIncrease),
+      ...(args.baselinePatch && args.candidatePatch
+        ? {
+            patch: {
+              maxAiqDrop: Number.parseFloat(args.maxPatchAiqDrop),
+              maxCostIncrease: Number.parseFloat(args.maxPatchCostIncrease),
+              maxLatencyIncrease: Number.parseFloat(args.maxPatchLatencyIncrease),
+              maxRegressionIncrease: Number.parseFloat(args.maxPatchRegressionIncrease),
+            },
+          }
+        : {}),
+    },
+    inputs: {
+      baseline: path.relative(runDir, baselineCopy),
+      candidate: path.relative(runDir, candidateCopy),
+      ...(baselinePatchCopy && candidatePatchCopy
+        ? {
+            baselinePatch: path.relative(runDir, baselinePatchCopy),
+            candidatePatch: path.relative(runDir, candidatePatchCopy),
+          }
+        : {}),
+    },
+    outputs: {
+      markdown: path.relative(runDir, args.output),
+      json: path.relative(runDir, args.jsonOutput),
+      ...(args.baselinePatch && args.candidatePatch
+        ? {
+            patchMarkdown: path.relative(runDir, args.patchOutput),
+            patchJson: path.relative(runDir, args.patchJsonOutput),
+          }
+        : {}),
+    },
+    environment: {
+      runtime: isBunRuntime ? "bun" : "node",
+      platform: process.platform,
+    },
+    result: {
+      status,
+    },
+  };
+
+  fs.writeFileSync(path.join(runDir, "manifest.json"), `${JSON.stringify(manifest, null, 2)}\n`);
+}
+
+function runPatchGate(args: Args): number {
+  if (!args.baselinePatch && !args.candidatePatch) return 0;
+  if (!args.baselinePatch || !args.candidatePatch) {
+    console.error("[router-eval] --baseline-patch and --candidate-patch must be provided together");
+    return 2;
+  }
+  ensureReadable(args.baselinePatch, "baseline patch");
+  ensureReadable(args.candidatePatch, "candidate patch");
+  fs.mkdirSync(path.dirname(args.patchOutput), { recursive: true });
+  fs.mkdirSync(path.dirname(args.patchJsonOutput), { recursive: true });
+
+  const result = runTypeScriptScript([
+    "scripts/router-eval/patch-compare.ts",
+    "--baseline",
+    args.baselinePatch,
+    "--candidate",
+    args.candidatePatch,
+    "--output",
+    args.patchOutput,
+    "--json-output",
+    args.patchJsonOutput,
+    "--run-id",
+    args.runId,
+    "--max-aiq-drop",
+    args.maxPatchAiqDrop,
+    "--max-cost-increase",
+    args.maxPatchCostIncrease,
+    "--max-latency-increase",
+    args.maxPatchLatencyIncrease,
+    "--max-regression-increase",
+    args.maxPatchRegressionIncrease,
+    "--fail-on-regression",
+  ]);
+
+  if (result.error) {
+    console.error(`[router-eval] failed to launch patch compare: ${result.error.message}`);
+    return 1;
+  }
+  if (result.stdout) process.stdout.write(result.stdout);
+  if (result.stderr) process.stderr.write(result.stderr);
+  return result.status ?? 1;
+}
+
+function main(): void {
+  const args = readArgs();
+  ensureReadable(args.baseline, "baseline corpus");
+  ensureReadable(args.candidate, "candidate corpus");
+  fs.mkdirSync(path.dirname(args.output), { recursive: true });
+  fs.mkdirSync(path.dirname(args.jsonOutput), { recursive: true });
+
+  const result = runTypeScriptScript([
+    "scripts/router-eval/index.ts",
+    "--input",
+    args.candidate,
+    "--baseline-input",
+    args.baseline,
+    "--max-aiq-drop",
+    args.maxAiqDrop,
+    "--max-cost-increase",
+    args.maxCostIncrease,
+    "--output",
+    args.output,
+    "--json-output",
+    args.jsonOutput,
+    "--fail-on-regression",
+  ]);
+
+  if (result.error) {
+    console.error(`[router-eval] failed to launch evaluator: ${result.error.message}`);
+    process.exit(1);
+  }
+
+  if (result.stdout) process.stdout.write(result.stdout);
+  if (result.stderr) process.stderr.write(result.stderr);
+
+  if (result.status === 0) {
+    const patchStatus = runPatchGate(args);
+    writeRetainedRun(args, patchStatus);
+    if (patchStatus !== 0) {
+      console.error(`[router-eval] patch gate failed with exit code ${patchStatus}`);
+      process.exit(patchStatus);
+    }
+    const retention = args.artifactDir ? ` retained run ${args.runId}` : " temp run";
+    console.log(`[router-eval] OK -${retention}; artifacts: ${args.output}, ${args.jsonOutput}`);
+    return;
+  }
+
+  writeRetainedRun(args, result.status ?? 1);
+  console.error(`[router-eval] regression gate failed with exit code ${result.status ?? 1}`);
+  process.exit(result.status ?? 1);
+}
+
+main();
diff --git a/scripts/router-eval/compare.ts b/scripts/router-eval/compare.ts
new file mode 100644
index 00000000000..31ea7f36771
--- /dev/null
+++ b/scripts/router-eval/compare.ts
@@ -0,0 +1,135 @@
+#!/usr/bin/env node
+import fs from "node:fs";
+import path from "node:path";
+import { spawnSync } from "node:child_process";
+import { fileURLToPath } from "node:url";
+
+type Args = {
+  baseline: string;
+  candidate: string;
+  baselineName: string;
+  candidateName: string;
+  artifactDir: string;
+  runId: string;
+  maxAiqDrop: string;
+  maxCostIncrease: string;
+  failOnRegression: boolean;
+};
+
+const repoRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../..");
+const isBunRuntime = "Bun" in globalThis;
+
+function runTypeScriptScript(args: string[]) {
+  return spawnSync(process.execPath, isBunRuntime ? args : ["--import", "tsx", ...args], {
+    cwd: repoRoot,
+    encoding: "utf8",
+    stdio: ["ignore", "pipe", "pipe"],
+  });
+}
+
+function getArgValue(name: string): string | undefined {
+  const index = process.argv.indexOf(`--${name}`);
+  if (index < 0) return undefined;
+  const value = process.argv[index + 1];
+  return value && !value.startsWith("--") ? value : undefined;
+}
+
+function usage(): string {
+  return [
+    "Usage:",
+    "  npm run eval:router:compare -- --baseline <baseline.ndjson> --candidate <candidate.ndjson>",
+    "       [--baseline-name <name>] [--candidate-name <name>] [--artifact-dir <dir>]",
+    "       [--run-id <id>] [--max-aiq-drop <n>] [--max-cost-increase <n>] [--fail-on-regression]",
+    "",
+    "Runs a named baseline-vs-candidate router-eval comparison and retains artifacts.",
+  ].join("\n");
+}
+
+function requireArg(name: string): string {
+  const value = getArgValue(name);
+  if (!value) {
+    console.error(`Missing required --${name}`);
+    process.exit(2);
+  }
+  return value;
+}
+
+function readArgs(): Args {
+  const baselineName = getArgValue("baseline-name") ?? "baseline";
+  const candidateName = getArgValue("candidate-name") ?? "candidate";
+  const timestamp = new Date().toISOString().replace(/[:.]/g, "-");
+  return {
+    baseline: requireArg("baseline"),
+    candidate: requireArg("candidate"),
+    baselineName,
+    candidateName,
+    artifactDir: getArgValue("artifact-dir") ?? "artifacts/router-eval/comparisons",
+    runId: getArgValue("run-id") ?? `${baselineName}-vs-${candidateName}-${timestamp}`,
+    maxAiqDrop: getArgValue("max-aiq-drop") ?? "1",
+    maxCostIncrease: getArgValue("max-cost-increase") ?? "0.05",
+    failOnRegression: process.argv.includes("--fail-on-regression"),
+  };
+}
+
+function ensureReadable(filePath: string, label: string): void {
+  if (!fs.existsSync(filePath)) {
+    console.error(`[router-eval:compare] ${label} missing: ${filePath}`);
+    process.exit(2);
+  }
+}
+
+function main(): void {
+  if (process.argv.includes("--help") || process.argv.includes("-h")) {
+    console.log(usage());
+    return;
+  }
+
+  const args = readArgs();
+  ensureReadable(args.baseline, "baseline corpus");
+  ensureReadable(args.candidate, "candidate corpus");
+
+  const checkArgs = [
+    "scripts/check/check-router-eval-regression.ts",
+    "--baseline",
+    args.baseline,
+    "--candidate",
+    args.candidate,
+    "--artifact-dir",
+    args.artifactDir,
+    "--run-id",
+    args.runId,
+    "--max-aiq-drop",
+    args.maxAiqDrop,
+    "--max-cost-increase",
+    args.maxCostIncrease,
+  ];
+
+  const result = runTypeScriptScript(checkArgs);
+
+  if (result.error) {
+    console.error(`[router-eval:compare] failed to launch comparison: ${result.error.message}`);
+    process.exit(1);
+  }
+
+  if (result.stdout) process.stdout.write(result.stdout);
+  if (result.stderr) process.stderr.write(result.stderr);
+
+  const runDir = path.resolve(args.artifactDir, args.runId);
+  const labels = {
+    baselineName: args.baselineName,
+    candidateName: args.candidateName,
+    baseline: path.relative(runDir, path.resolve(args.baseline)),
+    candidate: path.relative(runDir, path.resolve(args.candidate)),
+  };
+  fs.mkdirSync(runDir, { recursive: true });
+  fs.writeFileSync(path.join(runDir, "comparison.json"), `${JSON.stringify(labels, null, 2)}\n`);
+
+  if (result.status === 0 || !args.failOnRegression) {
+    console.log(`[router-eval:compare] artifacts: ${runDir}`);
+    return;
+  }
+
+  process.exit(result.status ?? 1);
+}
+
+main();
diff --git a/scripts/router-eval/index.ts b/scripts/router-eval/index.ts
new file mode 100644
index 00000000000..6b0b7312e91
--- /dev/null
+++ b/scripts/router-eval/index.ts
@@ -0,0 +1,483 @@
+#!/usr/bin/env node
+import fs from "node:fs";
+import path from "node:path";
+
+import {
+  compareRouterEvalRuns,
+  createRouterEvalArtifact,
+  formatRouterEvalComparison,
+  formatRouterEvalReport,
+  runRouterEval,
+  toRouterObservation,
+  type RouterEvalArtifact,
+  type RouterEvalArtifactMetadata,
+  type RouterObservation,
+} from "@/lib/routerEval/index.ts";
+import { SQLITE_FILE } from "@/lib/db/core.ts";
+
+type DbCallLogRow = {
+  id: string;
+  model: string | null;
+  requested_model: string | null;
+  duration: number | null;
+  tokens_in: number | null;
+  tokens_out: number | null;
+  status: number | null;
+  combo_name: string | null;
+  provider: string | null;
+  error_summary: string | null;
+  timestamp: string | null;
+  correlation_id: string | null;
+};
+
+type DbUsageHistoryRow = {
+  id: number;
+  provider: string | null;
+  model: string | null;
+  tokens_input: number | null;
+  tokens_output: number | null;
+  service_tier: string | null;
+  status: string | null;
+  success: number | null;
+  latency_ms: number | null;
+  error_code: string | null;
+  combo_strategy: string | null;
+  timestamp: string | null;
+};
+
+type DbReplaySource = "auto" | "call-logs" | "usage-history";
+
+type SqliteStatement = {
+  get: (...params: unknown[]) => unknown;
+  all: (...params: unknown[]) => unknown[];
+};
+
+type SqliteDatabase = {
+  prepare: (sql: string) => SqliteStatement;
+  close: () => void;
+};
+
+type ArgSpec = {
+  input?: string;
+  db?: string;
+  dbSource?: DbReplaySource;
+  baselineInput?: string;
+  baselineDb?: string;
+  baselineDbSource?: DbReplaySource;
+  since?: string;
+  limit?: number;
+  aiqDrop?: number;
+  costIncrease?: number;
+  output?: string;
+  jsonOutput?: string;
+  exportCorpus?: string;
+  failOnRegression?: boolean;
+  help?: boolean;
+};
+
+function getArgValue(name: string): string | undefined {
+  const index = process.argv.indexOf(`--${name}`);
+  if (index < 0) return undefined;
+  const value = process.argv[index + 1];
+  if (!value || value.startsWith("--")) return undefined;
+  return value;
+}
+
+function getNumericArg(name: string): number | undefined {
+  const value = getArgValue(name);
+  if (!value) return undefined;
+  const parsed = Number.parseInt(value, 10);
+  return Number.isFinite(parsed) && parsed >= 0 ? parsed : undefined;
+}
+
+function getFloatArg(name: string): number | undefined {
+  const value = getArgValue(name);
+  if (!value) return undefined;
+  const parsed = Number.parseFloat(value);
+  return Number.isFinite(parsed) ? parsed : undefined;
+}
+
+function parseArgs(): ArgSpec {
+  return {
+    input: getArgValue("input"),
+    db: getArgValue("db"),
+    dbSource: parseReplaySource(getArgValue("db-source")),
+    baselineInput: getArgValue("baseline-input"),
+    baselineDb: getArgValue("baseline-db"),
+    baselineDbSource: parseReplaySource(getArgValue("baseline-db-source")),
+    since: getArgValue("since"),
+    limit: getNumericArg("limit"),
+    aiqDrop: getFloatArg("max-aiq-drop"),
+    costIncrease: getFloatArg("max-cost-increase"),
+    output: getArgValue("output"),
+    jsonOutput: getArgValue("json-output"),
+    exportCorpus: getArgValue("export-corpus"),
+    failOnRegression: process.argv.includes("--fail-on-regression"),
+    help: process.argv.includes("--help") || process.argv.includes("-h"),
+  };
+}
+
+function usage() {
+  return [
+    "Usage:",
+    "  npm run eval:router -- --input <file.ndjson> [--since <iso>] [--limit <n>]",
+    "  npm run eval:router -- --db [path] [--db-source usage-history|call-logs|auto] [--since <iso>] [--limit <n>]",
+    "  npm run eval:router -- --input <candidate> --baseline-input <baseline>",
+    "  npm run eval:router -- --db <path> --db-source usage-history",
+    "       [--max-aiq-drop <n>] [--max-cost-increase <n>] [--fail-on-regression]",
+    "",
+    "Options:",
+    "  --input <path>           JSONL observation corpus (or omit for stdin)",
+    "  --db [path]              Read SQLite rows from the routing-replay source",
+    "  --db-source <auto|call-logs|usage-history>  Source for --db reads (default: auto => call-logs then usage-history)",
+    "  --baseline-input <path>   Baseline corpus in JSONL",
+    "  --baseline-db <path>      Baseline corpus in SQLite",
+    "  --baseline-db-source <auto|call-logs|usage-history>  Source for baseline DB reads",
+    "  --since <ISO8601>        Filter rows newer than this value",
+    "  --limit <n>              Limit sample count",
+    "  --max-aiq-drop <n>       Regression threshold (default: 0)",
+    "  --max-cost-increase <n>   Relative increase threshold (default: 0)",
+    "  --output <path>          Write report to file",
+    "  --json-output <path>     Write machine-readable artifact JSON",
+    "  --export-corpus <path>   Write normalized RouterObservation JSONL",
+    "  --fail-on-regression      Exit 1 if candidate regresses vs baseline",
+  ].join("\n");
+}
+
+function parseInputLine(rawLine: string): RouterObservation | null {
+  const trimmed = rawLine.trim();
+  if (!trimmed) return null;
+  try {
+    const parsed = JSON.parse(trimmed);
+    return toRouterObservation(parsed);
+  } catch {
+    return null;
+  }
+}
+
+async function readJsonl(inputPath?: string): Promise<RouterObservation[]> {
+  let text: string;
+  if (!inputPath) {
+    text = await new Response(process.stdin, { duplex: "half" }).text();
+  } else {
+    text = await fs.promises.readFile(path.resolve(inputPath), "utf8");
+  }
+
+  const observations: RouterObservation[] = [];
+  for (const line of text.split(/\r?\n/)) {
+    const parsed = parseInputLine(line);
+    if (parsed) observations.push(parsed);
+  }
+  return observations;
+}
+
+function estimateCost(tokensIn: unknown, tokensOut: unknown): number {
+  const inTokens = typeof tokensIn === "number" ? tokensIn : 0;
+  const outTokens = typeof tokensOut === "number" ? tokensOut : 0;
+  return Number(((inTokens + outTokens) * 0.000001).toFixed(6));
+}
+
+function parseReplaySource(rawSource?: string): DbReplaySource {
+  if (!rawSource) return "auto";
+  const normalized = rawSource.toLowerCase();
+  if (normalized === "auto") return "auto";
+  if (normalized === "usage_history" || normalized === "usage-history") return "usage-history";
+  if (normalized === "call_logs" || normalized === "call-logs") return "call-logs";
+  throw new Error(`Unsupported db source: ${rawSource}`);
+}
+
+function hasReplayTable(database: SqliteDatabase, tableName: string): boolean {
+  return Boolean(
+    database.prepare("SELECT 1 FROM sqlite_master WHERE type='table' AND name=?").get(tableName)
+  );
+}
+
+function resolveReplaySource(
+  database: SqliteDatabase,
+  requestedSource: DbReplaySource
+): DbReplaySource {
+  const hasUsageHistory = hasReplayTable(database, "usage_history");
+  const hasCallLogs = hasReplayTable(database, "call_logs");
+
+  if (requestedSource === "usage-history") {
+    if (!hasUsageHistory) throw new Error("Table 'usage_history' missing in database");
+    return "usage-history";
+  }
+
+  if (requestedSource === "call-logs") {
+    if (!hasCallLogs) throw new Error("Table 'call_logs' missing in database");
+    return "call-logs";
+  }
+
+  if (requestedSource === "auto") {
+    if (hasCallLogs) return "call-logs";
+    if (hasUsageHistory) return "usage-history";
+  }
+
+  throw new Error("No replay table found in database (expected usage_history or call_logs)");
+}
+
+function toSuccessFromStatus(status: unknown): boolean {
+  if (typeof status === "number") return status >= 200 && status < 400;
+  if (typeof status === "string") {
+    const parsed = Number.parseInt(status, 10);
+    if (Number.isFinite(parsed)) return parsed >= 200 && parsed < 400;
+    const normalized = status.trim().toLowerCase();
+    if (normalized === "ok" || normalized === "success" || normalized === "true") return true;
+  }
+  return false;
+}
+
+function readCallLogDb(db: SqliteDatabase, since?: string, limit?: number): RouterObservation[] {
+  const queryParts = [
+    "SELECT id, model, requested_model, duration, tokens_in, tokens_out, status, combo_name, provider, error_summary, timestamp, correlation_id",
+    "FROM call_logs",
+    "WHERE 1=1",
+  ];
+  const params: unknown[] = [];
+
+  if (since) {
+    queryParts.push("AND timestamp >= ?");
+    params.push(since);
+  }
+
+  queryParts.push("ORDER BY timestamp ASC");
+  if (limit) {
+    queryParts.push("LIMIT ?");
+    params.push(limit);
+  }
+
+  const rows = db.prepare(queryParts.join(" ")).all(...params) as DbCallLogRow[];
+
+  const observations: RouterObservation[] = [];
+  for (const row of rows) {
+    const mapped = toRouterObservation({
+      sampleId: row.id,
+      model: row.model,
+      requestedModel: row.requested_model,
+      latency: row.duration ?? 0,
+      costUsd: estimateCost(row.tokens_in, row.tokens_out),
+      configId: row.combo_name || row.provider || "default",
+      success: row.status != null && row.status >= 200 && row.status < 400,
+      status: row.status ?? 0,
+      error: row.error_summary,
+      routeInput: {
+        correlationId: row.correlation_id ?? "",
+      },
+      timestamp: row.timestamp ?? new Date().toISOString(),
+    });
+    if (mapped) observations.push(mapped);
+  }
+  return observations;
+}
+
+function readUsageHistoryDb(
+  db: SqliteDatabase,
+  since?: string,
+  limit?: number
+): RouterObservation[] {
+  const queryParts = [
+    "SELECT id, provider, model, tokens_input, tokens_output, service_tier, status, success, latency_ms, error_code, combo_strategy, timestamp",
+    "FROM usage_history",
+    "WHERE 1=1",
+  ];
+  const params: unknown[] = [];
+
+  if (since) {
+    queryParts.push("AND timestamp >= ?");
+    params.push(since);
+  }
+
+  queryParts.push("ORDER BY timestamp ASC");
+  if (limit) {
+    queryParts.push("LIMIT ?");
+    params.push(limit);
+  }
+
+  const rows = db.prepare(queryParts.join(" ")).all(...params) as DbUsageHistoryRow[];
+
+  const observations: RouterObservation[] = [];
+  for (const row of rows) {
+    const cost = estimateCost(row.tokens_input, row.tokens_output);
+    const rowId = `${row.id}`;
+    const mapped = toRouterObservation({
+      sampleId: rowId,
+      model: row.model,
+      requestedModel: row.model,
+      latency: row.latency_ms ?? 0,
+      costUsd: cost,
+      configId: row.combo_strategy || row.provider || "default",
+      success: toSuccessFromStatus(row.status) || row.success === 1,
+      status: row.success === 1 ? 200 : 0,
+      routeInput: {},
+      metadata: {
+        provider: row.provider,
+        serviceTier: row.service_tier,
+        errorCode: row.error_code,
+      },
+      timestamp: row.timestamp ?? new Date().toISOString(),
+    });
+    if (mapped) observations.push(mapped);
+  }
+  return observations;
+}
+
+async function openSqliteDatabase(sqliteFile: string): Promise<SqliteDatabase> {
+  if ("Bun" in globalThis) {
+    const sqlite = await import("bun:sqlite");
+    return new sqlite.Database(sqliteFile, { readonly: true });
+  }
+
+  const sqlite = await import("better-sqlite3");
+  return new sqlite.default(sqliteFile, { readonly: true });
+}
+
+async function readDb(
+  filePath: string,
+  since?: string,
+  limit?: number,
+  source: DbReplaySource = "auto"
+): Promise<RouterObservation[]> {
+  const sqliteFile = filePath || SQLITE_FILE;
+  if (!sqliteFile) throw new Error("SQLite mode requires a path or SQLITE_FILE");
+  const db = await openSqliteDatabase(sqliteFile);
+  try {
+    const normalized = parseReplaySource(source);
+    const activeSource = resolveReplaySource(db, normalized);
+    if (activeSource === "usage-history") {
+      return readUsageHistoryDb(db, since, limit);
+    }
+    return readCallLogDb(db, since, limit);
+  } finally {
+    db.close();
+  }
+}
+
+function resolveDbPath(rawArg?: string): string {
+  if (rawArg) return path.resolve(rawArg);
+  if (SQLITE_FILE) return SQLITE_FILE;
+  throw new Error("No SQLITE_FILE and no --db path provided");
+}
+
+function describeInputSource(
+  inputPath: string | undefined,
+  dbPath: string | undefined,
+  dbSource: DbReplaySource | undefined,
+  usesDb: boolean
+): { source: string; path?: string; dbSource?: string } {
+  if (inputPath) return { source: "jsonl", path: path.resolve(inputPath) };
+  if (usesDb) {
+    return {
+      source: "sqlite",
+      path: resolveDbPath(dbPath),
+      dbSource: dbSource ?? "auto",
+    };
+  }
+  return { source: "stdin" };
+}
+
+function buildArtifactMetadata(args: ArgSpec, hasCandidateDb: boolean): RouterEvalArtifactMetadata {
+  const hasBaselineDb = Boolean(args.baselineDb);
+  return {
+    candidate: describeInputSource(args.input, args.db, args.dbSource, hasCandidateDb),
+    baseline:
+      args.baselineInput || hasBaselineDb
+        ? describeInputSource(
+            args.baselineInput,
+            args.baselineDb,
+            args.baselineDbSource,
+            hasBaselineDb
+          )
+        : undefined,
+    window: {
+      since: args.since,
+      limit: args.limit,
+    },
+    thresholds: {
+      maxAiqDrop: args.aiqDrop ?? 0,
+      maxCostIncrease: args.costIncrease ?? 0,
+    },
+    outputs: {
+      markdown: args.output ? path.resolve(args.output) : undefined,
+      json: args.jsonOutput ? path.resolve(args.jsonOutput) : undefined,
+      corpus: args.exportCorpus ? path.resolve(args.exportCorpus) : undefined,
+    },
+  };
+}
+
+async function writeCorpus(pathArg: string, observations: RouterObservation[]): Promise<void> {
+  const outPath = path.resolve(pathArg);
+  await fs.promises.mkdir(path.dirname(outPath), { recursive: true });
+  const lines = observations.map((observation) => JSON.stringify(observation));
+  await fs.promises.writeFile(outPath, `${lines.join("\n")}\n`, "utf8");
+}
+
+async function run() {
+  const args = parseArgs();
+  if (args.help) {
+    console.log(usage());
+    return;
+  }
+
+  const hasCandidateDb = Boolean(args.db || process.argv.includes("--db"));
+  const candidate: RouterObservation[] = args.input
+    ? await readJsonl(args.input)
+    : hasCandidateDb
+      ? await readDb(resolveDbPath(args.db), args.since, args.limit, args.dbSource)
+      : await readJsonl();
+
+  const baseline: RouterObservation[] | undefined = args.baselineInput
+    ? await readJsonl(args.baselineInput)
+    : args.baselineDb
+      ? await readDb(resolveDbPath(args.baselineDb), args.since, args.limit, args.baselineDbSource)
+      : undefined;
+
+  if (candidate.length === 0) {
+    console.error("No candidate observations found");
+    process.exitCode = 2;
+    return;
+  }
+
+  if (args.exportCorpus) {
+    await writeCorpus(args.exportCorpus, candidate);
+  }
+
+  const report = runRouterEval(candidate);
+  const metadata = buildArtifactMetadata(args, hasCandidateDb);
+  let output = formatRouterEvalReport(report);
+  let artifact: RouterEvalArtifact = createRouterEvalArtifact(report, metadata);
+
+  if (baseline && baseline.length > 0) {
+    const comparison = compareRouterEvalRuns(runRouterEval(baseline), report, {
+      aiqDrop: args.aiqDrop ?? 0,
+      relativeCostIncrease: args.costIncrease ?? 0,
+    });
+    output = formatRouterEvalComparison(comparison);
+    artifact = createRouterEvalArtifact(comparison, metadata);
+    console.log(output);
+    if (args.failOnRegression && comparison.regressions.length > 0) {
+      process.exitCode = 1;
+    }
+  } else {
+    console.log(output);
+  }
+
+  if (args.output) {
+    const outPath = path.resolve(args.output);
+    await fs.promises.writeFile(outPath, output, "utf8");
+  }
+
+  if (args.jsonOutput) {
+    const outPath = path.resolve(args.jsonOutput);
+    await fs.promises.writeFile(outPath, `${JSON.stringify(artifact, null, 2)}\n`, "utf8");
+  }
+}
+
+run().catch((error) => {
+  if (error && typeof error === "object" && "message" in error) {
+    console.error((error as Error).message);
+  } else {
+    console.error(String(error));
+  }
+  process.exitCode = 1;
+});
diff --git a/scripts/router-eval/patch-compare.ts b/scripts/router-eval/patch-compare.ts
new file mode 100644
index 00000000000..32fe80d9028
--- /dev/null
+++ b/scripts/router-eval/patch-compare.ts
@@ -0,0 +1,315 @@
+#!/usr/bin/env node
+import fs from "node:fs";
+import path from "node:path";
+
+type PatchOperation = {
+  op?: string;
+  path?: string;
+  value?: string;
+  evidence?: {
+    aiq?: number;
+    avgCostUsd?: number;
+    avgLatencyMs?: number;
+    regressions?: number;
+  };
+  rationale?: string;
+};
+
+type RouterConfigPatchArtifact = {
+  schemaVersion?: number;
+  kind?: string;
+  generatedAt?: string;
+  applyPolicy?: string;
+  source?: {
+    objective?: string;
+    runId?: string;
+    artifactPath?: string;
+  };
+  operations?: PatchOperation[];
+};
+
+type PatchComparison = {
+  schemaVersion: 1;
+  kind: "router-config-patch-comparison";
+  generatedAt: string;
+  runId: string;
+  thresholds: PatchThresholds;
+  baseline: PatchSummary;
+  candidate: PatchSummary;
+  delta: {
+    aiq: number;
+    avgCostUsd: number;
+    costIncreaseRatio: number;
+    avgLatencyMs: number;
+    latencyIncreaseRatio: number;
+    regressions: number;
+  };
+  changedRecommendation: boolean;
+  regressions: string[];
+  result: {
+    passed: boolean;
+    status: 0 | 1;
+  };
+};
+
+type PatchThresholds = {
+  maxAiqDrop: number;
+  maxCostIncrease: number;
+  maxLatencyIncrease: number;
+  maxRegressionIncrease: number;
+};
+
+type PatchSummary = {
+  name: string;
+  file: string;
+  objective: string;
+  runId: string;
+  recommendedConfigId: string;
+  aiq: number;
+  avgCostUsd: number;
+  avgLatencyMs: number;
+  regressions: number;
+  applyPolicy: string;
+};
+
+function getArgValue(name: string): string | undefined {
+  const index = process.argv.indexOf(`--${name}`);
+  if (index < 0) return undefined;
+  const value = process.argv[index + 1];
+  return value && !value.startsWith("--") ? value : undefined;
+}
+
+function usage(): string {
+  return [
+    "Usage:",
+    "  npm run eval:router:patch-compare -- --baseline <router-config.patch.json> --candidate <router-config.patch.json>",
+    "       [--baseline-name <name>] [--candidate-name <name>] [--artifact-dir <dir>] [--run-id <id>]",
+    "       [--output <file>] [--json-output <file>] [--fail-on-regression]",
+    "       [--max-aiq-drop <n>] [--max-cost-increase <ratio>] [--max-latency-increase <ratio>]",
+    "       [--max-regression-increase <n>]",
+    "",
+    "Compares two retained router config patch proposals without applying them.",
+  ].join("\n");
+}
+
+function getNumberArg(name: string, fallback: number): number {
+  const value = getArgValue(name);
+  if (value === undefined) return fallback;
+  const parsed = Number(value);
+  if (!Number.isFinite(parsed) || parsed < 0) {
+    console.error(`Invalid --${name} value: ${value}`);
+    process.exit(2);
+  }
+  return parsed;
+}
+
+function requireArg(name: string): string {
+  const value = getArgValue(name);
+  if (!value) {
+    console.error(`Missing required --${name}`);
+    process.exit(2);
+  }
+  return value;
+}
+
+function readPatch(file: string): RouterConfigPatchArtifact {
+  if (!fs.existsSync(file)) {
+    console.error(`[router-eval:patch-compare] patch file missing: ${file}`);
+    process.exit(2);
+  }
+  try {
+    return JSON.parse(fs.readFileSync(file, "utf8")) as RouterConfigPatchArtifact;
+  } catch (error) {
+    console.error(
+      `[router-eval:patch-compare] invalid JSON in ${file}: ${(error as Error).message}`
+    );
+    process.exit(2);
+  }
+}
+
+function requireNumber(value: unknown, field: string, file: string): number {
+  if (typeof value !== "number" || !Number.isFinite(value)) {
+    console.error(`[router-eval:patch-compare] invalid numeric evidence field ${field} in ${file}`);
+    process.exit(2);
+  }
+  return value;
+}
+
+function summarizePatch(
+  name: string,
+  file: string,
+  patch: RouterConfigPatchArtifact
+): PatchSummary {
+  if (patch.kind !== "router-config-patch") {
+    console.error(
+      `[router-eval:patch-compare] invalid patch kind in ${file}: ${patch.kind ?? "missing"}`
+    );
+    process.exit(2);
+  }
+  const operation = patch.operations?.[0];
+  if (operation?.op !== "recommend-router-config") {
+    console.error(
+      `[router-eval:patch-compare] missing recommend-router-config operation in ${file}`
+    );
+    process.exit(2);
+  }
+  if (typeof operation.value !== "string" || operation.value.length === 0) {
+    console.error(`[router-eval:patch-compare] invalid recommended config value in ${file}`);
+    process.exit(2);
+  }
+  return {
+    name,
+    file: path.resolve(file),
+    objective: patch.source?.objective ?? "unknown",
+    runId: patch.source?.runId ?? "unknown",
+    recommendedConfigId: operation.value,
+    aiq: requireNumber(operation.evidence?.aiq, "aiq", file),
+    avgCostUsd: requireNumber(operation.evidence?.avgCostUsd, "avgCostUsd", file),
+    avgLatencyMs: requireNumber(operation.evidence?.avgLatencyMs, "avgLatencyMs", file),
+    regressions: requireNumber(operation.evidence?.regressions, "regressions", file),
+    applyPolicy: patch.applyPolicy ?? "unknown",
+  };
+}
+
+function increaseRatio(delta: number, baseline: number): number {
+  if (baseline === 0) return delta > 0 ? Number.POSITIVE_INFINITY : 0;
+  return delta / baseline;
+}
+
+function findRegressions(
+  baseline: PatchSummary,
+  candidate: PatchSummary,
+  thresholds: PatchThresholds
+): string[] {
+  const aiqDrop = baseline.aiq - candidate.aiq;
+  const costDelta = candidate.avgCostUsd - baseline.avgCostUsd;
+  const latencyDelta = candidate.avgLatencyMs - baseline.avgLatencyMs;
+  const regressionDelta = candidate.regressions - baseline.regressions;
+  const regressions: string[] = [];
+  if (aiqDrop > thresholds.maxAiqDrop) regressions.push(`AIQ dropped by ${aiqDrop.toFixed(3)}`);
+  if (increaseRatio(costDelta, baseline.avgCostUsd) > thresholds.maxCostIncrease) {
+    regressions.push(
+      `average cost increased by ${increaseRatio(costDelta, baseline.avgCostUsd).toFixed(3)}`
+    );
+  }
+  if (increaseRatio(latencyDelta, baseline.avgLatencyMs) > thresholds.maxLatencyIncrease) {
+    regressions.push(
+      `average latency increased by ${increaseRatio(latencyDelta, baseline.avgLatencyMs).toFixed(3)}`
+    );
+  }
+  if (regressionDelta > thresholds.maxRegressionIncrease) {
+    regressions.push(`regression count increased by ${regressionDelta}`);
+  }
+  return regressions;
+}
+
+function comparePatches(
+  runId: string,
+  thresholds: PatchThresholds,
+  failOnRegression: boolean,
+  baseline: PatchSummary,
+  candidate: PatchSummary
+): PatchComparison {
+  const regressions = findRegressions(baseline, candidate, thresholds);
+  const status = failOnRegression && regressions.length > 0 ? 1 : 0;
+  const costDelta = candidate.avgCostUsd - baseline.avgCostUsd;
+  const latencyDelta = candidate.avgLatencyMs - baseline.avgLatencyMs;
+  return {
+    schemaVersion: 1,
+    kind: "router-config-patch-comparison",
+    generatedAt: new Date().toISOString(),
+    runId,
+    thresholds,
+    baseline,
+    candidate,
+    delta: {
+      aiq: candidate.aiq - baseline.aiq,
+      avgCostUsd: costDelta,
+      costIncreaseRatio: increaseRatio(costDelta, baseline.avgCostUsd),
+      avgLatencyMs: latencyDelta,
+      latencyIncreaseRatio: increaseRatio(latencyDelta, baseline.avgLatencyMs),
+      regressions: candidate.regressions - baseline.regressions,
+    },
+    changedRecommendation: baseline.recommendedConfigId !== candidate.recommendedConfigId,
+    regressions,
+    result: {
+      passed: regressions.length === 0,
+      status,
+    },
+  };
+}
+
+function formatComparison(comparison: PatchComparison): string {
+  return [
+    "# Router Config Patch Comparison",
+    "",
+    `Passed: ${comparison.result.passed ? "yes" : "no"}`,
+    `Changed recommendation: ${comparison.changedRecommendation ? "yes" : "no"}`,
+    ...(comparison.regressions.length > 0
+      ? ["", "## Regressions", "", ...comparison.regressions.map((item) => `- ${item}`)]
+      : []),
+    "",
+    "| Side | Name | Objective | Recommended Config | AIQ | Avg Cost | Avg Latency | Regressions | Apply Policy |",
+    "| --- | --- | --- | --- | ---: | ---: | ---: | ---: | --- |",
+    formatSummaryRow("Baseline", comparison.baseline),
+    formatSummaryRow("Candidate", comparison.candidate),
+    "",
+    "| Delta | AIQ | Avg Cost | Avg Latency | Regressions |",
+    "| --- | ---: | ---: | ---: | ---: |",
+    `| Candidate - Baseline | ${comparison.delta.aiq.toFixed(3)} | $${comparison.delta.avgCostUsd.toFixed(6)} | ${comparison.delta.avgLatencyMs.toFixed(2)}ms | ${comparison.delta.regressions} |`,
+    "",
+  ].join("\n");
+}
+
+function formatSummaryRow(side: string, summary: PatchSummary): string {
+  return `| ${side} | ${summary.name} | ${summary.objective} | ${summary.recommendedConfigId} | ${summary.aiq.toFixed(3)} | $${summary.avgCostUsd.toFixed(6)} | ${summary.avgLatencyMs.toFixed(2)}ms | ${summary.regressions} | ${summary.applyPolicy} |`;
+}
+
+function main(): void {
+  if (process.argv.includes("--help") || process.argv.includes("-h")) {
+    console.log(usage());
+    return;
+  }
+
+  const baselineFile = requireArg("baseline");
+  const candidateFile = requireArg("candidate");
+  const baselineName = getArgValue("baseline-name") ?? "baseline";
+  const candidateName = getArgValue("candidate-name") ?? "candidate";
+  const runId = getArgValue("run-id") ?? new Date().toISOString().replace(/[:.]/g, "-");
+  const thresholds: PatchThresholds = {
+    maxAiqDrop: getNumberArg("max-aiq-drop", Number.POSITIVE_INFINITY),
+    maxCostIncrease: getNumberArg("max-cost-increase", Number.POSITIVE_INFINITY),
+    maxLatencyIncrease: getNumberArg("max-latency-increase", Number.POSITIVE_INFINITY),
+    maxRegressionIncrease: getNumberArg("max-regression-increase", Number.POSITIVE_INFINITY),
+  };
+  const failOnRegression = process.argv.includes("--fail-on-regression");
+  const baseline = summarizePatch(baselineName, baselineFile, readPatch(baselineFile));
+  const candidate = summarizePatch(candidateName, candidateFile, readPatch(candidateFile));
+  const comparison = comparePatches(runId, thresholds, failOnRegression, baseline, candidate);
+  const markdown = formatComparison(comparison);
+  const artifactDir = getArgValue("artifact-dir");
+  const output = getArgValue("output");
+  const jsonOutput = getArgValue("json-output");
+
+  if (artifactDir) {
+    const runDir = path.resolve(artifactDir, runId);
+    fs.mkdirSync(runDir, { recursive: true });
+    fs.writeFileSync(path.join(runDir, "patch-comparison.md"), markdown);
+    fs.writeFileSync(
+      path.join(runDir, "patch-comparison.json"),
+      `${JSON.stringify(comparison, null, 2)}\n`
+    );
+  }
+  if (output) {
+    fs.mkdirSync(path.dirname(path.resolve(output)), { recursive: true });
+    fs.writeFileSync(output, markdown);
+  }
+  if (jsonOutput) {
+    fs.mkdirSync(path.dirname(path.resolve(jsonOutput)), { recursive: true });
+    fs.writeFileSync(jsonOutput, `${JSON.stringify(comparison, null, 2)}\n`);
+  }
+  console.log(markdown);
+  process.exit(comparison.result.status);
+}
+
+main();
diff --git a/scripts/router-eval/search.ts b/scripts/router-eval/search.ts
new file mode 100644
index 00000000000..1683585e558
--- /dev/null
+++ b/scripts/router-eval/search.ts
@@ -0,0 +1,439 @@
+#!/usr/bin/env node
+import fs from "node:fs";
+import path from "node:path";
+import { spawnSync } from "node:child_process";
+import { fileURLToPath } from "node:url";
+
+import type { RouterConfigAggregate, RouterEvalArtifact } from "@/lib/routerEval/index.ts";
+
+type Candidate = {
+  name: string;
+  path: string;
+};
+
+type SearchResult = {
+  candidateName: string;
+  configId: string;
+  runId: string;
+  aiq: number;
+  avgCostUsd: number;
+  avgLatencyMs: number;
+  regressions: number;
+  artifactPath: string;
+};
+
+type SearchObjective = "balanced" | "quality" | "cost" | "latency";
+
+type SearchRecommendation = SearchResult & {
+  objective: SearchObjective;
+  rank: number;
+  rationale: string;
+};
+
+type RouterConfigSuggestion = {
+  schemaVersion: 1;
+  kind: "router-config-suggestion";
+  generatedAt: string;
+  objective: SearchObjective;
+  recommendedConfigId: string;
+  sourceRunId: string;
+  sourceArtifactPath: string;
+  evidence: {
+    aiq: number;
+    avgCostUsd: number;
+    avgLatencyMs: number;
+    regressions: number;
+  };
+  applyPolicy: "manual-review";
+  rationale: string;
+};
+
+type RouterConfigPatchArtifact = {
+  schemaVersion: 1;
+  kind: "router-config-patch";
+  generatedAt: string;
+  applyPolicy: "manual-review";
+  source: {
+    objective: SearchObjective;
+    runId: string;
+    artifactPath: string;
+  };
+  operations: Array<{
+    op: "recommend-router-config";
+    path: "/router/recommendedConfigId";
+    value: string;
+    evidence: RouterConfigSuggestion["evidence"];
+    rationale: string;
+  }>;
+};
+
+const repoRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../..");
+const isBunRuntime = "Bun" in globalThis;
+
+function runTypeScriptScript(args: string[]) {
+  return spawnSync(process.execPath, isBunRuntime ? args : ["--import", "tsx", ...args], {
+    cwd: repoRoot,
+    encoding: "utf8",
+    stdio: ["ignore", "pipe", "pipe"],
+  });
+}
+
+function getArgValue(name: string): string | undefined {
+  const index = process.argv.indexOf(`--${name}`);
+  if (index < 0) return undefined;
+  const value = process.argv[index + 1];
+  return value && !value.startsWith("--") ? value : undefined;
+}
+
+function getArgValues(name: string): string[] {
+  const values: string[] = [];
+  for (let index = 0; index < process.argv.length; index += 1) {
+    if (process.argv[index] !== `--${name}`) continue;
+    const value = process.argv[index + 1];
+    if (value && !value.startsWith("--")) values.push(value);
+  }
+  return values;
+}
+
+function usage(): string {
+  return [
+    "Usage:",
+    "  npm run eval:router:search -- --baseline <baseline.ndjson>",
+    "       --candidate <name=path.ndjson> [--candidate <name=path.ndjson> ...]",
+    "       [--objective balanced|quality|cost|latency]",
+    "       [--artifact-dir <dir>] [--run-id <id>] [--max-aiq-drop <n>] [--max-cost-increase <n>]",
+    "",
+    "Ranks candidate corpora by router-eval AIQ while retaining comparison artifacts.",
+  ].join("\n");
+}
+
+function requireArg(name: string): string {
+  const value = getArgValue(name);
+  if (!value) {
+    console.error(`Missing required --${name}`);
+    process.exit(2);
+  }
+  return value;
+}
+
+function parseCandidate(raw: string): Candidate {
+  const splitAt = raw.indexOf("=");
+  if (splitAt <= 0 || splitAt === raw.length - 1) {
+    console.error(`Invalid --candidate value: ${raw}. Expected name=path.ndjson`);
+    process.exit(2);
+  }
+  return {
+    name: raw.slice(0, splitAt),
+    path: raw.slice(splitAt + 1),
+  };
+}
+
+function ensureReadable(filePath: string, label: string): void {
+  if (!fs.existsSync(filePath)) {
+    console.error(`[router-eval:search] ${label} missing: ${filePath}`);
+    process.exit(2);
+  }
+}
+
+function readArtifact(filePath: string): RouterEvalArtifact {
+  return JSON.parse(fs.readFileSync(filePath, "utf8")) as RouterEvalArtifact;
+}
+
+function parseObjective(value: string | undefined): SearchObjective {
+  if (!value) return "balanced";
+  if (value === "balanced" || value === "quality" || value === "cost" || value === "latency") {
+    return value;
+  }
+  console.error(
+    `Invalid --objective value: ${value}. Expected balanced, quality, cost, or latency.`
+  );
+  process.exit(2);
+}
+
+function compareConfigsByObjective(
+  objective: SearchObjective,
+  a: RouterConfigAggregate,
+  b: RouterConfigAggregate
+): number {
+  if (objective === "cost") {
+    return a.avgCostUsd - b.avgCostUsd || b.aiq - a.aiq || a.avgLatencyMs - b.avgLatencyMs;
+  }
+  if (objective === "latency") {
+    return a.avgLatencyMs - b.avgLatencyMs || b.aiq - a.aiq || a.avgCostUsd - b.avgCostUsd;
+  }
+  return b.aiq - a.aiq || a.avgCostUsd - b.avgCostUsd || a.avgLatencyMs - b.avgLatencyMs;
+}
+
+function selectBestConfig(
+  artifact: RouterEvalArtifact,
+  objective: SearchObjective
+): RouterConfigAggregate | undefined {
+  const configs =
+    artifact.comparison?.candidate.configurations ??
+    artifact.report?.configurations ??
+    artifact.report?.top ??
+    [];
+  return [...configs].sort((a, b) => compareConfigsByObjective(objective, a, b))[0];
+}
+
+function resultFromArtifact(
+  candidateName: string,
+  runId: string,
+  artifactPath: string,
+  objective: SearchObjective
+): SearchResult {
+  const artifact = readArtifact(artifactPath);
+  const best = selectBestConfig(artifact, objective);
+  if (!best) {
+    throw new Error(`No best candidate found in ${artifactPath}`);
+  }
+  return {
+    candidateName,
+    configId: best.configId,
+    runId,
+    aiq: best.aiq,
+    avgCostUsd: best.avgCostUsd,
+    avgLatencyMs: best.avgLatencyMs,
+    regressions: artifact.comparison?.regressions.length ?? 0,
+    artifactPath,
+  };
+}
+
+function formatSearch(results: SearchResult[]): string {
+  const lines = [
+    "# Router Eval Search",
+    "",
+    "| Rank | Candidate | AIQ | Avg Cost | Avg Latency | Regressions | Run |",
+    "| ---: | --- | ---: | ---: | ---: | ---: | --- |",
+  ];
+  results.forEach((result, index) => {
+    lines.push(
+      `| ${index + 1} | ${result.candidateName} | ${result.aiq.toFixed(3)} | $${result.avgCostUsd.toFixed(6)} | ${result.avgLatencyMs.toFixed(2)}ms | ${result.regressions} | ${result.runId} |`
+    );
+  });
+  return `${lines.join("\n")}\n`;
+}
+
+function compareByObjective(objective: SearchObjective, a: SearchResult, b: SearchResult): number {
+  if (objective === "cost") {
+    return (
+      a.regressions - b.regressions ||
+      a.avgCostUsd - b.avgCostUsd ||
+      b.aiq - a.aiq ||
+      a.avgLatencyMs - b.avgLatencyMs
+    );
+  }
+  if (objective === "latency") {
+    return (
+      a.regressions - b.regressions ||
+      a.avgLatencyMs - b.avgLatencyMs ||
+      b.aiq - a.aiq ||
+      a.avgCostUsd - b.avgCostUsd
+    );
+  }
+  if (objective === "quality") {
+    return (
+      b.aiq - a.aiq ||
+      a.regressions - b.regressions ||
+      a.avgCostUsd - b.avgCostUsd ||
+      a.avgLatencyMs - b.avgLatencyMs
+    );
+  }
+  return (
+    b.aiq - a.aiq ||
+    a.regressions - b.regressions ||
+    a.avgCostUsd - b.avgCostUsd ||
+    a.avgLatencyMs - b.avgLatencyMs
+  );
+}
+
+function recommendationRationale(objective: SearchObjective, result: SearchResult): string {
+  if (objective === "cost") {
+    return `${result.candidateName} has the best cost-first rank with ${result.regressions} regressions and $${result.avgCostUsd.toFixed(6)} average cost.`;
+  }
+  if (objective === "latency") {
+    return `${result.candidateName} has the best latency-first rank with ${result.regressions} regressions and ${result.avgLatencyMs.toFixed(2)}ms average latency.`;
+  }
+  if (objective === "quality") {
+    return `${result.candidateName} has the best quality-first rank with ${result.aiq.toFixed(3)} AIQ.`;
+  }
+  return `${result.candidateName} has the best balanced rank with ${result.aiq.toFixed(3)} AIQ, ${result.regressions} regressions, $${result.avgCostUsd.toFixed(6)} average cost, and ${result.avgLatencyMs.toFixed(2)}ms average latency.`;
+}
+
+function createRecommendation(
+  objective: SearchObjective,
+  results: SearchResult[]
+): SearchRecommendation {
+  const winner = results[0];
+  if (!winner) {
+    throw new Error("Cannot create a recommendation without search results");
+  }
+  return {
+    ...winner,
+    objective,
+    rank: 1,
+    rationale: recommendationRationale(objective, winner),
+  };
+}
+
+function createConfigSuggestion(
+  generatedAt: string,
+  recommendation: SearchRecommendation
+): RouterConfigSuggestion {
+  return {
+    schemaVersion: 1,
+    kind: "router-config-suggestion",
+    generatedAt,
+    objective: recommendation.objective,
+    recommendedConfigId: recommendation.configId,
+    sourceRunId: recommendation.runId,
+    sourceArtifactPath: recommendation.artifactPath,
+    evidence: {
+      aiq: recommendation.aiq,
+      avgCostUsd: recommendation.avgCostUsd,
+      avgLatencyMs: recommendation.avgLatencyMs,
+      regressions: recommendation.regressions,
+    },
+    applyPolicy: "manual-review",
+    rationale: recommendation.rationale,
+  };
+}
+
+function createConfigPatch(suggestion: RouterConfigSuggestion): RouterConfigPatchArtifact {
+  return {
+    schemaVersion: 1,
+    kind: "router-config-patch",
+    generatedAt: suggestion.generatedAt,
+    applyPolicy: "manual-review",
+    source: {
+      objective: suggestion.objective,
+      runId: suggestion.sourceRunId,
+      artifactPath: suggestion.sourceArtifactPath,
+    },
+    operations: [
+      {
+        op: "recommend-router-config",
+        path: "/router/recommendedConfigId",
+        value: suggestion.recommendedConfigId,
+        evidence: suggestion.evidence,
+        rationale: suggestion.rationale,
+      },
+    ],
+  };
+}
+
+function formatPatchOperations(patch: RouterConfigPatchArtifact): string {
+  const lines = [
+    "## Patch Operations",
+    "",
+    "| Op | Path | Value | AIQ | Avg Cost | Avg Latency | Regressions | Apply Policy |",
+    "| --- | --- | --- | ---: | ---: | ---: | ---: | --- |",
+  ];
+  for (const operation of patch.operations) {
+    lines.push(
+      `| ${operation.op} | ${operation.path} | ${operation.value} | ${operation.evidence.aiq.toFixed(3)} | $${operation.evidence.avgCostUsd.toFixed(6)} | ${operation.evidence.avgLatencyMs.toFixed(2)}ms | ${operation.evidence.regressions} | ${patch.applyPolicy} |`
+    );
+  }
+  return `${lines.join("\n")}\n`;
+}
+
+function main(): void {
+  if (process.argv.includes("--help") || process.argv.includes("-h")) {
+    console.log(usage());
+    return;
+  }
+
+  const baseline = requireArg("baseline");
+  ensureReadable(baseline, "baseline corpus");
+  const candidates = getArgValues("candidate").map(parseCandidate);
+  if (candidates.length === 0) {
+    console.error("At least one --candidate <name=path> is required");
+    process.exit(2);
+  }
+
+  const artifactDir = getArgValue("artifact-dir") ?? "artifacts/router-eval/search";
+  const searchId = getArgValue("run-id") ?? new Date().toISOString().replace(/[:.]/g, "-");
+  const objective = parseObjective(getArgValue("objective"));
+  const maxAiqDrop = getArgValue("max-aiq-drop") ?? "1";
+  const maxCostIncrease = getArgValue("max-cost-increase") ?? "0.05";
+  const searchDir = path.resolve(artifactDir, searchId);
+  fs.mkdirSync(searchDir, { recursive: true });
+
+  const results: SearchResult[] = [];
+  for (const candidate of candidates) {
+    ensureReadable(candidate.path, `${candidate.name} corpus`);
+    const runId = `${searchId}-${candidate.name}`;
+    const result = runTypeScriptScript([
+      "scripts/router-eval/compare.ts",
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate.path,
+      "--baseline-name",
+      "baseline",
+      "--candidate-name",
+      candidate.name,
+      "--artifact-dir",
+      searchDir,
+      "--run-id",
+      runId,
+      "--max-aiq-drop",
+      maxAiqDrop,
+      "--max-cost-increase",
+      maxCostIncrease,
+    ]);
+    if (result.stdout) process.stdout.write(result.stdout);
+    if (result.stderr) process.stderr.write(result.stderr);
+    if (result.error) {
+      console.error(
+        `[router-eval:search] failed to launch ${candidate.name}: ${result.error.message}`
+      );
+      process.exit(1);
+    }
+    if (result.status !== 0) {
+      console.error(
+        `[router-eval:search] comparison failed for ${candidate.name} with exit code ${result.status ?? 1}`
+      );
+      process.exit(result.status ?? 1);
+    }
+    const artifactPath = path.join(searchDir, runId, "router-eval.json");
+    results.push(resultFromArtifact(candidate.name, runId, artifactPath, objective));
+  }
+
+  results.sort((a, b) => compareByObjective(objective, a, b));
+
+  const recommendation = createRecommendation(objective, results);
+  const generatedAt = new Date().toISOString();
+  const suggestion = createConfigSuggestion(generatedAt, recommendation);
+  const patch = createConfigPatch(suggestion);
+  const markdown = `${formatSearch(results)}## Recommendation\n\n${recommendation.rationale}\n\n${formatPatchOperations(patch)}`;
+  const summary = {
+    schemaVersion: 1,
+    kind: "router-eval-search",
+    generatedAt,
+    baseline: path.resolve(baseline),
+    objective,
+    recommendation,
+    suggestion,
+    patch,
+    results,
+  };
+  fs.writeFileSync(path.join(searchDir, "search.md"), markdown);
+  fs.writeFileSync(path.join(searchDir, "search.json"), `${JSON.stringify(summary, null, 2)}\n`);
+  fs.writeFileSync(
+    path.join(searchDir, "recommendation.json"),
+    `${JSON.stringify(recommendation, null, 2)}\n`
+  );
+  fs.writeFileSync(
+    path.join(searchDir, "suggestion.json"),
+    `${JSON.stringify(suggestion, null, 2)}\n`
+  );
+  fs.writeFileSync(
+    path.join(searchDir, "router-config.patch.json"),
+    `${JSON.stringify(patch, null, 2)}\n`
+  );
+  console.log(markdown);
+  console.log(`[router-eval:search] artifacts: ${searchDir}`);
+}
+
+main();
diff --git a/scripts/router-eval/trends.ts b/scripts/router-eval/trends.ts
new file mode 100644
index 00000000000..70f277c1100
--- /dev/null
+++ b/scripts/router-eval/trends.ts
@@ -0,0 +1,198 @@
+#!/usr/bin/env node
+import fs from "node:fs";
+import path from "node:path";
+
+import type {
+  RouterConfigAggregate,
+  RouterEvalArtifact,
+  RouterEvalArtifactMetadata,
+  RouterEvalComparison,
+  RouterEvalReport,
+} from "@/lib/routerEval/index.ts";
+
+type TrendRow = {
+  runId: string;
+  generatedAt: string;
+  kind: RouterEvalArtifact["kind"];
+  bestConfig: string;
+  aiq: number;
+  avgCostUsd: number;
+  avgLatencyMs: number;
+  regressions: number;
+  source: string;
+  window: string;
+};
+
+function getArgValue(name: string): string | undefined {
+  const index = process.argv.indexOf(`--${name}`);
+  if (index < 0) return undefined;
+  const value = process.argv[index + 1];
+  return value && !value.startsWith("--") ? value : undefined;
+}
+
+function usage(): string {
+  return [
+    "Usage:",
+    "  npm run eval:router:trends -- --artifact-dir <dir> [--limit <n>] [--dashboard]",
+    "",
+    "Reads retained router-eval JSON artifacts and prints a markdown trend table or dashboard.",
+  ].join("\n");
+}
+
+function readJson(filePath: string): RouterEvalArtifact | null {
+  try {
+    return JSON.parse(fs.readFileSync(filePath, "utf8")) as RouterEvalArtifact;
+  } catch {
+    return null;
+  }
+}
+
+function bestFromReport(report: RouterEvalReport): RouterConfigAggregate | undefined {
+  return report.top[0];
+}
+
+function bestFromComparison(comparison: RouterEvalComparison): RouterConfigAggregate | undefined {
+  return comparison.candidate.top[0];
+}
+
+function toTrendRow(runId: string, artifact: RouterEvalArtifact): TrendRow | null {
+  const best = artifact.comparison
+    ? bestFromComparison(artifact.comparison)
+    : artifact.report
+      ? bestFromReport(artifact.report)
+      : undefined;
+
+  if (!best) return null;
+
+  return {
+    runId,
+    generatedAt: artifact.generatedAt,
+    kind: artifact.kind,
+    bestConfig: best.configId,
+    aiq: best.aiq,
+    avgCostUsd: best.avgCostUsd,
+    avgLatencyMs: best.avgLatencyMs,
+    regressions: artifact.comparison?.regressions.length ?? 0,
+    source: artifact.metadata?.candidate?.source ?? "unknown",
+    window: formatWindow(artifact.metadata?.window),
+  };
+}
+
+function formatWindow(window: RouterEvalArtifactMetadata["window"]): string {
+  if (!window || typeof window !== "object") return "all";
+  const parts: string[] = [];
+  if ("since" in window && typeof window.since === "string") parts.push(`since ${window.since}`);
+  if ("limit" in window && typeof window.limit === "number") parts.push(`limit ${window.limit}`);
+  return parts.length > 0 ? parts.join(", ") : "all";
+}
+
+function collectTrendRows(artifactDir: string): TrendRow[] {
+  if (!fs.existsSync(artifactDir)) return [];
+
+  const rows: TrendRow[] = [];
+  for (const entry of fs.readdirSync(artifactDir, { withFileTypes: true })) {
+    const runId = entry.name;
+    const jsonPath = entry.isDirectory()
+      ? path.join(artifactDir, runId, "router-eval.json")
+      : entry.isFile() && entry.name.endsWith(".json")
+        ? path.join(artifactDir, entry.name)
+        : "";
+    if (!jsonPath) continue;
+
+    const artifact = readJson(jsonPath);
+    if (!artifact || artifact.schemaVersion !== 1) continue;
+    const row = toTrendRow(runId.replace(/\.json$/, ""), artifact);
+    if (row) rows.push(row);
+  }
+
+  return rows.sort((a, b) => a.generatedAt.localeCompare(b.generatedAt));
+}
+
+function formatTrend(rows: TrendRow[], limit: number): string {
+  const limited = rows.slice(-limit);
+  const lines = [
+    "# Router Eval Trends",
+    "",
+    "| Run | Kind | Source | Window | Best Config | AIQ | Avg Cost | Avg Latency | Regressions |",
+    "| --- | --- | --- | --- | --- | ---: | ---: | ---: | ---: |",
+  ];
+
+  for (const row of limited) {
+    lines.push(
+      `| ${row.runId} | ${row.kind} | ${row.source} | ${row.window} | ${row.bestConfig} | ${row.aiq.toFixed(3)} | $${row.avgCostUsd.toFixed(6)} | ${row.avgLatencyMs.toFixed(2)}ms | ${row.regressions} |`
+    );
+  }
+
+  return `${lines.join("\n")}\n`;
+}
+
+function formatDelta(value: number): string {
+  if (value > 0) return `+${value.toFixed(3)}`;
+  return value.toFixed(3);
+}
+
+function average(values: number[]): number {
+  if (values.length === 0) return 0;
+  return values.reduce((sum, value) => sum + value, 0) / values.length;
+}
+
+function formatDashboard(rows: TrendRow[], limit: number): string {
+  const limited = rows.slice(-limit);
+  const latest = limited[limited.length - 1];
+  const previous = limited[limited.length - 2];
+  const aiqDelta = latest && previous ? latest.aiq - previous.aiq : 0;
+  const latencyDelta = latest && previous ? latest.avgLatencyMs - previous.avgLatencyMs : 0;
+  const costDelta = latest && previous ? latest.avgCostUsd - previous.avgCostUsd : 0;
+  const regressions = limited.reduce((sum, row) => sum + row.regressions, 0);
+  const lines = [
+    "# Router Eval Dashboard",
+    "",
+    `Runs: ${limited.length}`,
+    `Latest: ${latest?.runId ?? "n/a"}`,
+    `Best config: ${latest?.bestConfig ?? "n/a"}`,
+    `AIQ: ${latest ? latest.aiq.toFixed(3) : "0.000"} (${formatDelta(aiqDelta)})`,
+    `Avg latency: ${latest ? latest.avgLatencyMs.toFixed(2) : "0.00"}ms (${formatDelta(latencyDelta)}ms)`,
+    `Avg cost: $${latest ? latest.avgCostUsd.toFixed(6) : "0.000000"} (${formatDelta(costDelta)})`,
+    `Window: ${latest?.window ?? "all"}`,
+    `Source: ${latest?.source ?? "unknown"}`,
+    `Regression count: ${regressions}`,
+    "",
+    "## Rolling Averages",
+    "",
+    `AIQ: ${average(limited.map((row) => row.aiq)).toFixed(3)}`,
+    `Latency: ${average(limited.map((row) => row.avgLatencyMs)).toFixed(2)}ms`,
+    `Cost: $${average(limited.map((row) => row.avgCostUsd)).toFixed(6)}`,
+  ];
+
+  return `${lines.join("\n")}\n`;
+}
+
+function main(): void {
+  if (process.argv.includes("--help") || process.argv.includes("-h")) {
+    console.log(usage());
+    return;
+  }
+
+  const artifactDir = getArgValue("artifact-dir");
+  if (!artifactDir) {
+    console.error("Missing required --artifact-dir");
+    process.exit(2);
+  }
+
+  const limit = Number.parseInt(getArgValue("limit") ?? "20", 10);
+  const rows = collectTrendRows(path.resolve(artifactDir));
+  if (rows.length === 0) {
+    console.error(`No router-eval artifacts found in ${artifactDir}`);
+    process.exit(2);
+  }
+
+  const boundedLimit = Number.isFinite(limit) && limit > 0 ? limit : 20;
+  if (process.argv.includes("--dashboard")) {
+    console.log(formatDashboard(rows, boundedLimit));
+    return;
+  }
+
+  console.log(formatTrend(rows, boundedLimit));
+}
+
+main();
diff --git a/skills/omni-context-rtk/SKILL.md b/skills/omni-context-rtk/SKILL.md
index 8cb4d8193fd..8bad8466e57 100644
--- a/skills/omni-context-rtk/SKILL.md
+++ b/skills/omni-context-rtk/SKILL.md
@@ -43,6 +43,17 @@ curl https://localhost:20128/api/context/rtk/filters \
   -H "Authorization: Bearer $OMNIROUTE_TOKEN"
 ```
 
+### POST /api/context/rtk/import
+
+Validate or install an RTK TOML schema v1 filter file
+
+```bash
+curl -X POST https://localhost:20128/api/context/rtk/import \
+  -H "Authorization: Bearer $OMNIROUTE_TOKEN"
+  -H "Content-Type: application/json" \
+  -d '{}'
+```
+
 ### POST /api/context/rtk/test
 
 Run RTK compression preview for text
diff --git a/src/app/(dashboard)/dashboard/HomeProviderTopologySection.tsx b/src/app/(dashboard)/dashboard/HomeProviderTopologySection.tsx
index 53abbbeed21..8aef005e07a 100644
--- a/src/app/(dashboard)/dashboard/HomeProviderTopologySection.tsx
+++ b/src/app/(dashboard)/dashboard/HomeProviderTopologySection.tsx
@@ -27,9 +27,14 @@ export function HomeProviderTopologySection({
   enabled?: boolean;
 }) {
   const t = useTranslations("home");
+  const tCommon = useTranslations("common");
+  const tSettings = useTranslations("settings");
+  const tAnalytics = useTranslations("analytics");
   // #4596: gate the live-WS connection so it only opens while the topology
   // section is actually shown on the home page.
   const { activeRequests: liveActiveRequests } = useLiveRequests({ enabled });
+  const activeRequests = selectActiveRequests(liveActiveRequests);
+  const activeProviderCount = new Set(activeRequests.map(({ provider }) => provider)).size;
 
   return (
     <Card>
@@ -37,24 +42,27 @@ export function HomeProviderTopologySection({
         <div>
           <h2 className="text-base font-semibold">{t("providerTopology")}</h2>
           <p className="text-xs text-text-muted">
-            Connected providers routing through OmniRoute in real time
+            {t("activeError", { active: activeProviderCount, errors: errorProvider ? 1 : 0 })}
           </p>
         </div>
         <div className="flex items-center gap-3 text-[11px] text-text-muted">
           <span className="flex items-center gap-1.5">
-            <span className="size-2 rounded-full bg-green-500" /> Active
+            <span className="size-2 rounded-full bg-green-500" />
+            {tCommon("active")}
           </span>
           <span className="flex items-center gap-1.5">
-            <span className="size-2 rounded-full bg-amber-500" /> Recent
+            <span className="size-2 rounded-full bg-amber-500" />
+            {tSettings("recent")}
           </span>
           <span className="flex items-center gap-1.5">
-            <span className="size-2 rounded-full bg-red-500" /> Error
+            <span className="size-2 rounded-full bg-red-500" />
+            {tAnalytics("modelStatusError")}
           </span>
         </div>
       </div>
       <ProviderTopology
         providers={providers}
-        activeRequests={selectActiveRequests(liveActiveRequests)}
+        activeRequests={activeRequests}
         lastProvider={lastProvider}
         errorProvider={errorProvider}
       />
diff --git a/src/app/(dashboard)/dashboard/analytics/CompressionAnalyticsTab.tsx b/src/app/(dashboard)/dashboard/analytics/CompressionAnalyticsTab.tsx
index 1ad0ea99267..c5e9c4d7f11 100644
--- a/src/app/(dashboard)/dashboard/analytics/CompressionAnalyticsTab.tsx
+++ b/src/app/(dashboard)/dashboard/analytics/CompressionAnalyticsTab.tsx
@@ -83,7 +83,10 @@ function ModeBar({
           {skipped > 0 && (
             // #4268: attempted-but-no-op runs (e.g. Stacked saved nothing) are
             // recorded now, so this mode is visible even when count is 0.
-            <span className="text-text-muted/70"> · {skipped.toLocaleString()} skipped (no-op)</span>
+            <span className="text-text-muted/70">
+              {" "}
+              · {skipped.toLocaleString()} skipped (no-op)
+            </span>
           )}
         </span>
       </div>
@@ -174,6 +177,7 @@ export default function CompressionAnalyticsTab() {
 
   const modes = Object.entries(stats.byMode).sort(([, a], [, b]) => b.count - a.count);
   const providers = Object.entries(stats.byProvider).sort(([, a], [, b]) => b.count - a.count);
+  const totalAttempts = stats.totalRequests + (stats.totalSkipped ?? 0);
 
   // Calculate max tokens for hourly chart scaling
   const maxTokensPerHour = Math.max(...stats.last24h.map((h) => h.tokensSaved), 1);
@@ -209,7 +213,7 @@ export default function CompressionAnalyticsTab() {
         <StatCard
           icon="compress"
           label={t("compressionAnalyticsTotalRequests")}
-          value={stats.totalRequests.toLocaleString()}
+          value={totalAttempts.toLocaleString()}
         />
         <StatCard
           icon="token"
@@ -365,7 +369,7 @@ export default function CompressionAnalyticsTab() {
       )}
 
       {/* Empty state */}
-      {stats.totalRequests === 0 && (
+      {totalAttempts === 0 && (
         <div className="card p-8 text-center text-text-muted">
           <span className="material-symbols-outlined text-[48px] mb-3 block text-primary opacity-50">
             compress
diff --git a/src/app/(dashboard)/dashboard/combos/page.tsx b/src/app/(dashboard)/dashboard/combos/page.tsx
index 4efc9618a3c..ea29735d058 100644
--- a/src/app/(dashboard)/dashboard/combos/page.tsx
+++ b/src/app/(dashboard)/dashboard/combos/page.tsx
@@ -1,6 +1,6 @@
 "use client";
 
-import { useState, useEffect, useCallback, useMemo, useRef } from "react";
+import { useState, useEffect, useCallback, useMemo, useRef, memo } from "react";
 import dynamic from "next/dynamic";
 import Link from "next/link";
 import { useRouter, useSearchParams } from "next/navigation";
@@ -1549,7 +1549,7 @@ function ComboReadinessPanel({ checks, blockers, showDescription = true }) {
   );
 }
 
-function ComboCard({
+function ComboCardInner({
   combo,
   metrics,
   compressionEnabled,
@@ -1767,6 +1767,7 @@ function ComboCard({
     </Card>
   );
 }
+const ComboCard = memo(ComboCardInner);
 
 function TestResultsView({ results }) {
   const emailsVisible = useEmailPrivacyStore((s) => s.emailsVisible);
diff --git a/src/app/(dashboard)/dashboard/context/rtk/RtkContextPageClient.tsx b/src/app/(dashboard)/dashboard/context/rtk/RtkContextPageClient.tsx
index 13df110d249..c49f89b70f6 100644
--- a/src/app/(dashboard)/dashboard/context/rtk/RtkContextPageClient.tsx
+++ b/src/app/(dashboard)/dashboard/context/rtk/RtkContextPageClient.tsx
@@ -4,6 +4,7 @@ import { useEffect, useMemo, useState } from "react";
 import { useTranslations } from "next-intl";
 import { SegmentedControl, Collapsible } from "@/shared/components";
 import RtkLearnDiscoverCard from "./RtkLearnDiscoverCard";
+import RtkTomlImportCard from "./RtkTomlImportCard";
 
 type RtkFilter = {
   id: string;
@@ -78,11 +79,14 @@ export default function RtkContextPageClient() {
       .catch(() => {});
   }, []);
 
-  useEffect(() => {
+  const loadFilters = () =>
     fetch("/api/context/rtk/filters")
       .then((res) => (res.ok ? res.json() : null))
       .then((data) => setFilters(Array.isArray(data?.filters) ? data.filters : []))
       .catch(() => {});
+
+  useEffect(() => {
+    void loadFilters();
     fetch("/api/context/rtk/config")
       .then((res) => (res.ok ? res.json() : null))
       .then((data) => setConfig(data))
@@ -363,6 +367,8 @@ export default function RtkContextPageClient() {
         </div>
       </Collapsible>
 
+      {viewMode === "advanced" && <RtkTomlImportCard onInstalled={loadFilters} />}
+
       <RtkLearnDiscoverCard />
     </div>
   );
diff --git a/src/app/(dashboard)/dashboard/context/rtk/RtkTomlImportCard.tsx b/src/app/(dashboard)/dashboard/context/rtk/RtkTomlImportCard.tsx
new file mode 100644
index 00000000000..f6a3718bbf7
--- /dev/null
+++ b/src/app/(dashboard)/dashboard/context/rtk/RtkTomlImportCard.tsx
@@ -0,0 +1,255 @@
+"use client";
+
+import { useState } from "react";
+import { useTranslations } from "next-intl";
+
+const RTK_TOML_MAX_BYTES = 1024 * 1024;
+
+interface ImportFilterSummary {
+  id: string;
+  description: string;
+  category: string;
+  commandPatterns: string[];
+  testCount: number;
+}
+
+interface ImportTestOutcome {
+  filterId: string;
+  testName: string;
+  passed: boolean;
+}
+
+interface ImportResult {
+  sha256: string;
+  passed: boolean;
+  filters: ImportFilterSummary[];
+  outcomes: ImportTestOutcome[];
+  warnings: string[];
+  installedPath?: string;
+  backupCreated?: boolean;
+}
+
+interface RtkTomlImportCardProps {
+  onInstalled?: () => void | Promise<void>;
+}
+
+interface RtkTomlEditorProps {
+  content: string;
+  processing: "validate" | "install" | null;
+  overwrite: boolean;
+  onContentChange: (content: string) => void;
+  onFileChange: (file: File | undefined) => void;
+  onProcess: (action: "validate" | "install") => void;
+  onOverwriteChange: (overwrite: boolean) => void;
+}
+
+async function readErrorMessage(response: Response): Promise<string | null> {
+  try {
+    const body = (await response.json()) as { error?: { message?: unknown } };
+    return typeof body.error?.message === "string" ? body.error.message : null;
+  } catch {
+    return null;
+  }
+}
+
+function RtkTomlEditor({
+  content,
+  processing,
+  overwrite,
+  onContentChange,
+  onFileChange,
+  onProcess,
+  onOverwriteChange,
+}: RtkTomlEditorProps) {
+  const t = useTranslations("contextRtk");
+  return (
+    <>
+      <div className="mt-3 flex flex-col gap-3">
+        <label className="text-xs font-medium text-text-main">
+          {t("tomlChooseFile")}
+          <input
+            type="file"
+            accept=".toml,text/plain,application/toml"
+            onChange={(event) => onFileChange(event.target.files?.[0])}
+            data-testid="rtk-toml-file"
+            className="mt-1 block w-full text-xs text-text-muted file:mr-3 file:rounded file:border file:border-border file:bg-bg file:px-2.5 file:py-1 file:text-xs file:text-text-main"
+          />
+        </label>
+        <textarea
+          value={content}
+          onChange={(event) => onContentChange(event.target.value)}
+          placeholder={t("tomlImportPlaceholder")}
+          data-testid="rtk-toml-content"
+          className="h-56 w-full rounded-lg border border-border bg-bg p-3 font-mono text-xs text-text-main"
+        />
+      </div>
+
+      <div className="mt-3 flex flex-wrap items-center gap-3">
+        <button
+          type="button"
+          onClick={() => onProcess("validate")}
+          disabled={processing !== null || !content.trim()}
+          data-testid="rtk-toml-validate"
+          className="rounded border border-border px-3 py-1.5 text-xs font-medium text-text-main hover:bg-surface-hover disabled:opacity-50"
+        >
+          {processing === "validate" ? t("tomlValidating") : t("tomlValidate")}
+        </button>
+        <button
+          type="button"
+          onClick={() => onProcess("install")}
+          disabled={processing !== null || !content.trim()}
+          data-testid="rtk-toml-install"
+          className="rounded bg-primary px-3 py-1.5 text-xs font-medium text-white disabled:opacity-50"
+        >
+          {processing === "install" ? t("tomlInstalling") : t("tomlInstall")}
+        </button>
+        <label className="flex items-center gap-2 text-xs text-text-main">
+          <input
+            type="checkbox"
+            checked={overwrite}
+            onChange={(event) => onOverwriteChange(event.target.checked)}
+            data-testid="rtk-toml-overwrite"
+          />
+          {t("tomlConfirmOverwrite")}
+        </label>
+      </div>
+    </>
+  );
+}
+
+function RtkTomlResult({ result }: { result: ImportResult }) {
+  const t = useTranslations("contextRtk");
+  return (
+    <div className="mt-4 rounded-lg border border-border bg-bg p-3" data-testid="rtk-toml-result">
+      <p className="text-xs font-medium text-text-main">
+        {result.installedPath
+          ? t("tomlInstalled", { path: result.installedPath })
+          : result.passed
+            ? t("tomlValidationPassed")
+            : t("tomlValidationFailed")}
+      </p>
+      <p className="mt-1 text-[11px] text-text-muted">
+        {t("tomlValidationSummary", {
+          filters: result.filters.length,
+          tests: result.outcomes.length,
+        })}
+      </p>
+      {result.backupCreated && (
+        <p className="mt-1 text-[11px] text-text-muted">{t("tomlBackupCreated")}</p>
+      )}
+      {result.filters.length > 0 && (
+        <ul className="mt-2 space-y-1 text-[11px] text-text-main">
+          {result.filters.map((filter) => (
+            <li key={filter.id}>
+              <code>{filter.id}</code> · {filter.category} ·{" "}
+              {t("tomlTestCount", { count: filter.testCount })}
+            </li>
+          ))}
+        </ul>
+      )}
+      {result.outcomes.length > 0 && (
+        <ul className="mt-2 space-y-1 text-[11px] text-text-main">
+          {result.outcomes.map((outcome) => (
+            <li key={`${outcome.filterId}:${outcome.testName}`}>
+              <span
+                className={
+                  outcome.passed
+                    ? "text-emerald-700 dark:text-emerald-300"
+                    : "text-red-600 dark:text-red-400"
+                }
+              >
+                {outcome.passed ? t("tomlTestPassed") : t("tomlTestFailed")}
+              </span>{" "}
+              · <code>{outcome.filterId}</code> · {outcome.testName}
+            </li>
+          ))}
+        </ul>
+      )}
+      {result.warnings.length > 0 && (
+        <ul className="mt-2 space-y-1 text-[11px] text-amber-700 dark:text-amber-300">
+          {result.warnings.map((warning) => (
+            <li key={warning}>{warning}</li>
+          ))}
+        </ul>
+      )}
+    </div>
+  );
+}
+
+export default function RtkTomlImportCard({ onInstalled }: RtkTomlImportCardProps) {
+  const t = useTranslations("contextRtk");
+  const [content, setContent] = useState("");
+  const [result, setResult] = useState<ImportResult | null>(null);
+  const [error, setError] = useState<string | null>(null);
+  const [processing, setProcessing] = useState<"validate" | "install" | null>(null);
+  const [overwrite, setOverwrite] = useState(false);
+
+  async function processImport(action: "validate" | "install") {
+    if (!content.trim()) return;
+    setProcessing(action);
+    setError(null);
+    try {
+      const response = await fetch("/api/context/rtk/import", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ action, content, overwrite: action === "install" && overwrite }),
+      });
+      if (!response.ok) {
+        throw new Error((await readErrorMessage(response)) ?? t("tomlImportError"));
+      }
+      const nextResult = (await response.json()) as ImportResult;
+      setResult(nextResult);
+      if (action === "install") await onInstalled?.();
+    } catch (caught) {
+      setResult(null);
+      setError(caught instanceof Error ? caught.message : t("tomlImportError"));
+    } finally {
+      setProcessing(null);
+    }
+  }
+
+  async function loadFile(file: File | undefined) {
+    if (!file) return;
+    setError(null);
+    if (file.size > RTK_TOML_MAX_BYTES) {
+      setError(t("tomlFileReadError"));
+      return;
+    }
+    try {
+      setContent(await file.text());
+      setResult(null);
+    } catch {
+      setError(t("tomlFileReadError"));
+    }
+  }
+
+  return (
+    <section
+      className="rounded-lg border border-border bg-surface p-4"
+      data-testid="rtk-toml-import"
+    >
+      <h2 className="text-sm font-semibold text-text-main">{t("tomlImportTitle")}</h2>
+      <p className="mt-1 text-xs text-text-muted">{t("tomlImportDesc")}</p>
+      <RtkTomlEditor
+        content={content}
+        processing={processing}
+        overwrite={overwrite}
+        onContentChange={(nextContent) => {
+          setContent(nextContent);
+          setResult(null);
+        }}
+        onFileChange={(file) => void loadFile(file)}
+        onProcess={(action) => void processImport(action)}
+        onOverwriteChange={setOverwrite}
+      />
+
+      {error && (
+        <p className="mt-3 text-xs text-red-600 dark:text-red-400" data-testid="rtk-toml-error">
+          {error}
+        </p>
+      )}
+
+      {result && <RtkTomlResult result={result} />}
+    </section>
+  );
+}
diff --git a/src/app/(dashboard)/dashboard/context/settings/CompressionPanel.tsx b/src/app/(dashboard)/dashboard/context/settings/CompressionPanel.tsx
index 3e416834c95..3d9bbd9362b 100644
--- a/src/app/(dashboard)/dashboard/context/settings/CompressionPanel.tsx
+++ b/src/app/(dashboard)/dashboard/context/settings/CompressionPanel.tsx
@@ -67,6 +67,7 @@ interface CompressionConfig {
   // The panel currently surfaces the computed target read-only; mode/policy editors are a
   // follow-up (the load/save path does not yet populate this field).
   contextBudget?: ContextBudgetConfig;
+  liveZone?: { enabled: boolean };
 }
 
 const CAVEMAN_OUTPUT_LEVELS: CavemanIntensity[] = ["lite", "full", "ultra"];
@@ -81,6 +82,7 @@ const DEFAULT_CONFIG: CompressionConfig = {
   outputStyles: [],
   ultraEngine: "heuristic",
   ultraSlmPrewarm: false,
+  liveZone: { enabled: false },
 };
 
 function normalizeEngines(raw: unknown): Record<string, EngineToggle> {
@@ -88,11 +90,40 @@ function normalizeEngines(raw: unknown): Record<string, EngineToggle> {
   const source = (raw && typeof raw === "object" ? raw : {}) as Record<string, EngineToggle>;
   for (const id of ENGINE_IDS) {
     const cur = source[id];
-    engines[id] = cur ? { enabled: cur.enabled === true, ...(cur.level ? { level: cur.level } : {}) } : { enabled: false };
+    engines[id] = cur
+      ? { enabled: cur.enabled === true, ...(cur.level ? { level: cur.level } : {}) }
+      : { enabled: false };
   }
   return engines;
 }
 
+function LiveZoneToggle({
+  enabled,
+  saving,
+  onChange,
+}: {
+  enabled: boolean;
+  saving: boolean;
+  onChange: (enabled: boolean) => void;
+}) {
+  const t = useTranslations("settings");
+  return (
+    <label className="flex items-center justify-between gap-4">
+      <span className="space-y-0.5">
+        <span className="block text-sm text-text-muted">{t("compressionLiveZoneTitle")}</span>
+        <span className="block text-xs text-text-muted">{t("compressionLiveZoneDesc")}</span>
+      </span>
+      <Toggle
+        size="sm"
+        checked={enabled}
+        onChange={onChange}
+        disabled={saving}
+        ariaLabel={t("compressionLiveZoneTitle")}
+      />
+    </label>
+  );
+}
+
 export default function CompressionPanel() {
   const t = useTranslations("settings");
   // D-A6/§7: locale-gated styles (e.g. terse-cjk → zh) are only OFFERED under their locale.
@@ -256,8 +287,7 @@ export default function CompressionPanel() {
           )}
           {status === "error" && (
             <span className="flex items-center gap-1 text-xs font-medium text-red-500">
-              <span className="material-symbols-outlined text-[14px]">error</span>{" "}
-              {t("saveFailed")}
+              <span className="material-symbols-outlined text-[14px]">error</span> {t("saveFailed")}
             </span>
           )}
           <Toggle
@@ -371,9 +401,7 @@ export default function CompressionPanel() {
             >
               <div className="min-w-0">
                 <p className="text-sm text-text-main">{meta.label}</p>
-                {meta.description && (
-                  <p className="text-xs text-text-muted">{meta.description}</p>
-                )}
+                {meta.description && <p className="text-xs text-text-muted">{meta.description}</p>}
               </div>
               <div className="flex shrink-0 items-center gap-2">
                 <select
@@ -410,15 +438,11 @@ export default function CompressionPanel() {
           or the opt-in LLMLingua-2 SLM Tier-B) + best-effort pre-warm. */}
       <div className="mt-2 flex flex-col gap-3 border-t border-border/30 py-3">
         <label className="flex items-center justify-between">
-          <span className="text-sm font-medium text-text-main">
-            {t("compressionUltraEngine")}
-          </span>
+          <span className="text-sm font-medium text-text-main">{t("compressionUltraEngine")}</span>
           <select
             data-testid="ultra-engine-select"
             value={config.ultraEngine ?? "heuristic"}
-            onChange={(e) =>
-              save({ ultraEngine: e.target.value === "slm" ? "slm" : "heuristic" })
-            }
+            onChange={(e) => save({ ultraEngine: e.target.value === "slm" ? "slm" : "heuristic" })}
             disabled={saving}
             className="w-44 rounded border border-border bg-surface px-2 py-1 text-sm text-text-main"
           >
@@ -431,9 +455,7 @@ export default function CompressionPanel() {
           <>
             <p className="text-xs text-text-muted">{t("compressionUltraSlmHint")}</p>
             <label className="flex items-center justify-between">
-              <span className="text-sm text-text-muted">
-                {t("compressionUltraSlmPrewarm")}
-              </span>
+              <span className="text-sm text-text-muted">{t("compressionUltraSlmPrewarm")}</span>
               <span data-testid="ultra-slm-prewarm-toggle">
                 <Toggle
                   size="sm"
@@ -505,6 +527,11 @@ export default function CompressionPanel() {
             <option value="never">{t("compressionPreserveSystemNever")}</option>
           </select>
         </label>
+        <LiveZoneToggle
+          enabled={config.liveZone?.enabled === true}
+          saving={saving}
+          onChange={(enabled) => save({ liveZone: { enabled } })}
+        />
       </div>
     </Card>
   );
diff --git a/src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.tsx b/src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.tsx
index 54e27f4991e..f2972f1ccc1 100644
--- a/src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.tsx
+++ b/src/app/(dashboard)/dashboard/endpoint/EndpointPageClient.tsx
@@ -5,7 +5,7 @@ import Link from "next/link";
 import { Card, Button, Input, Modal, CardSkeleton, SegmentedControl } from "@/shared/components";
 import Toggle from "@/shared/components/Toggle";
 import { useCopyToClipboard } from "@/shared/hooks/useCopyToClipboard";
-import { useDisplayBaseUrl } from "@/shared/hooks";
+import { isPublicDisplayBaseUrl, useDisplayBaseUrl } from "@/shared/hooks";
 import { AI_PROVIDERS, getProviderByAlias } from "@/shared/constants/providers";
 import { getProviderDisplayName } from "@/lib/display/names";
 import { useTranslations } from "next-intl";
@@ -19,12 +19,7 @@ const CLOUD_ACTION_TIMEOUT_MS = 15000;
 
 type TranslationValues = Record<string, string | number | boolean | Date>;
 type CloudflaredTunnelPhase =
-  | "unsupported"
-  | "not_installed"
-  | "stopped"
-  | "starting"
-  | "running"
-  | "error";
+  "unsupported" | "not_installed" | "stopped" | "starting" | "running" | "error";
 
 type CloudflaredTunnelStatus = {
   supported: boolean;
@@ -43,12 +38,7 @@ type CloudflaredTunnelStatus = {
 };
 
 type TailscaleTunnelPhase =
-  | "unsupported"
-  | "not_installed"
-  | "needs_login"
-  | "stopped"
-  | "running"
-  | "error";
+  "unsupported" | "not_installed" | "needs_login" | "stopped" | "running" | "error";
 
 type TailscaleTunnelStatus = {
   supported: boolean;
@@ -70,13 +60,7 @@ type TailscaleTunnelStatus = {
 };
 
 type NgrokTunnelPhase =
-  | "unsupported"
-  | "not_installed"
-  | "stopped"
-  | "needs_auth"
-  | "starting"
-  | "running"
-  | "error";
+  "unsupported" | "not_installed" | "stopped" | "needs_auth" | "starting" | "running" | "error";
 
 type NgrokTunnelStatus = {
   supported: boolean;
@@ -187,6 +171,7 @@ export default function APIPageClient({ machineId }: Readonly<APIPageClientProps
   const [ngrokToken, setNgrokToken] = useState("");
   const [showNgrokTunnel, setShowNgrokTunnel] = useState(true);
   const [expandedTunnel, setExpandedTunnel] = useState<string | null>(null);
+  const [localApiUrl, setLocalApiUrl] = useState("http://localhost:20128/v1");
   const [lanUrls, setLanUrls] = useState<string[]>([]);
   const [tailscaleIpUrl, setTailscaleIpUrl] = useState<string | null>(null);
   const [activeEndpointTab, setActiveEndpointTab] = useState<EndpointTab>("apis");
@@ -335,6 +320,7 @@ export default function APIPageClient({ machineId }: Readonly<APIPageClientProps
           if (res.ok) {
             const data = await res.json();
             if (mounted) {
+              if (data.localUrl) setLocalApiUrl(data.localUrl);
               setLanUrls(data.lanUrls ?? []);
               if (data.tailscaleIpUrl) setTailscaleIpUrl(data.tailscaleIpUrl);
             }
@@ -1080,7 +1066,8 @@ export default function APIPageClient({ machineId }: Readonly<APIPageClientProps
   }, [fetchTailscaleStatus, handleTailscaleEnable, tailscalePassword, translateOrFallback]);
 
   const displayBaseUrl = useDisplayBaseUrl();
-  const baseUrl = `${displayBaseUrl}/v1`;
+  const displayApiUrl = `${displayBaseUrl}/v1`;
+  const publicDisplayApiUrl = isPublicDisplayBaseUrl(displayBaseUrl) ? displayApiUrl : null;
   const normalizedCloudBaseUrl = cloudBaseUrl
     ? resolvedMachineId && !cloudBaseUrl.endsWith(`/${resolvedMachineId}`)
       ? `${cloudBaseUrl}/${resolvedMachineId}`
@@ -1097,13 +1084,9 @@ export default function APIPageClient({ machineId }: Readonly<APIPageClientProps
     );
   }
 
-  // Use new format endpoint (machineId embedded in key)
-  const currentEndpoint = cloudEnabled && cloudEndpointNew ? cloudEndpointNew : baseUrl;
-
-  const activeUrls = [
-    { label: "Local", url: baseUrl, key: "active_local" },
-    ...(cloudEnabled && cloudEndpointNew
-      ? [{ label: "Cloud", url: cloudEndpointNew, key: "active_cloud" }]
+  const activeTunnelUrls = [
+    ...(publicDisplayApiUrl
+      ? [{ label: t("tierPublic"), url: publicDisplayApiUrl, key: "active_public" }]
       : []),
     ...(cloudflaredStatus?.running && cloudflaredStatus.apiUrl
       ? [{ label: "Cloudflare", url: cloudflaredStatus.apiUrl, key: "active_cf" }]
@@ -1121,6 +1104,21 @@ export default function APIPageClient({ machineId }: Readonly<APIPageClientProps
       ? [{ label: "ngrok", url: ngrokStatus.apiUrl, key: "active_ngrok" }]
       : []),
   ];
+  const preferredTunnelUrl = activeTunnelUrls[0]?.url ?? null;
+  const currentEndpoint =
+    preferredTunnelUrl ??
+    (cloudEnabled && cloudEndpointNew ? cloudEndpointNew : null) ??
+    displayApiUrl;
+  const activeUrls = [
+    ...activeTunnelUrls,
+    ...(cloudEnabled && cloudEndpointNew
+      ? [{ label: "Cloud", url: cloudEndpointNew, key: "active_cloud" }]
+      : []),
+    { label: "Local", url: localApiUrl, key: "active_local" },
+  ].filter(
+    (candidate, index, candidates) =>
+      candidates.findIndex((other) => other.url === candidate.url) === index
+  );
   const visibleTunnelCount = [showCloudflaredTunnel, showTailscaleFunnel, showNgrokTunnel].filter(
     Boolean
   ).length;
@@ -1357,7 +1355,7 @@ export default function APIPageClient({ machineId }: Readonly<APIPageClientProps
               Running
             </span>
             <button
-              onClick={() => void copy(baseUrl, "endpoint_url")}
+              onClick={() => void copy(localApiUrl, "endpoint_url")}
               className="shrink-0 flex items-center gap-1 text-xs px-2.5 py-1.5 rounded-md border border-border/70 text-text-muted hover:text-text hover:border-border transition-colors"
             >
               <span className="material-symbols-outlined text-[14px]">
diff --git a/src/app/(dashboard)/dashboard/endpoint/__tests__/ApiEndpointsTab.test.tsx b/src/app/(dashboard)/dashboard/endpoint/__tests__/ApiEndpointsTab.test.tsx
index 58b46fdd080..9c316a1e3d1 100644
--- a/src/app/(dashboard)/dashboard/endpoint/__tests__/ApiEndpointsTab.test.tsx
+++ b/src/app/(dashboard)/dashboard/endpoint/__tests__/ApiEndpointsTab.test.tsx
@@ -29,10 +29,14 @@ vi.mock("next-intl", () => ({
       "endpoint.showInternal": "Show internal",
       "endpoint.hideInternal": "Hide internal",
       "endpoint.vscodeAliasTitle": "VS Code Token Alias",
-      "endpoint.vscodeAliasDescriptionReady": "Ready-to-paste compatibility URLs using the /api/v1/vscode/{token}/... endpoint.",
-      "endpoint.vscodeAliasDescriptionError": "Showing placeholder URLs because CLI keys could not be loaded in this session.",
-      "endpoint.vscodeAliasDescriptionLoading": "Loading CLI keys. Placeholder URLs are shown until a key is available.",
-      "endpoint.vscodeAliasDescriptionPlaceholder": "Showing placeholder URLs. Create or activate an API key in CLI Tools to replace {token}.",
+      "endpoint.vscodeAliasDescriptionReady":
+        "Ready-to-paste compatibility URLs using the /api/v1/vscode/{token}/... endpoint.",
+      "endpoint.vscodeAliasDescriptionError":
+        "Showing placeholder URLs because CLI keys could not be loaded in this session.",
+      "endpoint.vscodeAliasDescriptionLoading":
+        "Loading CLI keys. Placeholder URLs are shown until a key is available.",
+      "endpoint.vscodeAliasDescriptionPlaceholder":
+        "Showing placeholder URLs. Create or activate an API key in CLI Tools to replace {token}.",
       "endpoint.vscodeAliasManage": "CLI Tools",
       "endpoint.vscodeAliasBaseLabel": "VS Code base",
       "endpoint.vscodeAliasModelsLabel": "VS Code models",
@@ -114,6 +118,7 @@ describe("ApiEndpointsTab", () => {
       cleanupCallbacks.pop()?.();
     }
     document.body.innerHTML = "";
+    vi.unstubAllEnvs();
     vi.unstubAllGlobals();
   });
 
@@ -174,6 +179,7 @@ describe("ApiEndpointsTab", () => {
   });
 
   it("renders curl example using window.location.origin when NEXT_PUBLIC_BASE_URL is unset", async () => {
+    vi.stubEnv("NEXT_PUBLIC_BASE_URL", "");
     fetchMock.mockImplementation(async (input) => {
       if (input === "/api/cli-tools/keys") {
         return jsonResponse({ keys: [] });
@@ -221,7 +227,7 @@ describe("ApiEndpointsTab", () => {
 
     expect(
       expectedOrigins.some((origin) =>
-        renderedText.includes(`curl -X POST ${origin}/v1/chat/completions`)
+        renderedText.includes(`curl -X POST ${origin}/api/v1/chat/completions`)
       )
     ).toBe(true);
   });
diff --git a/src/app/(dashboard)/dashboard/endpoint/__tests__/EndpointPageClient.test.tsx b/src/app/(dashboard)/dashboard/endpoint/__tests__/EndpointPageClient.test.tsx
index b0adec3a2f1..49f0e37236b 100644
--- a/src/app/(dashboard)/dashboard/endpoint/__tests__/EndpointPageClient.test.tsx
+++ b/src/app/(dashboard)/dashboard/endpoint/__tests__/EndpointPageClient.test.tsx
@@ -186,9 +186,61 @@ describe("EndpointPageClient", () => {
       cleanupCallbacks.pop()?.();
     }
     document.body.innerHTML = "";
+    vi.unstubAllEnvs();
     vi.unstubAllGlobals();
   });
 
+  it("shows the public browser endpoint before the runtime local endpoint", async () => {
+    vi.stubEnv("NEXT_PUBLIC_BASE_URL", "http://localhost:20128");
+    vi.stubGlobal("location", { origin: "https://api.example.com" });
+
+    fetchMock.mockImplementation((input: RequestInfo | URL) => {
+      const path = getRequestPath(input);
+      if (path === "/api/settings") {
+        return Promise.resolve(
+          jsonResponse({
+            cloudEnabled: false,
+            cloudConfigured: false,
+            hideEndpointCloudflaredTunnel: true,
+            hideEndpointTailscaleFunnel: true,
+            hideEndpointNgrokTunnel: true,
+          })
+        );
+      }
+      if (path === "/api/network/info") {
+        return Promise.resolve(
+          jsonResponse({
+            localUrl: "http://localhost:20131/v1",
+            lanUrls: [],
+            tailscaleIpUrl: null,
+          })
+        );
+      }
+      if (path === "/v1/models") return Promise.resolve(jsonResponse({ data: [] }));
+      if (path === "/api/mcp/status") return Promise.resolve(jsonResponse({ online: false }));
+      if (path === "/api/a2a/status") {
+        return Promise.resolve(jsonResponse({ status: "ok", tasks: { activeStreams: 0 } }));
+      }
+      if (path === "/api/search/providers") {
+        return Promise.resolve(jsonResponse({ providers: [] }));
+      }
+      if (path === "/api/cli-tools/keys") return Promise.resolve(jsonResponse({ keys: [] }));
+      throw new Error(`Unexpected request: ${path}`);
+    });
+
+    renderEndpointPage();
+
+    await waitForText("https://api.example.com/v1");
+    await waitForText("http://localhost:20131/v1");
+
+    const displayedUrls = Array.from(document.body.querySelectorAll("code")).map(
+      (element) => element.textContent
+    );
+    expect(displayedUrls.indexOf("https://api.example.com/v1")).toBeLessThan(
+      displayedUrls.indexOf("http://localhost:20131/v1")
+    );
+  });
+
   it("renders the endpoint shell before models finish and skips hidden tunnel probes", async () => {
     const modelsDeferred = createDeferred<Response>();
 
diff --git a/src/app/(dashboard)/dashboard/onboarding/page.tsx b/src/app/(dashboard)/dashboard/onboarding/page.tsx
index de8b44e3cf2..7385bc0a86c 100644
--- a/src/app/(dashboard)/dashboard/onboarding/page.tsx
+++ b/src/app/(dashboard)/dashboard/onboarding/page.tsx
@@ -25,7 +25,7 @@ export default function OnboardingWizard() {
   const baseUrl = useDisplayBaseUrl();
   const [step, setStep] = useState(0);
   const [loading, setLoading] = useState(true);
-  const [apiEndpoint, setApiEndpoint] = useState(`${baseUrl}/api/v1`);
+  const apiEndpoint = `${baseUrl}/api/v1`;
 
   // Security step state
   const [password, setPassword] = useState("");
@@ -45,20 +45,11 @@ export default function OnboardingWizard() {
 
   // Check if setup is already complete
   useEffect(() => {
-    const resolveApiEndpoint = (apiPort) => {
-      if (typeof window === "undefined") return;
-      const protocol = window.location.protocol;
-      const hostname = window.location.hostname;
-      const effectiveApiPort = apiPort || 20128;
-      setApiEndpoint(`${protocol}//${hostname}:${effectiveApiPort}/api/v1`);
-    };
-
     const checkSetup = async () => {
       try {
         const res = await fetch("/api/settings");
         if (res.ok) {
           const settings = await res.json();
-          resolveApiEndpoint(settings?.apiPort);
           if (settings.setupComplete) {
             router.replace("/dashboard");
             return;
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/ProviderDetailPageClient.tsx b/src/app/(dashboard)/dashboard/providers/[id]/ProviderDetailPageClient.tsx
index 1c4465a2b87..7340e7e6fce 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/ProviderDetailPageClient.tsx
+++ b/src/app/(dashboard)/dashboard/providers/[id]/ProviderDetailPageClient.tsx
@@ -123,6 +123,7 @@ export default function ProviderDetailPageClient() {
     deleteConfirm,
     handleUpdateConnectionStatus,
     handleToggleRateLimit,
+    handleToggleQuotaVisibility,
     handleToggleClaudeExtraUsage,
     handleToggleCodexLimit,
     handleToggleCliproxyapiMode,
@@ -486,7 +487,10 @@ export default function ProviderDetailPageClient() {
       )}
       {!isUpstreamProxyProvider && !isFreeNoAuth && (
         <Card>
-          <ProviderAccountRoutingCard providerKey={providerId} connectionCount={connections.length} />
+          <ProviderAccountRoutingCard
+            providerKey={providerId}
+            connectionCount={connections.length}
+          />
           <ConnectionsHeaderToolbar
             providerId={providerId}
             providerInfo={providerInfo}
@@ -585,6 +589,7 @@ export default function ProviderDetailPageClient() {
                 deleteConfirm={deleteConfirm}
                 handleUpdateConnectionStatus={handleUpdateConnectionStatus}
                 handleToggleRateLimit={handleToggleRateLimit}
+                handleToggleQuotaVisibility={handleToggleQuotaVisibility}
                 handleToggleClaudeExtraUsage={handleToggleClaudeExtraUsage}
                 handleToggleCliproxyapiMode={handleToggleCliproxyapiMode}
                 handleToggleCodexLimit={handleToggleCodexLimit}
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/__tests__/phase1f.test.tsx b/src/app/(dashboard)/dashboard/providers/[id]/__tests__/phase1f.test.tsx
index a8591cb3c8b..4c12396eab9 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/__tests__/phase1f.test.tsx
+++ b/src/app/(dashboard)/dashboard/providers/[id]/__tests__/phase1f.test.tsx
@@ -129,6 +129,7 @@ describe("useProviderConnections — initial state", () => {
       "fetchProxyConfig",
       "handleUpdateConnectionStatus",
       "handleToggleRateLimit",
+      "handleToggleQuotaVisibility",
       "handleToggleClaudeExtraUsage",
       "handleToggleCodexLimit",
       "handleToggleCliproxyapiMode",
@@ -334,10 +335,7 @@ describe("useProviderModels — initial state", () => {
 // Resolve the hooks dir from the repo root (vitest runs from cwd). Was a
 // hardcoded absolute worktree path that broke the test outside that worktree
 // (#3501 Phase 1g-1j).
-const HOOKS_DIR = path.join(
-  process.cwd(),
-  "src/app/(dashboard)/dashboard/providers/[id]/hooks"
-);
+const HOOKS_DIR = path.join(process.cwd(), "src/app/(dashboard)/dashboard/providers/[id]/hooks");
 
 describe("Cycle-safety — hooks do not import ProviderDetailPageClient", () => {
   // We allow the name in JSDoc comments; what we forbid is an actual ES import statement.
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionRow.tsx b/src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionRow.tsx
index 72e9229c02b..6a9ff815a06 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionRow.tsx
+++ b/src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionRow.tsx
@@ -21,6 +21,7 @@ import {
   ERROR_TYPE_LABELS,
 } from "../providerPageHelpers";
 import { getCodexPlanLabel } from "../codexPlanLabel";
+import ProviderQuotaVisibilityToggle from "./ProviderQuotaVisibilityToggle";
 
 // ---------------------------------------------------------------------------
 // Types (exported so the client can reference them without re-importing)
@@ -48,6 +49,7 @@ export interface ConnectionRowConnection {
   authType?: string;
   proxyEnabled?: boolean;
   perKeyProxyEnabled?: boolean;
+  quotaVisible?: boolean;
 }
 
 export interface ConnectionRowProps {
@@ -64,6 +66,7 @@ export interface ConnectionRowProps {
   onMoveDown: () => void;
   onToggleActive: (isActive?: boolean) => void | Promise<void>;
   onToggleRateLimit: (enabled?: boolean) => void;
+  onToggleQuotaVisibility?: (visible: boolean) => void;
   onToggleClaudeExtraUsage?: (enabled?: boolean) => void;
   onToggleCodex5h?: (enabled?: boolean) => void;
   onToggleCodexWeekly?: (enabled?: boolean) => void;
@@ -347,6 +350,7 @@ export default function ConnectionRow({
   onMoveDown,
   onToggleActive,
   onToggleRateLimit,
+  onToggleQuotaVisibility,
   onToggleClaudeExtraUsage,
   onToggleCodex5h,
   onToggleCodexWeekly,
@@ -447,6 +451,7 @@ export default function ConnectionRow({
 
   const statusPresentation = getStatusPresentation(connection, effectiveStatus, isCooldown, t);
   const rateLimitEnabled = !!connection.rateLimitProtection;
+  const quotaVisible = connection.quotaVisible !== false;
   const codexPolicy =
     connection.providerSpecificData &&
     typeof connection.providerSpecificData === "object" &&
@@ -623,6 +628,12 @@ export default function ConnectionRow({
               <span className="material-symbols-outlined text-[13px]">shield</span>
               {rateLimitEnabled ? t("rateLimitProtected") : t("rateLimitUnprotected")}
             </button>
+            {onToggleQuotaVisibility && (
+              <ProviderQuotaVisibilityToggle
+                visible={quotaVisible}
+                onToggle={onToggleQuotaVisibility}
+              />
+            )}
             {isClaude && (
               <>
                 <span className="text-text-muted/30 select-none">|</span>
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionsListPanel.tsx b/src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionsListPanel.tsx
index 266ac8707d3..fc1de3069a6 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionsListPanel.tsx
+++ b/src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionsListPanel.tsx
@@ -7,6 +7,7 @@ import { pickDisplayValue } from "@/shared/utils/maskEmail";
 import { readBooleanToggle, providerCountText } from "../providerPageHelpers";
 import { compareTr } from "@/shared/utils/turkishText";
 import type { CodexGlobalServiceMode } from "@/lib/providers/codexFastTier";
+import { supportsProviderQuota } from "@/shared/utils/providerQuotaVisibility";
 import type { ConnectionDeleteConfirmState } from "../hooks/useConnectionDeleteConfirm";
 
 type ConnectionsListPanelProps = {
@@ -41,6 +42,7 @@ type ConnectionsListPanelProps = {
   deleteConfirm: ConnectionDeleteConfirmState;
   handleUpdateConnectionStatus: (id: string, isActive: boolean) => void;
   handleToggleRateLimit: (id: string, enabled: boolean) => void;
+  handleToggleQuotaVisibility: (id: string, visible: boolean) => void;
   handleToggleClaudeExtraUsage: (id: string, enabled: boolean) => void;
   handleToggleCliproxyapiMode: (id: string, enabled: boolean) => void;
   handleToggleCodexLimit: (id: string, type: "use5h" | "useWeekly", enabled: boolean) => void;
@@ -68,6 +70,14 @@ type ConnectionsListPanelProps = {
   t: any; // ProviderMessageTranslator
 };
 
+function quotaVisibilityHandler(
+  supported: boolean,
+  connectionId: string,
+  toggle: (id: string, visible: boolean) => void
+) {
+  return supported ? (visible: boolean) => toggle(connectionId, visible) : undefined;
+}
+
 export default function ConnectionsListPanel({
   connections,
   providerId,
@@ -98,6 +108,7 @@ export default function ConnectionsListPanel({
   deleteConfirm,
   handleUpdateConnectionStatus,
   handleToggleRateLimit,
+  handleToggleQuotaVisibility,
   handleToggleClaudeExtraUsage,
   handleToggleCliproxyapiMode,
   handleToggleCodexLimit,
@@ -124,6 +135,7 @@ export default function ConnectionsListPanel({
   t,
 }: ConnectionsListPanelProps) {
   const sorted = [...connections].sort((a, b) => (a.priority || 0) - (b.priority || 0));
+  const quotaSupported = supportsProviderQuota(providerId);
   const hasAnyTag = sorted.some((c) => c.providerSpecificData?.tag as string | undefined);
   const allSelected = selectedIds.size === connections.length && connections.length > 0;
   const someSelected = selectedIds.size > 0 && selectedIds.size < connections.length;
@@ -332,6 +344,11 @@ export default function ConnectionsListPanel({
                 onMoveDown={() => handleSwapPriority(conn, sorted[index + 1])}
                 onToggleActive={(isActive) => handleUpdateConnectionStatus(conn.id, isActive)}
                 onToggleRateLimit={(enabled) => handleToggleRateLimit(conn.id, enabled)}
+                onToggleQuotaVisibility={quotaVisibilityHandler(
+                  quotaSupported,
+                  conn.id,
+                  handleToggleQuotaVisibility
+                )}
                 onToggleClaudeExtraUsage={(enabled) =>
                   handleToggleClaudeExtraUsage(conn.id, enabled)
                 }
@@ -504,6 +521,11 @@ export default function ConnectionsListPanel({
                     onMoveDown={() => handleSwapPriority(conn, sorted[sorted.indexOf(conn) + 1])}
                     onToggleActive={(isActive) => handleUpdateConnectionStatus(conn.id, isActive)}
                     onToggleRateLimit={(enabled) => handleToggleRateLimit(conn.id, enabled)}
+                    onToggleQuotaVisibility={quotaVisibilityHandler(
+                      quotaSupported,
+                      conn.id,
+                      handleToggleQuotaVisibility
+                    )}
                     onToggleClaudeExtraUsage={(enabled) =>
                       handleToggleClaudeExtraUsage(conn.id, enabled)
                     }
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/components/ProviderQuotaVisibilityToggle.tsx b/src/app/(dashboard)/dashboard/providers/[id]/components/ProviderQuotaVisibilityToggle.tsx
new file mode 100644
index 00000000000..cebeb3c7550
--- /dev/null
+++ b/src/app/(dashboard)/dashboard/providers/[id]/components/ProviderQuotaVisibilityToggle.tsx
@@ -0,0 +1,43 @@
+"use client";
+
+import { useTranslations } from "next-intl";
+
+import { providerText } from "../providerPageHelpers";
+
+interface ProviderQuotaVisibilityToggleProps {
+  visible: boolean;
+  onToggle: (visible: boolean) => void;
+}
+
+export default function ProviderQuotaVisibilityToggle({
+  visible,
+  onToggle,
+}: ProviderQuotaVisibilityToggleProps) {
+  const t = useTranslations("providers");
+  const actionLabel = visible
+    ? providerText(t, "hideConnectionFromProviderQuota", "Hide this account from Provider Quota")
+    : providerText(t, "showConnectionInProviderQuota", "Show this account in Provider Quota");
+
+  return (
+    <>
+      <span className="text-text-muted/30 select-none">|</span>
+      <button
+        type="button"
+        onClick={() => onToggle(!visible)}
+        aria-pressed={visible}
+        aria-label={actionLabel}
+        className={`inline-flex items-center gap-1 px-1.5 py-0.5 rounded text-xs font-medium transition-all cursor-pointer ${
+          visible
+            ? "bg-sky-500/15 text-sky-500 hover:bg-sky-500/25"
+            : "bg-black/[0.03] dark:bg-white/[0.03] text-text-muted/50 hover:text-text-muted hover:bg-black/[0.06] dark:hover:bg-white/[0.06]"
+        }`}
+        title={actionLabel}
+      >
+        <span className="material-symbols-outlined text-[13px]">
+          {visible ? "visibility" : "visibility_off"}
+        </span>
+        {providerText(t, "providerQuotaShort", "Quota")}
+      </button>
+    </>
+  );
+}
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/components/__tests__/phase1d.test.tsx b/src/app/(dashboard)/dashboard/providers/[id]/components/__tests__/phase1d.test.tsx
index ae3cdacb471..577b292c4b9 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/components/__tests__/phase1d.test.tsx
+++ b/src/app/(dashboard)/dashboard/providers/[id]/components/__tests__/phase1d.test.tsx
@@ -119,6 +119,31 @@ describe("phase-1d extractions (#3501)", () => {
     expect(c).toBeDefined();
   });
 
+  it("ConnectionRow toggles Provider Quota visibility", () => {
+    const onToggleQuotaVisibility = vi.fn();
+    const c = renderComponent(
+      <ConnectionRow
+        connection={{ id: "conn-quota", name: "Hidden quota", quotaVisible: false }}
+        isOAuth={false}
+        isFirst={true}
+        isLast={true}
+        onMoveUp={vi.fn()}
+        onMoveDown={vi.fn()}
+        onToggleActive={vi.fn()}
+        onToggleRateLimit={vi.fn()}
+        onToggleQuotaVisibility={onToggleQuotaVisibility}
+        onRetest={vi.fn()}
+        onEdit={vi.fn()}
+        onDelete={vi.fn()}
+      />
+    );
+
+    const button = c.querySelector('button[aria-pressed="false"]') as HTMLButtonElement;
+    expect(button).not.toBeNull();
+    act(() => button.click());
+    expect(onToggleQuotaVisibility).toHaveBeenCalledWith(true);
+  });
+
   it("ConnectionRow renders cooldown badge when rateLimitedUntil is in the future", () => {
     const conn = {
       id: "conn-3",
@@ -180,22 +205,14 @@ describe("phase-1d extractions (#3501)", () => {
 
   it("SiliconFlowEndpointModal mounts when isOpen=false (renders nothing visible)", () => {
     const c = renderComponent(
-      <SiliconFlowEndpointModal
-        isOpen={false}
-        onSelect={vi.fn()}
-        onClose={vi.fn()}
-      />
+      <SiliconFlowEndpointModal isOpen={false} onSelect={vi.fn()} onClose={vi.fn()} />
     );
     expect(c).toBeDefined();
   });
 
   it("SiliconFlowEndpointModal mounts when isOpen=true without throwing", () => {
     const c = renderComponent(
-      <SiliconFlowEndpointModal
-        isOpen={true}
-        onSelect={vi.fn()}
-        onClose={vi.fn()}
-      />
+      <SiliconFlowEndpointModal isOpen={true} onSelect={vi.fn()} onClose={vi.fn()} />
     );
     expect(c).toBeDefined();
   });
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderConnections.ts b/src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderConnections.ts
index fc358c358d2..8d6b9f3109c 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderConnections.ts
+++ b/src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderConnections.ts
@@ -27,6 +27,7 @@ import { useNotificationStore } from "@/store/notificationStore";
 import { isClaudeCodeCompatibleProvider } from "@/shared/constants/providers";
 import type { ConnectionRowConnection } from "../components/ConnectionRow";
 import { normalizeCodexLimitPolicy } from "../providerPageHelpers";
+import { useProviderQuotaVisibility } from "./useProviderQuotaVisibility";
 import { useReorderByAvailability } from "./useReorderByAvailability";
 import {
   useConnectionDeleteConfirm,
@@ -74,8 +75,7 @@ export interface UseProviderConnectionsReturn {
   setBatchTestResults: (r: BatchTestResults) => void;
   setConnections: (
     updater:
-      | ConnectionRowConnection[]
-      | ((prev: ConnectionRowConnection[]) => ConnectionRowConnection[])
+      ConnectionRowConnection[] | ((prev: ConnectionRowConnection[]) => ConnectionRowConnection[])
   ) => void;
   setProviderNode: (node: any) => void;
 
@@ -87,6 +87,7 @@ export interface UseProviderConnectionsReturn {
   deleteConfirm: ConnectionDeleteConfirmState;
   handleUpdateConnectionStatus: (id: string, isActive: boolean) => Promise<void>;
   handleToggleRateLimit: (connectionId: string, enabled: boolean) => Promise<void>;
+  handleToggleQuotaVisibility: (connectionId: string, visible: boolean) => Promise<void>;
   handleToggleClaudeExtraUsage: (connectionId: string, enabled: boolean) => Promise<void>;
   handleToggleCodexLimit: (connectionId: string, field: string, enabled: boolean) => Promise<void>;
   handleToggleCliproxyapiMode: (connectionId: string, enabled: boolean) => Promise<void>;
@@ -135,6 +136,7 @@ export function useProviderConnections(
 
   // ── core state ──────────────────────────────────────────────────────────
   const [connections, setConnections] = useState<ConnectionRowConnection[]>([]);
+  const handleToggleQuotaVisibility = useProviderQuotaVisibility(setConnections, notify, t);
   const [providerNode, setProviderNode] = useState<any>(null);
   const [loading, setLoading] = useState(true);
 
@@ -894,6 +896,7 @@ export function useProviderConnections(
     deleteConfirm,
     handleUpdateConnectionStatus,
     handleToggleRateLimit,
+    handleToggleQuotaVisibility,
     handleToggleClaudeExtraUsage,
     handleToggleCodexLimit,
     handleToggleCliproxyapiMode,
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderQuotaVisibility.ts b/src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderQuotaVisibility.ts
new file mode 100644
index 00000000000..dda6f29564c
--- /dev/null
+++ b/src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderQuotaVisibility.ts
@@ -0,0 +1,45 @@
+"use client";
+
+import { useCallback, type Dispatch, type SetStateAction } from "react";
+
+import type { ConnectionRowConnection } from "../components/ConnectionRow";
+import { providerText, type ProviderMessageTranslator } from "../providerPageHelpers";
+
+interface NotificationStore {
+  error: (message: string) => void;
+}
+
+export function useProviderQuotaVisibility(
+  setConnections: Dispatch<SetStateAction<ConnectionRowConnection[]>>,
+  notify: NotificationStore,
+  t: ProviderMessageTranslator
+) {
+  return useCallback(
+    async (connectionId: string, visible: boolean) => {
+      try {
+        const response = await fetch(`/api/providers/${connectionId}`, {
+          method: "PUT",
+          headers: { "Content-Type": "application/json" },
+          body: JSON.stringify({ quotaVisible: visible }),
+        });
+        if (!response.ok) throw new Error(`HTTP ${response.status}`);
+
+        setConnections((previous) =>
+          previous.map((connection) =>
+            connection.id === connectionId ? { ...connection, quotaVisible: visible } : connection
+          )
+        );
+      } catch (error) {
+        console.error("Error toggling provider quota visibility:", error);
+        notify.error(
+          providerText(
+            t,
+            "quotaVisibilityUpdateFailed",
+            "Failed to update Provider Quota visibility"
+          )
+        );
+      }
+    },
+    [notify, setConnections, t]
+  );
+}
diff --git a/src/app/(dashboard)/dashboard/settings/components/ProxyTab.tsx b/src/app/(dashboard)/dashboard/settings/components/ProxyTab.tsx
index f08fa359fdb..3cfc2902b91 100644
--- a/src/app/(dashboard)/dashboard/settings/components/ProxyTab.tsx
+++ b/src/app/(dashboard)/dashboard/settings/components/ProxyTab.tsx
@@ -9,11 +9,11 @@ import DocumentationTab from "./proxy/DocumentationTab";
 
 type TabId = "global-config" | "proxy-pool" | "free-pool" | "documentation";
 
-const TABS: Array<{ id: TabId; labelKey: string }> = [
-  { id: "global-config", labelKey: "proxyGlobalConfigTab" },
-  { id: "proxy-pool", labelKey: "proxyPoolTab" },
-  { id: "free-pool", labelKey: "freePoolTab" },
-  { id: "documentation", labelKey: "proxyDocumentationTab" },
+const TABS: Array<{ id: TabId; labelKey: string; fallback: string }> = [
+  { id: "global-config", labelKey: "proxyGlobalConfigTab", fallback: "Global config" },
+  { id: "proxy-pool", labelKey: "proxyPoolTab", fallback: "Proxy pool" },
+  { id: "free-pool", labelKey: "freePoolTab", fallback: "Free pool" },
+  { id: "documentation", labelKey: "proxyDocumentationTab", fallback: "Documentation" },
 ];
 
 export default function ProxyTab() {
@@ -21,6 +21,8 @@ export default function ProxyTab() {
   const searchParams = useSearchParams();
   const router = useRouter();
   const pathname = usePathname();
+  const translateOrFallback = (key: string, fallback: string) =>
+    typeof t.has === "function" && !t.has(key) ? fallback : t(key);
 
   const activeTab = useMemo<TabId>(() => {
     const tabParam = searchParams.get("tab") as TabId | null;
@@ -38,7 +40,7 @@ export default function ProxyTab() {
       <div
         className="flex gap-1 border-b border-border overflow-x-auto"
         role="tablist"
-        aria-label={t("proxySubTabsAria")}
+        aria-label={translateOrFallback("proxySubTabsAria", "Proxy sections")}
       >
         {TABS.map((tab) => (
           <button
@@ -52,7 +54,7 @@ export default function ProxyTab() {
                 : "border-transparent text-text-muted hover:text-text"
             }`}
           >
-            {t(tab.labelKey)}
+            {translateOrFallback(tab.labelKey, tab.fallback)}
           </button>
         ))}
       </div>
diff --git a/src/app/(dashboard)/dashboard/settings/components/SidebarTab.tsx b/src/app/(dashboard)/dashboard/settings/components/SidebarTab.tsx
index 32cebb1e9be..d3fc9652232 100644
--- a/src/app/(dashboard)/dashboard/settings/components/SidebarTab.tsx
+++ b/src/app/(dashboard)/dashboard/settings/components/SidebarTab.tsx
@@ -36,7 +36,9 @@ import {
   applySectionOrder,
   applyItemOrder,
   normalizeHiddenSidebarItems,
+  HIDEABLE_SIDEBAR_ITEM_IDS,
   type HideableSidebarItemId,
+  type SidebarItemId,
   type SidebarSectionId,
   type SidebarItemOrder,
   type SidebarPresetId,
@@ -215,11 +217,48 @@ interface ItemRowProps {
 }
 
 // Items that must always remain visible (safety guard)
-const PROTECTED_ITEM_IDS = new Set(["settings-sidebar"]);
+const PROTECTED_ITEM_IDS = new Set<SidebarItemId>(["proxy", "settings-sidebar"]);
+
+function isHideableSidebarItemId(id: SidebarItemId): id is HideableSidebarItemId {
+  return HIDEABLE_SIDEBAR_ITEM_IDS.includes(id as HideableSidebarItemId);
+}
+
+function GroupItemVisibilityControl({
+  item,
+  hiddenSet,
+  onToggleItem,
+}: {
+  item: SidebarItemDefinition;
+  hiddenSet: Set<HideableSidebarItemId>;
+  onToggleItem: (id: HideableSidebarItemId) => void;
+}) {
+  const tSidebar = useTranslations("sidebar");
+  const hideableId = isHideableSidebarItemId(item.id) ? item.id : null;
+  if (hideableId !== null) {
+    return (
+      <Toggle
+        size="sm"
+        checked={!hiddenSet.has(hideableId)}
+        onChange={() => onToggleItem(hideableId)}
+      />
+    );
+  }
+
+  return (
+    <span
+      className="material-symbols-outlined text-[16px] text-text-muted/40"
+      title={tSidebar("cannotHide")}
+      aria-label={tSidebar("alwaysVisible")}
+    >
+      lock
+    </span>
+  );
+}
 
 function ItemRow({ item, hiddenSet, onToggleItem, getLabel }: ItemRowProps) {
   const tSidebar = useTranslations("sidebar");
-  const isProtected = PROTECTED_ITEM_IDS.has(item.id);
+  const hideableId = isHideableSidebarItemId(item.id) ? item.id : null;
+  const isProtected = PROTECTED_ITEM_IDS.has(item.id) || hideableId === null;
   return (
     <div className="flex items-center justify-between gap-4 px-4 py-3">
       <div className="flex items-center gap-2 min-w-0">
@@ -237,7 +276,7 @@ function ItemRow({ item, hiddenSet, onToggleItem, getLabel }: ItemRowProps) {
           lock
         </span>
       ) : (
-        <Toggle checked={!hiddenSet.has(item.id)} onChange={() => onToggleItem(item.id)} />
+        <Toggle checked={!hiddenSet.has(hideableId)} onChange={() => onToggleItem(hideableId)} />
       )}
     </div>
   );
@@ -288,7 +327,8 @@ function GroupRow({
           </span>
         </button>
         <span className="text-xs text-text-muted/40">
-          {group.items.filter((i) => !hiddenSet.has(i.id)).length}/{group.items.length}
+          {group.items.filter((i) => !isHideableSidebarItemId(i.id) || !hiddenSet.has(i.id)).length}/
+          {group.items.length}
         </span>
         {canToggleSeparator && (
           <div className="flex items-center gap-2 border-l border-border/60 pl-3">
@@ -311,10 +351,10 @@ function GroupRow({
                 </span>
                 <p className="text-sm font-medium truncate">{getLabel(item.i18nKey, item.id)}</p>
               </div>
-              <Toggle
-                size="sm"
-                checked={!hiddenSet.has(item.id)}
-                onChange={() => onToggleItem(item.id)}
+              <GroupItemVisibilityControl
+                item={item}
+                hiddenSet={hiddenSet}
+                onToggleItem={onToggleItem}
               />
             </div>
           ))}
@@ -418,11 +458,14 @@ export default function SidebarTab() {
     patch({ [HIDDEN_SIDEBAR_GROUP_LABELS_SETTING_KEY]: next, [SIDEBAR_PRESET_KEY]: null });
   };
 
-  const visibleSections = SIDEBAR_SECTIONS.filter((s) => s.visibility !== "debug" || showDebug).map(
-    (s) => ({ ...s, title: getLabel(s.titleKey, s.titleFallback) })
+  const visibleSections = SIDEBAR_SECTIONS.filter(
+    (s) => s.visibility !== "debug" || showDebug
   );
 
-  const orderedSections = applySectionOrder(visibleSections, sectionOrder);
+  const orderedSections = applySectionOrder(visibleSections, sectionOrder).map((s) => ({
+    ...s,
+    title: getLabel(s.titleKey, s.titleFallback),
+  }));
 
   const sectionIds = orderedSections.map((s) => s.id);
 
diff --git a/src/app/(dashboard)/dashboard/settings/components/proxy/FreePoolTab.tsx b/src/app/(dashboard)/dashboard/settings/components/proxy/FreePoolTab.tsx
index a30aa5d7965..1c51952c27f 100644
--- a/src/app/(dashboard)/dashboard/settings/components/proxy/FreePoolTab.tsx
+++ b/src/app/(dashboard)/dashboard/settings/components/proxy/FreePoolTab.tsx
@@ -22,6 +22,11 @@ const MAX_VISIBLE_PAGES = 7;
 
 export default function FreePoolTab() {
   const t = useTranslations("settings");
+  const translateOrFallback = (
+    key: string,
+    fallback: string,
+    values?: Record<string, string | number>
+  ) => (typeof t.has === "function" && !t.has(key) ? fallback : t(key, values));
   const [proxies, setProxies] = useState<FreeProxyRowData[]>([]);
   const [stats, setStats] = useState<FreePoolStats | null>(null);
   const [disabledSources, setDisabledSources] = useState<Set<SourceId>>(new Set());
@@ -106,9 +111,19 @@ export default function FreePoolTab() {
         headers: { "Content-Type": "application/json" },
         body: JSON.stringify(body),
       });
-      const data = await res.json();
-      if (!res.ok) {
-        setSyncErrors(data.errors ?? null);
+      const data = await res.json().catch(() => null);
+      if (data?.results) {
+        const errors: Record<string, string[]> = {};
+        for (const [source, result] of Object.entries(
+          data.results as Record<string, { errors?: string[] }>
+        )) {
+          if (Array.isArray(result?.errors) && result.errors.length > 0) {
+            errors[source] = result.errors;
+          }
+        }
+        if (Object.keys(errors).length > 0) setSyncErrors(errors);
+      } else if (!res.ok) {
+        setSyncErrors(data?.errors ?? null);
       }
       await loadData();
     } catch {}
@@ -222,9 +237,9 @@ export default function FreePoolTab() {
             value={filterProtocol}
             onChange={(e) => setFilterProtocol(e.target.value)}
             className="text-xs bg-surface-alt border border-border rounded px-2 py-1"
-            aria-label={t("proxyFreePoolFilterProtocol")}
+            aria-label={translateOrFallback("proxyFreePoolFilterProtocol", "Filter by protocol")}
           >
-            <option value="">{t("proxyFreePoolProtocol")}</option>
+            <option value="">{translateOrFallback("proxyFreePoolProtocol", "Protocol")}</option>
             {["http", "https", "socks4", "socks5"].map((p) => (
               <option key={p} value={p}>
                 {p.toUpperCase()}
@@ -233,24 +248,32 @@ export default function FreePoolTab() {
           </select>
           <input
             type="text"
-            placeholder={t("proxyFreePoolCountryPlaceholder")}
+            placeholder={translateOrFallback(
+              "proxyFreePoolCountryPlaceholder",
+              "Country (e.g. US)"
+            )}
             value={filterCountry}
             onChange={(e) => setFilterCountry(e.target.value.toUpperCase().slice(0, 2))}
             className="text-xs bg-surface-alt border border-border rounded px-2 py-1 w-28"
-            aria-label={t("proxyFreePoolFilterCountry")}
+            aria-label={translateOrFallback("proxyFreePoolFilterCountry", "Filter by country")}
           />
           <input
             type="number"
-            placeholder={t("proxyFreePoolMinQualityPlaceholder")}
+            placeholder={translateOrFallback("proxyFreePoolMinQualityPlaceholder", "Min quality")}
             value={minQuality}
             onChange={(e) => setMinQuality(e.target.value)}
             min={0}
             max={100}
             className="text-xs bg-surface-alt border border-border rounded px-2 py-1 w-24"
-            aria-label={t("proxyFreePoolMinQualityLabel")}
+            aria-label={translateOrFallback(
+              "proxyFreePoolMinQualityLabel",
+              "Minimum quality score"
+            )}
           />
           <Button size="sm" variant="secondary" icon="sync" onClick={handleSync} disabled={syncing}>
-            {syncing ? t("syncing") : t("proxyFreePoolSyncAll")}
+            {syncing
+              ? translateOrFallback("syncing", "Syncing...")
+              : translateOrFallback("proxyFreePoolSyncAll", "Sync all")}
           </Button>
         </div>
       </div>
@@ -258,19 +281,21 @@ export default function FreePoolTab() {
       {stats && (
         <div className="text-xs text-text-muted flex gap-4 flex-wrap">
           <span>
-            {t("proxyFreePoolTotal")}: {stats.total}
+            {translateOrFallback("proxyFreePoolTotal", "Total")}: {stats.total}
           </span>
           <span>
-            {t("proxyFreePoolInPool")}: {stats.inPool}
+            {translateOrFallback("proxyFreePoolInPool", "In pool")}: {stats.inPool}
           </span>
           {stats.avgQuality != null && (
             <span>
-              {t("proxyFreePoolAvgQuality")}: {stats.avgQuality}
+              {translateOrFallback("proxyFreePoolAvgQuality", "Average quality")}:{" "}
+              {stats.avgQuality}
             </span>
           )}
           {stats.lastSyncAt && (
-            <span>
-              {t("lastSync")}: {new Date(stats.lastSyncAt).toLocaleTimeString()}
+            <span suppressHydrationWarning>
+              {translateOrFallback("lastSync", "Last sync")}:{" "}
+              {new Date(stats.lastSyncAt).toLocaleTimeString()}
             </span>
           )}
         </div>
@@ -292,9 +317,13 @@ export default function FreePoolTab() {
 
       {selected.size > 0 && (
         <div className="flex items-center gap-2 p-2 bg-primary/10 rounded border border-primary/20">
-          <span className="text-xs">{t("proxyFreePoolSelected", { count: selected.size })}</span>
+          <span className="text-xs">
+            {translateOrFallback("proxyFreePoolSelected", `${selected.size} selected`, {
+              count: selected.size,
+            })}
+          </span>
           <Button size="sm" variant="primary" onClick={() => handleBulkAdd(Array.from(selected))}>
-            {t("proxyFreePoolAddSelected")}
+            {translateOrFallback("proxyFreePoolAddSelected", "Add selected to pool")}
           </Button>
           {bulkProgress && <span className="text-xs text-text-muted">{bulkProgress}</span>}
         </div>
@@ -307,7 +336,7 @@ export default function FreePoolTab() {
             variant="secondary"
             onClick={() => handleBulkAdd(notInPoolProxies.slice(0, 100).map((p) => p.id))}
           >
-            {t("proxyFreePoolAddVisible")}
+            {translateOrFallback("proxyFreePoolAddVisible", "Add all visible to pool")}
           </Button>
         </div>
       )}
@@ -318,22 +347,22 @@ export default function FreePoolTab() {
             <tr>
               <th className="px-3 py-2 text-left w-8" scope="col"></th>
               <th className="px-3 py-2 text-left" scope="col">
-                {t("proxyFreePoolSource")}
+                {translateOrFallback("proxyFreePoolSource", "Source")}
               </th>
               <th className="px-3 py-2 text-left" scope="col">
-                {t("proxyFreePoolHostPort")}
+                {translateOrFallback("proxyFreePoolHostPort", "Host:Port")}
               </th>
               <th className="px-3 py-2 text-left" scope="col">
-                {t("proxyFreePoolType")}
+                {translateOrFallback("proxyFreePoolType", "Type")}
               </th>
               <th className="px-3 py-2 text-left" scope="col">
-                {t("proxyFreePoolCountry")}
+                {translateOrFallback("proxyFreePoolCountry", "Country")}
               </th>
               <th className="px-3 py-2 text-left" scope="col">
-                {t("proxyFreePoolQuality")}
+                {translateOrFallback("proxyFreePoolQuality", "Quality")}
               </th>
               <th className="px-3 py-2 text-left" scope="col">
-                {t("proxyFreePoolLatency")}
+                {translateOrFallback("proxyFreePoolLatency", "Latency")}
               </th>
               <th className="px-3 py-2 text-left" scope="col"></th>
             </tr>
@@ -342,13 +371,16 @@ export default function FreePoolTab() {
             {loading ? (
               <tr>
                 <td colSpan={8} className="px-3 py-8 text-center text-text-muted">
-                  {t("loading")}
+                  {translateOrFallback("loading", "Loading...")}
                 </td>
               </tr>
             ) : proxies.length === 0 ? (
               <tr>
                 <td colSpan={8} className="px-3 py-8 text-center text-text-muted">
-                  {t("proxyFreePoolEmpty")}
+                  {translateOrFallback(
+                    "proxyFreePoolEmpty",
+                    "No proxies found. Click Sync all to fetch from sources."
+                  )}
                 </td>
               </tr>
             ) : (
diff --git a/src/app/(dashboard)/dashboard/usage/components/EvalsTab.tsx b/src/app/(dashboard)/dashboard/usage/components/EvalsTab.tsx
index 6b417efd98a..bc5922e8be3 100644
--- a/src/app/(dashboard)/dashboard/usage/components/EvalsTab.tsx
+++ b/src/app/(dashboard)/dashboard/usage/components/EvalsTab.tsx
@@ -1,6 +1,6 @@
 "use client";
 
-import { useEffect, useState } from "react";
+import { memo, useEffect, useState } from "react";
 import { useTranslations } from "next-intl";
 import {
   Card,
@@ -1846,7 +1846,7 @@ export default function EvalsTab() {
   );
 }
 
-function HeroSection({ t }: { t: (key: string, values?: Record<string, unknown>) => string }) {
+const HeroSection = memo(function HeroSection({ t }: { t: (key: string, values?: Record<string, unknown>) => string }) {
   return (
     <Card className="p-0 overflow-hidden">
       <div
@@ -1892,7 +1892,7 @@ function HeroSection({ t }: { t: (key: string, values?: Record<string, unknown>)
       </div>
     </Card>
   );
-}
+});
 
 function SuiteBuilderModal({
   draft,
diff --git a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/CodexResetCreditsModal.tsx b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/CodexResetCreditsModal.tsx
new file mode 100644
index 00000000000..5817c6fa013
--- /dev/null
+++ b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/CodexResetCreditsModal.tsx
@@ -0,0 +1,277 @@
+"use client";
+
+import { useMemo, useState } from "react";
+import { useTranslations } from "next-intl";
+
+import Button from "@/shared/components/Button";
+import Modal from "@/shared/components/Modal";
+import { translateUsageOrFallback, type UsageTranslationValues } from "./i18nFallback";
+import type { CodexResetCreditView } from "./useCodexResetCreditRedemption";
+
+interface Props {
+  credits: CodexResetCreditView[];
+  availableCount: number;
+  isOpen: boolean;
+  loading: boolean;
+  onClose: () => void;
+  onRedeem: (selectionToken: string) => Promise<void>;
+}
+
+function formatRelativeExpiry(expiresAt: string | null | undefined): string | null {
+  if (!expiresAt) return null;
+  const diffMs = new Date(expiresAt).getTime() - Date.now();
+  if (!Number.isFinite(diffMs)) return null;
+  if (diffMs <= 0) return "0m";
+
+  const minutes = Math.max(1, Math.ceil(diffMs / 60_000));
+  if (minutes < 60) return `${minutes}m`;
+  const hours = Math.ceil(minutes / 60);
+  if (hours < 24) return `${hours}h`;
+  return `${Math.ceil(hours / 24)}d`;
+}
+
+export function getCodexResetCreditExpiryLabel(
+  expiresAt: string | null | undefined,
+  locale?: string
+): { absolute: string | null; relative: string | null } {
+  if (!expiresAt) return { absolute: null, relative: null };
+  const date = new Date(expiresAt);
+  if (!Number.isFinite(date.getTime())) return { absolute: null, relative: null };
+  return {
+    absolute: date.toLocaleString(locale, { dateStyle: "medium", timeStyle: "short" }),
+    relative: formatRelativeExpiry(expiresAt),
+  };
+}
+
+function ResetCreditModalFooter({
+  confirming,
+  loading,
+  onBack,
+  onClose,
+  onConfirm,
+  tr,
+}: {
+  confirming: boolean;
+  loading: boolean;
+  onBack: () => void;
+  onClose: () => void;
+  onConfirm: () => void;
+  tr: (key: string, fallback: string, values?: UsageTranslationValues) => string;
+}) {
+  if (!confirming) {
+    return (
+      <Button variant="ghost" onClick={onClose} disabled={loading}>
+        {tr("close", "Close")}
+      </Button>
+    );
+  }
+
+  return (
+    <>
+      <Button variant="ghost" onClick={onBack} disabled={loading}>
+        {tr("back", "Back")}
+      </Button>
+      <Button onClick={onConfirm} loading={loading}>
+        {tr("confirmRedeemResetCreditButton", "Redeem credit")}
+      </Button>
+    </>
+  );
+}
+
+function ResetCreditConfirmation({
+  credit,
+  tr,
+}: {
+  credit: CodexResetCreditView;
+  tr: (key: string, fallback: string, values?: UsageTranslationValues) => string;
+}) {
+  return (
+    <div className="space-y-4">
+      <div className="rounded-lg border border-amber-500/30 bg-amber-500/10 p-4">
+        <div className="flex items-start gap-3">
+          <span className="material-symbols-outlined text-amber-500">warning</span>
+          <div>
+            <p className="font-semibold text-text-main">
+              {tr("confirmRedeemResetCreditTitle", "Redeem this reset credit?")}
+            </p>
+            <p className="mt-1 text-sm text-text-muted">
+              {tr(
+                "confirmRedeemResetCredit",
+                "Redeeming immediately resets the eligible Codex usage windows and permanently consumes this credit."
+              )}
+            </p>
+          </div>
+        </div>
+      </div>
+      <CreditSummary credit={credit} tr={tr} />
+    </div>
+  );
+}
+
+function ResetCreditList({
+  availableCount,
+  credits,
+  loading,
+  onSelect,
+  tr,
+}: {
+  availableCount: number;
+  credits: CodexResetCreditView[];
+  loading: boolean;
+  onSelect: (selectionToken: string) => void;
+  tr: (key: string, fallback: string, values?: UsageTranslationValues) => string;
+}) {
+  if (credits.length === 0) {
+    return (
+      <div className="rounded-lg border border-border p-5 text-center text-sm text-text-muted">
+        {availableCount > 0
+          ? tr(
+              "resetCreditsDetailsUnavailable",
+              "Credit details are currently unavailable. Refresh and try again."
+            )
+          : tr("noResetCreditsAvailable", "No reset credits are available.")}
+      </div>
+    );
+  }
+
+  return (
+    <div className="space-y-2">
+      {credits.map((credit, index) => (
+        <div
+          key={credit.selectionToken}
+          className="flex flex-col gap-3 rounded-lg border border-border bg-bg-subtle/40 p-3 sm:flex-row sm:items-center"
+        >
+          <CreditSummary credit={credit} tr={tr} recommended={index === 0} />
+          <Button
+            size="sm"
+            className="shrink-0"
+            onClick={() => onSelect(credit.selectionToken)}
+            disabled={loading}
+          >
+            {tr("redeemThisResetCredit", "Redeem")}
+          </Button>
+        </div>
+      ))}
+    </div>
+  );
+}
+
+export default function CodexResetCreditsModal({
+  credits,
+  availableCount,
+  isOpen,
+  loading,
+  onClose,
+  onRedeem,
+}: Props) {
+  const t = useTranslations("usage");
+  const tr = (key: string, fallback: string, values?: UsageTranslationValues) =>
+    translateUsageOrFallback(t, key, fallback, values);
+  const [selectedCreditId, setSelectedCreditId] = useState<string | null>(null);
+  const [confirming, setConfirming] = useState(false);
+
+  const selectedCredit = useMemo(
+    () => credits.find((credit) => credit.selectionToken === selectedCreditId) ?? null,
+    [credits, selectedCreditId]
+  );
+
+  const close = () => {
+    if (loading) return;
+    setConfirming(false);
+    setSelectedCreditId(null);
+    onClose();
+  };
+
+  const beginRedeem = (selectionToken: string) => {
+    setSelectedCreditId(selectionToken);
+    setConfirming(true);
+  };
+
+  const confirmRedeem = async () => {
+    if (!selectedCreditId) return;
+    await onRedeem(selectedCreditId);
+  };
+
+  return (
+    <Modal
+      isOpen={isOpen}
+      onClose={close}
+      title={tr("resetCreditsModalTitle", "Codex reset credits")}
+      size="lg"
+      closeOnOverlay={!loading}
+      footer={
+        <ResetCreditModalFooter
+          confirming={confirming}
+          loading={loading}
+          onBack={() => setConfirming(false)}
+          onClose={close}
+          onConfirm={confirmRedeem}
+          tr={tr}
+        />
+      }
+    >
+      {confirming && selectedCredit ? (
+        <ResetCreditConfirmation credit={selectedCredit} tr={tr} />
+      ) : (
+        <div className="space-y-4">
+          <p className="text-sm text-text-muted">
+            {tr(
+              "resetCreditsModalExplainer",
+              "Credits are ordered by expiration. Automatic redemption always uses the credit that expires first."
+            )}
+          </p>
+          <ResetCreditList
+            availableCount={availableCount}
+            credits={credits}
+            loading={loading}
+            onSelect={beginRedeem}
+            tr={tr}
+          />
+        </div>
+      )}
+    </Modal>
+  );
+}
+
+function CreditSummary({
+  credit,
+  recommended = false,
+  tr,
+}: {
+  credit: CodexResetCreditView;
+  recommended?: boolean;
+  tr: (key: string, fallback: string, values?: UsageTranslationValues) => string;
+}) {
+  const { absolute: expiry, relative: relativeExpiry } = getCodexResetCreditExpiryLabel(
+    credit.expiresAt
+  );
+
+  return (
+    <div className="min-w-0 flex-1">
+      <div className="flex flex-wrap items-center gap-2">
+        <span className="font-medium text-text-main">
+          {credit.title || tr("resetCreditDefaultTitle", "Full reset")}
+        </span>
+        {recommended && (
+          <span className="rounded-full bg-primary/10 px-2 py-0.5 text-[10px] font-semibold text-primary">
+            {tr("resetCreditExpiresFirst", "Expires first")}
+          </span>
+        )}
+      </div>
+      {credit.description && (
+        <p className="mt-1 text-xs leading-relaxed text-text-muted">{credit.description}</p>
+      )}
+      <div className="mt-2 flex items-center gap-1.5 text-xs text-text-muted">
+        <span className="material-symbols-outlined text-[15px]">schedule</span>
+        {expiry ? (
+          <span>
+            {tr("resetCreditExpiresAt", `Expires ${expiry}`, { date: expiry })}
+            {relativeExpiry ? ` (${relativeExpiry})` : ""}
+          </span>
+        ) : (
+          <span>{tr("resetCreditNoExpiry", "No expiration date")}</span>
+        )}
+      </div>
+    </div>
+  );
+}
diff --git a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/QuotaCard.tsx b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/QuotaCard.tsx
index 7c7833b7f91..97dbcfb68d3 100644
--- a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/QuotaCard.tsx
+++ b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/QuotaCard.tsx
@@ -34,10 +34,11 @@ interface QuotaCardProps {
   providerLabel: string;
   onRefresh: () => void;
   onOpenCutoff: () => void;
-  onRedeemResetCredit?: () => void;
+  onOpenResetCredits?: () => void;
   onToggleActive: (nextActive: boolean) => void;
   togglingActive: boolean;
   redeemingResetCredit?: boolean;
+  loadingResetCredits?: boolean;
 }
 
 export default function QuotaCard({
@@ -50,10 +51,11 @@ export default function QuotaCard({
   providerLabel,
   onRefresh,
   onOpenCutoff,
-  onRedeemResetCredit,
+  onOpenResetCredits,
   onToggleActive,
   togglingActive,
   redeemingResetCredit = false,
+  loadingResetCredits = false,
 }: QuotaCardProps) {
   const isActive = connection.isActive ?? true;
   const [costModalOpen, setCostModalOpen] = useState(false);
@@ -87,9 +89,7 @@ export default function QuotaCard({
   const hasStaleData = !!quota?.stale;
   const displayRefreshedAt = quota?.stale?.since || refreshedAt;
   const canEditCutoff = quotas.some((q: any) => q && typeof q.name === "string" && !q.isCredits);
-  const canRedeemResetCredit =
-    connection.provider === "codex" &&
-    quotas.some((q: any) => q?.isResetCredits && Number(q.creditCount ?? q.remaining ?? 0) > 0);
+  const canRedeemResetCredit = connection.provider === "codex" && hasAvailableResetCredits(quotas);
 
   return (
     <Card
@@ -119,11 +119,12 @@ export default function QuotaCard({
         onRefresh={onRefresh}
         onOpenCutoff={onOpenCutoff}
         onOpenCost={() => setCostModalOpen(true)}
-        onRedeemResetCredit={onRedeemResetCredit}
+        onOpenResetCredits={onOpenResetCredits}
         canEditCutoff={canEditCutoff}
         hasCutoffOverrides={hasOverrides}
         canRedeemResetCredit={canRedeemResetCredit}
         redeemingResetCredit={redeemingResetCredit}
+        loadingResetCredits={loadingResetCredits}
       />
       <ProviderUsdCostModal
         isOpen={costModalOpen}
@@ -135,3 +136,9 @@ export default function QuotaCard({
     </Card>
   );
 }
+
+function hasAvailableResetCredits(quotas: any[]): boolean {
+  return quotas.some(
+    (quota: any) => quota?.isResetCredits && Number(quota.creditCount ?? quota.remaining ?? 0) > 0
+  );
+}
diff --git a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/QuotaCardGrid.tsx b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/QuotaCardGrid.tsx
index 9beaec11dea..dfdc46a9c72 100644
--- a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/QuotaCardGrid.tsx
+++ b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/QuotaCardGrid.tsx
@@ -14,10 +14,11 @@ interface Props {
   renderInlineQuotaSummary?: (quota: any) => ReactNode;
   onRefresh: (id: string, provider: string) => void;
   onOpenCutoff: (connection: any) => void;
-  onRedeemResetCredit?: (id: string, provider: string) => void;
+  onOpenResetCredits?: (id: string, provider: string) => void;
   onToggleActive: (id: string, nextActive: boolean) => void;
   togglingActiveId: string | null;
   redeemingResetCreditId?: string | null;
+  loadingResetCreditsId?: string | null;
 }
 
 export default function QuotaCardGrid({
@@ -31,10 +32,11 @@ export default function QuotaCardGrid({
   renderInlineQuotaSummary: _renderInlineQuotaSummary,
   onRefresh,
   onOpenCutoff,
-  onRedeemResetCredit,
+  onOpenResetCredits,
   onToggleActive,
   togglingActiveId,
   redeemingResetCreditId = null,
+  loadingResetCreditsId = null,
 }: Props) {
   if (connections.length === 0) return null;
 
@@ -69,10 +71,11 @@ export default function QuotaCardGrid({
                 providerLabel={providerLabels[conn.provider] || conn.provider}
                 onRefresh={() => onRefresh(conn.id, conn.provider)}
                 onOpenCutoff={() => onOpenCutoff(conn)}
-                onRedeemResetCredit={() => onRedeemResetCredit?.(conn.id, conn.provider)}
+                onOpenResetCredits={() => onOpenResetCredits?.(conn.id, conn.provider)}
                 onToggleActive={(nextActive) => onToggleActive(conn.id, nextActive)}
                 togglingActive={togglingActiveId === conn.id}
                 redeemingResetCredit={redeemingResetCreditId === conn.id}
+                loadingResetCredits={loadingResetCreditsId === conn.id}
               />
             ))}
           </div>
diff --git a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/index.tsx b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/index.tsx
index 9ca661017c7..2ead860172c 100644
--- a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/index.tsx
+++ b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/index.tsx
@@ -21,6 +21,7 @@ import useEmailPrivacyStore from "@/store/emailPrivacyStore";
 import { useNotificationStore } from "@/store/notificationStore";
 import QuotaCutoffModal from "./QuotaCutoffModal";
 import QuotaCardGrid from "./QuotaCardGrid";
+import CodexResetCreditsModal from "./CodexResetCreditsModal";
 import { useVisibleQuotaData } from "./useVisibleQuotaData";
 import { useCodexResetCreditRedemption } from "./useCodexResetCreditRedemption";
 import { PROVIDER_LABEL, PROVIDER_ORDER, TIER_FILTERS } from "./constants";
@@ -28,6 +29,7 @@ import { formatAutoRefreshCountdown } from "./formatters";
 import { translateUsageOrFallback, type UsageTranslationValues } from "./i18nFallback";
 import { compareTr } from "@/shared/utils/turkishText";
 import { fetchWithTimeout } from "@/shared/utils/fetchTimeout";
+import { isProviderQuotaVisible } from "@/shared/utils/providerQuotaVisibility";
 
 // Bound the two first-paint requests so a stalled connection cannot wedge
 // `initialLoading` on `true` and freeze the quota page on its skeleton forever
@@ -516,6 +518,7 @@ export default function ProviderLimits({
     () =>
       connections.filter(
         (conn) =>
+          isProviderQuotaVisible(conn) &&
           USAGE_SUPPORTED_PROVIDERS.includes(conn.provider) &&
           (conn.authType === "oauth" || conn.authType === "apikey")
       ),
@@ -1034,13 +1037,25 @@ export default function ProviderLimits({
             setCutoffModalWindows(windows);
             setCutoffModalConn(conn);
           }}
-          onRedeemResetCredit={resetCreditRedemption.redeemCodexResetCredit}
+          onOpenResetCredits={resetCreditRedemption.openCodexResetCredits}
           onToggleActive={handleToggleActive}
           togglingActiveId={togglingActiveId}
           redeemingResetCreditId={resetCreditRedemption.redeemingResetCreditId}
+          loadingResetCreditsId={resetCreditRedemption.loadingResetCreditsId}
         />
       </div>
 
+      {resetCreditRedemption.resetCreditPicker && (
+        <CodexResetCreditsModal
+          isOpen={true}
+          credits={resetCreditRedemption.resetCreditPicker.credits}
+          availableCount={resetCreditRedemption.resetCreditPicker.availableCount}
+          loading={resetCreditRedemption.redeemingResetCreditId !== null}
+          onClose={resetCreditRedemption.closeResetCreditPicker}
+          onRedeem={resetCreditRedemption.redeemCodexResetCredit}
+        />
+      )}
+
       {cutoffModalConn && (
         <QuotaCutoffModal
           isOpen={!!cutoffModalConn}
diff --git a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/parts/QuotaCardExpanded.tsx b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/parts/QuotaCardExpanded.tsx
index a672cd978dc..21b0f48a398 100644
--- a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/parts/QuotaCardExpanded.tsx
+++ b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/parts/QuotaCardExpanded.tsx
@@ -60,14 +60,23 @@ interface Props {
   onRefresh: () => void;
   onOpenCutoff: () => void;
   onOpenCost: () => void;
-  onRedeemResetCredit?: () => void;
+  onOpenResetCredits?: () => void;
   canEditCutoff: boolean;
   hasCutoffOverrides: boolean;
   canRedeemResetCredit?: boolean;
   redeemingResetCredit?: boolean;
+  loadingResetCredits?: boolean;
 }
 
-function QuotaDetailRow({ q }: { q: any }) {
+function QuotaDetailRow({
+  q,
+  onOpenResetCredits,
+  loadingResetCredits = false,
+}: {
+  q: any;
+  onOpenResetCredits?: () => void;
+  loadingResetCredits?: boolean;
+}) {
   const t = useTranslations("usage");
   if (q.isResetCredits) {
     const count = Number(q.creditCount ?? q.remaining ?? 0);
@@ -87,12 +96,25 @@ function QuotaDetailRow({ q }: { q: any }) {
             {translateUsageOrFallback(t, "resetCreditsLabel", "Reset credits")}
           </span>
         </span>
-        <span
-          className="inline-flex h-6 shrink-0 items-center text-[12px] font-bold leading-none tabular-nums"
+        <button
+          type="button"
+          disabled={!onOpenResetCredits || loadingResetCredits}
+          onClick={(event) => {
+            event.stopPropagation();
+            onOpenResetCredits?.();
+          }}
+          aria-label={translateUsageOrFallback(t, "viewResetCredits", "View reset credits")}
+          className="inline-flex h-6 shrink-0 items-center gap-1 rounded-md px-1.5 text-[12px] font-bold leading-none tabular-nums hover:bg-black/[0.05] disabled:cursor-default dark:hover:bg-white/[0.05]"
           style={{ color: colors.text }}
         >
+          {loadingResetCredits && (
+            <span className="material-symbols-outlined animate-spin text-[12px]">
+              progress_activity
+            </span>
+          )}
           {count.toLocaleString(undefined, { maximumFractionDigits: 0 })}
-        </span>
+          <span className="material-symbols-outlined text-[13px]">chevron_right</span>
+        </button>
       </div>
     );
   }
@@ -178,11 +200,12 @@ export default function QuotaCardExpanded({
   onRefresh,
   onOpenCutoff,
   onOpenCost,
-  onRedeemResetCredit,
+  onOpenResetCredits,
   canEditCutoff,
   hasCutoffOverrides,
   canRedeemResetCredit = false,
   redeemingResetCredit = false,
+  loadingResetCredits = false,
 }: Props) {
   const t = useTranslations("usage");
   const tr = (key: string, fallback: string, values?: UsageTranslationValues) =>
@@ -231,7 +254,12 @@ export default function QuotaCardExpanded({
       ) : (
         <div className="flex flex-col divide-y divide-border/40">
           {visibleQuotas.map((q, i) => (
-            <QuotaDetailRow key={`${q.name}-${q.modelKey ?? ""}-${i}`} q={q} />
+            <QuotaDetailRow
+              key={`${q.name}-${q.modelKey ?? ""}-${i}`}
+              q={q}
+              onOpenResetCredits={q.isResetCredits ? onOpenResetCredits : undefined}
+              loadingResetCredits={loadingResetCredits}
+            />
           ))}
         </div>
       )}
@@ -273,21 +301,21 @@ export default function QuotaCardExpanded({
           {canRedeemResetCredit && (
             <button
               type="button"
-              disabled={loading || redeemingResetCredit}
+              disabled={loading || redeemingResetCredit || loadingResetCredits}
               onClick={(e) => {
                 e.stopPropagation();
-                onRedeemResetCredit?.();
+                onOpenResetCredits?.();
               }}
               className="inline-flex items-center gap-1 text-[11px] font-medium px-2 py-1 rounded-md border border-primary/40 text-primary bg-bg-subtle hover:bg-black/[0.04] dark:hover:bg-white/[0.04] disabled:opacity-40 disabled:cursor-not-allowed cursor-pointer"
             >
               <span
                 className={`material-symbols-outlined text-[12px] ${
-                  redeemingResetCredit ? "animate-spin" : ""
+                  redeemingResetCredit || loadingResetCredits ? "animate-spin" : ""
                 }`}
               >
-                {redeemingResetCredit ? "progress_activity" : "restart_alt"}
+                {redeemingResetCredit || loadingResetCredits ? "progress_activity" : "restart_alt"}
               </span>
-              {tr("redeemResetCredit", "Redeem reset")}
+              {tr("manageResetCredits", "View credits")}
             </button>
           )}
           <button
diff --git a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/useCodexResetCreditRedemption.ts b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/useCodexResetCreditRedemption.ts
index e5decda962a..da439a00035 100644
--- a/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/useCodexResetCreditRedemption.ts
+++ b/src/app/(dashboard)/dashboard/usage/components/ProviderLimits/useCodexResetCreditRedemption.ts
@@ -1,6 +1,6 @@
 "use client";
 
-import { useCallback, useState } from "react";
+import { useCallback, useRef, useState } from "react";
 
 import { useNotificationStore } from "@/store/notificationStore";
 import { parseQuotaData } from "./utils";
@@ -8,69 +8,198 @@ import type { UsageTranslationValues } from "./i18nFallback";
 
 type TranslateUsage = (key: string, fallback: string, values?: UsageTranslationValues) => string;
 
-export function useCodexResetCreditRedemption(
-  tr: TranslateUsage,
-  setErrors: React.Dispatch<React.SetStateAction<Record<string, string | null>>>,
-  setQuotaData: React.Dispatch<React.SetStateAction<Record<string, any>>>,
-  setLastRefreshedAt: React.Dispatch<React.SetStateAction<Record<string, string>>>
-) {
-  const notify = useNotificationStore();
-  const [redeemingResetCreditId, setRedeemingResetCreditId] = useState<string | null>(null);
+export interface CodexResetCreditView {
+  selectionToken: string;
+  resetType?: string;
+  status?: string;
+  grantedAt?: string;
+  expiresAt?: string | null;
+  title?: string;
+  description?: string;
+}
 
-  const redeemCodexResetCredit = useCallback(
-    async (connectionId: string, provider: string) => {
-      if (provider !== "codex" || redeemingResetCreditId) return;
+interface ResetCreditPickerState {
+  connectionId: string;
+  provider: string;
+  credits: CodexResetCreditView[];
+  availableCount: number;
+}
 
-      const confirmed = window.confirm(
-        tr(
-          "confirmRedeemResetCredit",
-          "Redeem one Codex reset credit for this account? This consumes one reset credit."
-        )
-      );
-      if (!confirmed) return;
+type SetErrors = React.Dispatch<React.SetStateAction<Record<string, string | null>>>;
+type SetQuotaData = React.Dispatch<React.SetStateAction<Record<string, any>>>;
+type SetLastRefreshedAt = React.Dispatch<React.SetStateAction<Record<string, string>>>;
 
-      const idempotencyKey =
-        typeof crypto !== "undefined" && typeof crypto.randomUUID === "function"
-          ? crypto.randomUUID()
-          : `${Date.now()}-${Math.random().toString(16).slice(2)}`;
+interface ResetCreditRequestState {
+  idempotencyKeysRef: React.MutableRefObject<Record<string, string>>;
+  redeemingResetCreditId: string | null;
+  resetCreditPicker: ResetCreditPickerState | null;
+  setErrors: SetErrors;
+  setLastRefreshedAt: SetLastRefreshedAt;
+  setQuotaData: SetQuotaData;
+  setRedeemingResetCreditId: React.Dispatch<React.SetStateAction<string | null>>;
+  setResetCreditPicker: React.Dispatch<React.SetStateAction<ResetCreditPickerState | null>>;
+  tr: TranslateUsage;
+}
 
-      setRedeemingResetCreditId(connectionId);
+function createIdempotencyKey(): string {
+  return typeof crypto !== "undefined" && typeof crypto.randomUUID === "function"
+    ? crypto.randomUUID()
+    : `${Date.now()}-${Math.random().toString(16).slice(2)}`;
+}
+
+function getRequestErrorMessage(error: unknown, fallback: string): string {
+  return error instanceof Error && error.message ? error.message : fallback;
+}
+
+function useOpenCodexResetCredits(
+  loadingResetCreditsId: string | null,
+  redeemingResetCreditId: string | null,
+  setErrors: SetErrors,
+  setLoadingResetCreditsId: React.Dispatch<React.SetStateAction<string | null>>,
+  setResetCreditPicker: React.Dispatch<React.SetStateAction<ResetCreditPickerState | null>>,
+  idempotencyKeysRef: React.MutableRefObject<Record<string, string>>,
+  tr: TranslateUsage
+) {
+  const notify = useNotificationStore();
+  return useCallback(
+    async (connectionId: string, provider: string) => {
+      if (provider !== "codex" || loadingResetCreditsId || redeemingResetCreditId) return;
+      setLoadingResetCreditsId(connectionId);
       setErrors((prev) => ({ ...prev, [connectionId]: null }));
+      try {
+        const response = await fetch(
+          `/api/usage/codex-reset-credit?connectionId=${encodeURIComponent(connectionId)}`,
+          { cache: "no-store" }
+        );
+        const data = await response.json().catch(() => ({}));
+        if (!response.ok) throw new Error(data.error || response.statusText);
+        idempotencyKeysRef.current = {};
+        setResetCreditPicker({
+          connectionId,
+          provider,
+          credits: Array.isArray(data.credits) ? data.credits : [],
+          availableCount: Number.isFinite(Number(data.availableCount))
+            ? Number(data.availableCount)
+            : 0,
+        });
+      } catch (error) {
+        const message = getRequestErrorMessage(
+          error,
+          tr("resetCreditsLoadFailed", "Failed to load reset credits")
+        );
+        setErrors((prev) => ({ ...prev, [connectionId]: message }));
+        notify.error(message);
+      } finally {
+        setLoadingResetCreditsId(null);
+      }
+    },
+    [
+      idempotencyKeysRef,
+      loadingResetCreditsId,
+      notify,
+      redeemingResetCreditId,
+      setErrors,
+      setLoadingResetCreditsId,
+      setResetCreditPicker,
+      tr,
+    ]
+  );
+}
 
+function useRedeemCodexResetCredit(state: ResetCreditRequestState) {
+  const notify = useNotificationStore();
+  return useCallback(
+    async (selectionToken: string) => {
+      const picker = state.resetCreditPicker;
+      if (!picker || state.redeemingResetCreditId || !selectionToken) return;
+      const idempotencyKey =
+        state.idempotencyKeysRef.current[selectionToken] ??
+        (state.idempotencyKeysRef.current[selectionToken] = createIdempotencyKey());
+      state.setRedeemingResetCreditId(picker.connectionId);
+      state.setErrors((prev) => ({ ...prev, [picker.connectionId]: null }));
       try {
         const response = await fetch("/api/usage/codex-reset-credit", {
           method: "POST",
           headers: { "Content-Type": "application/json" },
-          body: JSON.stringify({ connectionId, idempotencyKey }),
+          body: JSON.stringify({
+            connectionId: picker.connectionId,
+            idempotencyKey,
+            creditId: selectionToken,
+          }),
         });
         const data = await response.json().catch(() => ({}));
-
         if (!response.ok) throw new Error(data.error || response.statusText);
-
         const usage = data.usage || {};
-        setQuotaData((prev) => ({
+        state.setQuotaData((prev) => ({
           ...prev,
-          [connectionId]: {
-            quotas: parseQuotaData(provider, usage),
+          [picker.connectionId]: {
+            quotas: parseQuotaData(picker.provider, usage),
             plan: usage.plan || null,
             message: usage.message || null,
             raw: usage,
             stale: usage._stale ? { since: usage._staleSince, reason: usage._staleReason } : null,
           },
         }));
-        setLastRefreshedAt((prev) => ({ ...prev, [connectionId]: new Date().toISOString() }));
-        notify.success(tr("resetCreditRedeemed", "Reset redeemed"));
-      } catch (error: any) {
-        const message =
-          error?.message || tr("resetCreditRedeemFailed", "Failed to redeem reset credit");
-        setErrors((prev) => ({ ...prev, [connectionId]: message }));
+        state.setLastRefreshedAt((prev) => ({
+          ...prev,
+          [picker.connectionId]: new Date().toISOString(),
+        }));
+        state.setResetCreditPicker(null);
+        state.idempotencyKeysRef.current = {};
+        notify.success(state.tr("resetCreditRedeemed", "Reset redeemed"));
+      } catch (error) {
+        const message = getRequestErrorMessage(
+          error,
+          state.tr("resetCreditRedeemFailed", "Failed to redeem reset credit")
+        );
+        state.setErrors((prev) => ({ ...prev, [picker.connectionId]: message }));
         notify.error(message);
       } finally {
-        setRedeemingResetCreditId(null);
+        state.setRedeemingResetCreditId(null);
       }
     },
-    [notify, redeemingResetCreditId, setErrors, setLastRefreshedAt, setQuotaData, tr]
+    [notify, state]
+  );
+}
+
+export function useCodexResetCreditRedemption(
+  tr: TranslateUsage,
+  setErrors: React.Dispatch<React.SetStateAction<Record<string, string | null>>>,
+  setQuotaData: React.Dispatch<React.SetStateAction<Record<string, any>>>,
+  setLastRefreshedAt: React.Dispatch<React.SetStateAction<Record<string, string>>>
+) {
+  const [redeemingResetCreditId, setRedeemingResetCreditId] = useState<string | null>(null);
+  const [loadingResetCreditsId, setLoadingResetCreditsId] = useState<string | null>(null);
+  const [resetCreditPicker, setResetCreditPicker] = useState<ResetCreditPickerState | null>(null);
+  const idempotencyKeysRef = useRef<Record<string, string>>({});
+
+  const openCodexResetCredits = useOpenCodexResetCredits(
+    loadingResetCreditsId,
+    redeemingResetCreditId,
+    setErrors,
+    setLoadingResetCreditsId,
+    setResetCreditPicker,
+    idempotencyKeysRef,
+    tr
   );
+  const redeemCodexResetCredit = useRedeemCodexResetCredit({
+    idempotencyKeysRef,
+    redeemingResetCreditId,
+    resetCreditPicker,
+    setErrors,
+    setLastRefreshedAt,
+    setQuotaData,
+    setRedeemingResetCreditId,
+    setResetCreditPicker,
+    tr,
+  });
 
-  return { redeemCodexResetCredit, redeemingResetCreditId };
+  return {
+    closeResetCreditPicker: () => setResetCreditPicker(null),
+    loadingResetCreditsId,
+    openCodexResetCredits,
+    redeemCodexResetCredit,
+    redeemingResetCreditId,
+    resetCreditPicker,
+  };
 }
diff --git a/src/app/(dashboard)/home/ProviderQuotaWidget.tsx b/src/app/(dashboard)/home/ProviderQuotaWidget.tsx
index 79c4204bd2e..0144ac458eb 100644
--- a/src/app/(dashboard)/home/ProviderQuotaWidget.tsx
+++ b/src/app/(dashboard)/home/ProviderQuotaWidget.tsx
@@ -6,6 +6,7 @@ import Card from "@/shared/components/Card";
 import ProviderIcon from "@/shared/components/ProviderIcon";
 import { USAGE_SUPPORTED_PROVIDERS } from "@/shared/constants/providers";
 import { translateUsageOrFallback } from "../dashboard/usage/components/ProviderLimits/i18nFallback";
+import { isProviderQuotaVisible } from "@/shared/utils/providerQuotaVisibility";
 
 type Connection = {
   id: string;
@@ -13,6 +14,7 @@ type Connection = {
   authType?: string;
   email?: string;
   name?: string;
+  quotaVisible?: boolean;
 };
 
 type QuotaData = Record<string, any>;
@@ -121,6 +123,7 @@ export default function ProviderQuotaWidget({ autoRefreshInterval = 0 }: Provide
     // Only keep connections that are usage/quota supported
     const relevant = conns.filter(
       (c) =>
+        isProviderQuotaVisible(c) &&
         USAGE_SUPPORTED_PROVIDERS.includes(c.provider) &&
         (c.authType === "oauth" || c.authType === "apikey")
     );
diff --git a/src/app/api/cli-tools/all-statuses/route.ts b/src/app/api/cli-tools/all-statuses/route.ts
index 02037eee310..06532820f9c 100644
--- a/src/app/api/cli-tools/all-statuses/route.ts
+++ b/src/app/api/cli-tools/all-statuses/route.ts
@@ -92,6 +92,7 @@ export async function GET(request: Request): Promise<Response> {
   if (authError) return authError;
 
   try {
+    const forceRefresh = new URL(request.url).searchParams.get("refresh") === "true";
     const toolIds = Object.keys(CLI_TOOLS);
     const statuses: ToolBatchStatusMap = {};
 
@@ -117,7 +118,7 @@ export async function GET(request: Request): Promise<Response> {
     await Promise.allSettled(
       toolIds.map(async (toolId) => {
         const mtimeMs = mtimesMap[toolId] ?? 0;
-        const cached = getCached(toolId, mtimeMs);
+        const cached = forceRefresh ? null : getCached(toolId, mtimeMs);
 
         if (cached) {
           statuses[toolId] = cached;
@@ -153,14 +154,13 @@ export async function GET(request: Request): Promise<Response> {
 
           // Try to extract endpoint from config file
           const configPath = getCliPrimaryConfigPath(toolId);
-          const endpoint = configPath
-            ? await extractEndpointFromConfig(toolId, configPath)
-            : null;
+          const endpoint = configPath ? await extractEndpointFromConfig(toolId, configPath) : null;
 
           const result: ToolBatchStatus = {
             detection: {
               installed: runtime.installed,
               runnable: runtime.runnable,
+              version: (runtime as Record<string, unknown>).version as string | undefined,
               command: runtime.command ?? undefined,
               commandPath: (runtime as Record<string, unknown>).commandPath as string | undefined,
               reason: runtime.reason ?? undefined,
@@ -203,8 +203,11 @@ export async function GET(request: Request): Promise<Response> {
     return NextResponse.json(statuses);
   } catch (err) {
     logger.error({ err }, "Unexpected error in /api/cli-tools/all-statuses");
-    return NextResponse.json(buildErrorBody(500, err instanceof Error ? err.message : String(err)), {
-      status: 500,
-    });
+    return NextResponse.json(
+      buildErrorBody(500, err instanceof Error ? err.message : String(err)),
+      {
+        status: 500,
+      }
+    );
   }
 }
diff --git a/src/app/api/context/rtk/import/route.ts b/src/app/api/context/rtk/import/route.ts
new file mode 100644
index 00000000000..cfe92d93610
--- /dev/null
+++ b/src/app/api/context/rtk/import/route.ts
@@ -0,0 +1,81 @@
+import { NextResponse } from "next/server";
+import { z } from "zod";
+import { requireManagementAuth } from "@/lib/api/requireManagementAuth";
+import { buildErrorBody } from "@omniroute/open-sse/utils/error";
+import { loadRtkFilters } from "@omniroute/open-sse/services/compression/engines/rtk/filterLoader";
+import {
+  installGlobalRtkTomlV1,
+  parseRtkTomlV1,
+  RTK_TOML_MAX_BYTES,
+  RtkTomlCompatibilityError,
+  type RtkTomlCompatibilityResult,
+} from "@omniroute/open-sse/services/compression/engines/rtk/tomlCompatibility";
+
+const RequestSchema = z
+  .object({
+    action: z.enum(["validate", "install"]),
+    content: z.string().min(1).max(RTK_TOML_MAX_BYTES),
+    overwrite: z.boolean().optional(),
+  })
+  .strict();
+
+function responseBody(
+  result: RtkTomlCompatibilityResult & {
+    installedPath?: string;
+    backupCreated?: boolean;
+  }
+) {
+  return {
+    schemaVersion: result.schemaVersion,
+    sha256: result.sha256,
+    passed: result.passed,
+    filters: result.filters.map((filter) => ({
+      id: filter.id,
+      description: filter.description,
+      category: filter.category,
+      commandPatterns: filter.commandPatterns,
+      testCount: filter.tests.length,
+    })),
+    outcomes: result.outcomes,
+    filtersWithoutTests: result.filtersWithoutTests,
+    warnings: result.warnings,
+    installedPath: result.installedPath,
+    backupCreated: result.backupCreated,
+  };
+}
+
+export async function POST(request: Request) {
+  const authError = await requireManagementAuth(request);
+  if (authError) return authError;
+
+  let body: unknown;
+  try {
+    body = await request.json();
+  } catch {
+    return NextResponse.json(buildErrorBody(400, "Invalid JSON body"), { status: 400 });
+  }
+  const parsed = RequestSchema.safeParse(body);
+  if (!parsed.success) {
+    return NextResponse.json(buildErrorBody(400, "Invalid RTK TOML import request"), {
+      status: 400,
+    });
+  }
+
+  try {
+    const result =
+      parsed.data.action === "install"
+        ? installGlobalRtkTomlV1(parsed.data.content, { overwrite: parsed.data.overwrite })
+        : parseRtkTomlV1(parsed.data.content);
+    if (parsed.data.action === "install") {
+      loadRtkFilters({ refresh: true });
+    }
+    return NextResponse.json(responseBody(result));
+  } catch (error) {
+    if (error instanceof RtkTomlCompatibilityError) {
+      return NextResponse.json(buildErrorBody(400, error.publicMessage), { status: 400 });
+    }
+    return NextResponse.json(buildErrorBody(500, "Failed to process RTK TOML filter import"), {
+      status: 500,
+    });
+  }
+}
diff --git a/src/app/api/issue-agent/runs/route.ts b/src/app/api/issue-agent/runs/route.ts
new file mode 100644
index 00000000000..dd5b8c33b51
--- /dev/null
+++ b/src/app/api/issue-agent/runs/route.ts
@@ -0,0 +1,141 @@
+import { NextResponse } from "next/server";
+import { z } from "zod";
+import { appendIssueAgentAuditRecord } from "@/lib/issueAgent/audit";
+import {
+  executeRecordedTriageChatCompletion,
+  RecordedTriageTimeoutError,
+} from "@/lib/issueAgent/execution";
+import { normalizeGitHubIssueExport } from "@/lib/issueAgent/githubExport";
+import { createRecordedTriageRun } from "@/lib/issueAgent/recordedTriage";
+import { POST as postChatCompletion } from "@/app/api/v1/chat/completions/route";
+import { isValidationFailure, validateBody } from "@/shared/validation/helpers";
+import { sanitizeErrorMessage } from "@omniroute/open-sse/utils/error";
+
+const issueAgentRunRequestSchema = z.object({
+  mode: z.string().optional(),
+  issueUrl: z.string().optional(),
+  dryRun: z.boolean().optional(),
+  model: z.string().min(1).max(256).optional(),
+  provider: z.string().min(1).max(128).optional(),
+  routingPolicy: z.string().min(1).max(128).optional(),
+  timeoutMs: z.number().int().min(1).max(120_000).optional(),
+  recordedContext: z.unknown().optional(),
+  githubExport: z.unknown().optional(),
+});
+
+const ENABLED_VALUES = new Set(["1", "true", "yes", "on"]);
+
+function isIssueAgentEnabled(): boolean {
+  return ENABLED_VALUES.has((process.env.OMNIROUTE_ISSUE_AGENT_ENABLED ?? "").toLowerCase());
+}
+
+/**
+ * Node's own fs/system errors (ENOENT, EACCES, EEXIST, ...) always set `.code`
+ * to an uppercase errno identifier and embed the absolute path in `.message`
+ * (e.g. `appendIssueAgentAuditRecord`'s mkdir/appendFile on DATA_DIR). Hand-thrown
+ * validation errors in this module (createRecordedTriageRun,
+ * normalizeGitHubIssueExport) are plain `new Error("...")` with a curated,
+ * path-free message and never set `.code` — so this check safely distinguishes
+ * "safe to show the client" validation failures from opaque internal/system
+ * errors that could otherwise leak the server's filesystem layout.
+ */
+function isNodeSystemError(error: unknown): error is NodeJS.ErrnoException {
+  return (
+    error instanceof Error &&
+    typeof (error as NodeJS.ErrnoException).code === "string" &&
+    /^[A-Z]/.test((error as NodeJS.ErrnoException).code as string)
+  );
+}
+
+export async function GET() {
+  return NextResponse.json({
+    ok: true,
+    enabled: isIssueAgentEnabled(),
+    supportedModes: ["recorded-triage"],
+    execution: "disabled-by-default",
+  });
+}
+
+export async function POST(request: Request) {
+  let body: unknown;
+  try {
+    body = await request.json();
+  } catch {
+    return NextResponse.json({ error: "Invalid JSON body" }, { status: 400 });
+  }
+
+  const validation = validateBody(issueAgentRunRequestSchema, body);
+  if (isValidationFailure(validation)) {
+    return NextResponse.json({ error: validation.error }, { status: 400 });
+  }
+
+  const parsed = validation.data;
+  if (parsed.mode !== "recorded-triage") {
+    return NextResponse.json(
+      { error: "Unsupported issue-agent mode", supportedModes: ["recorded-triage"] },
+      { status: 400 }
+    );
+  }
+
+  if (!isIssueAgentEnabled()) {
+    return NextResponse.json(
+      {
+        error: "Issue Agent execution is disabled",
+        enabled: false,
+        requiredEnv: "OMNIROUTE_ISSUE_AGENT_ENABLED=true",
+      },
+      { status: 403 }
+    );
+  }
+
+  try {
+    const normalized = parsed.githubExport ? normalizeGitHubIssueExport(parsed.githubExport) : null;
+    const run = createRecordedTriageRun({
+      ...parsed,
+      issueUrl: parsed.issueUrl ?? normalized?.issueUrl,
+      recordedContext: parsed.recordedContext ?? normalized?.recordedContext,
+    });
+    const audit = await appendIssueAgentAuditRecord(run);
+    if (run.dryRun) {
+      return NextResponse.json({ ...run, auditPath: audit.path });
+    }
+
+    const completion = await executeRecordedTriageChatCompletion(
+      {
+        run,
+        model: parsed.model,
+        provider: parsed.provider,
+        routingPolicy: parsed.routingPolicy,
+        timeoutMs: parsed.timeoutMs,
+      },
+      postChatCompletion
+    );
+    return NextResponse.json(
+      {
+        ...run,
+        runner: "omniroute-chat-completions",
+        auditPath: audit.path,
+        completion: completion.body,
+      },
+      { status: completion.status }
+    );
+  } catch (error) {
+    if (error instanceof RecordedTriageTimeoutError) {
+      return NextResponse.json(
+        { error: sanitizeErrorMessage(error.message), code: "ISSUE_AGENT_TIMEOUT" },
+        { status: 504 }
+      );
+    }
+    // Hard Rule #12: never put a raw err.message in a response. Validation
+    // failures thrown by this module (bad issue URL, malformed GitHub export)
+    // are safe, curated messages meant for the client — sanitizeErrorMessage()
+    // is a no-op for them. An opaque Node system/fs error (e.g. audit.ts
+    // failing to mkdir/appendFile under DATA_DIR) is replaced with a generic
+    // message instead, since its raw text would otherwise embed the server's
+    // absolute filesystem path.
+    const message = isNodeSystemError(error)
+      ? "Issue Agent request failed due to an internal error"
+      : sanitizeErrorMessage(error instanceof Error ? error.message : "Invalid issue-agent request");
+    return NextResponse.json({ error: message }, { status: 400 });
+  }
+}
diff --git a/src/app/api/network/info/route.ts b/src/app/api/network/info/route.ts
index a8dc33cca72..0cbde43c0dd 100644
--- a/src/app/api/network/info/route.ts
+++ b/src/app/api/network/info/route.ts
@@ -12,6 +12,7 @@ export async function GET(request: Request) {
 
   const { apiPort } = getRuntimePorts();
   const interfaces = os.networkInterfaces();
+  const localUrl = `http://localhost:${apiPort}/v1`;
   const lanUrls: string[] = [];
   let tailscaleIpUrl: string | null = null;
 
@@ -28,5 +29,5 @@ export async function GET(request: Request) {
     }
   }
 
-  return NextResponse.json({ lanUrls, tailscaleIpUrl });
+  return NextResponse.json({ localUrl, lanUrls, tailscaleIpUrl });
 }
diff --git a/src/app/api/oauth/[provider]/[action]/route.ts b/src/app/api/oauth/[provider]/[action]/route.ts
index 0774bd8d2c3..c89336e4871 100755
--- a/src/app/api/oauth/[provider]/[action]/route.ts
+++ b/src/app/api/oauth/[provider]/[action]/route.ts
@@ -260,7 +260,12 @@ export async function GET(
     return NextResponse.json({ error: "Unknown action" }, { status: 400 });
   } catch (error) {
     console.error("OAuth GET error:", error);
-    return NextResponse.json({ error: "Internal server error" }, { status: 500 });
+    // Surface the SANITIZED upstream reason instead of a generic 500 that hides WHY the flow failed.
+    // device-code providers (qwen → qwen.ai, codebuddy-cn → copilot.tencent.com) throw a descriptive
+    // message ("Device code request failed: …", "CodeBuddy state request failed (403)") that was being
+    // swallowed, so a geo-block / upstream outage looked identical to a real server bug in the UI.
+    const detail = sanitizeErrorMessage(error instanceof Error ? error.message : String(error));
+    return NextResponse.json({ error: detail || "Internal server error" }, { status: 500 });
   }
 }
 
diff --git a/src/app/api/providers/[id]/models/discovery/providerModelsConfig.ts b/src/app/api/providers/[id]/models/discovery/providerModelsConfig.ts
index ab0b1de1cbb..ce1df845ef8 100644
--- a/src/app/api/providers/[id]/models/discovery/providerModelsConfig.ts
+++ b/src/app/api/providers/[id]/models/discovery/providerModelsConfig.ts
@@ -4,6 +4,7 @@ import { parseGeminiModelsList } from "@/lib/providerModels/geminiModelsParser";
 import { filterClinepassModels } from "@omniroute/open-sse/services/clinepassModels.ts";
 import { normalizeOpenAiLikeModelsResponse } from "./normalizers";
 import { extractKimiJwt } from "@/lib/providers/webCookieAuth";
+import { buildClaudeCodeCompatibleHeaders } from "@omniroute/open-sse/services/claudeCodeCompatible.ts";
 
 export type ProviderModelsConfigEntry = {
   url: string;
@@ -133,6 +134,29 @@ export const PROVIDER_MODELS_CONFIG: Record<string, ProviderModelsConfigEntry> =
     body: {},
     parseResponse: (data) => data.models || [],
   },
+  // #7016: AgentRouter rejects /v1/models unless the request carries the same
+  // Claude Code wire image the chat path uses (it adopts the dynamic CC wire
+  // image while keeping its own x-api-key auth — see #6056). Without these
+  // headers the gateway WAF 4xx's the request and model import silently falls
+  // back to the local catalog ("API unavailable — using local catalog").
+  agentrouter: {
+    url: "https://agentrouter.org/v1/models",
+    method: "GET",
+    headers: { "Content-Type": "application/json" },
+    buildHeaders: (token: string) => {
+      const wire = buildClaudeCodeCompatibleHeaders(token, false, undefined, {});
+      const out: Record<string, string> = { ...wire };
+      // Keep AgentRouter's own x-api-key auth scheme (#6056); the CC helper
+      // adds a Bearer Authorization we must not send.
+      for (const key of Object.keys(out)) {
+        if (key.toLowerCase() === "authorization") delete out[key];
+      }
+      if (token) out["x-api-key"] = token;
+      return out;
+    },
+    parseResponse: (data: any) =>
+      Array.isArray(data) ? data : (data?.data || data?.models || []),
+  },
   openai: {
     url: "https://api.openai.com/v1/models",
     method: "GET",
diff --git a/src/app/api/providers/[id]/route.ts b/src/app/api/providers/[id]/route.ts
index c92a7a2c675..a60b4a5c3a7 100644
--- a/src/app/api/providers/[id]/route.ts
+++ b/src/app/api/providers/[id]/route.ts
@@ -24,7 +24,10 @@ import {
 } from "@/lib/providers/claudeExtraUsage";
 import { requireManagementAuth } from "@/lib/api/requireManagementAuth";
 import { isApiKeyRevealEnabled, maskStoredApiKey } from "@/lib/apiKeyExposure";
-import { refreshConnectionRateLimits, enableRateLimitProtection } from "@/../open-sse/services/rateLimitManager";
+import {
+  refreshConnectionRateLimits,
+  enableRateLimitProtection,
+} from "@/../open-sse/services/rateLimitManager";
 
 function normalizeCodexLimitPolicy(
   incoming: unknown,
@@ -135,6 +138,7 @@ export async function PUT(request: Request, { params }: { params: Promise<{ id:
       quotaWindowThresholds: incomingWindowThresholds,
       proxyEnabled,
       perKeyProxyEnabled,
+      quotaVisible,
       projectId,
       providerSpecificData: incomingPsd,
       rateLimitOverrides,
@@ -191,6 +195,7 @@ export async function PUT(request: Request, { params }: { params: Promise<{ id:
     if (rateLimitOverrides !== undefined) updateData.rateLimitOverrides = rateLimitOverrides;
     if (proxyEnabled !== undefined) updateData.proxyEnabled = proxyEnabled;
     if (perKeyProxyEnabled !== undefined) updateData.perKeyProxyEnabled = perKeyProxyEnabled;
+    if (quotaVisible !== undefined) updateData.quotaVisible = quotaVisible;
 
     // Merge providerSpecificData (partial update — preserve existing keys not sent by caller)
     if (incomingPsd !== undefined && incomingPsd !== null && typeof incomingPsd === "object") {
diff --git a/src/app/api/tools/agent-bridge/agents/[id]/dns/route.ts b/src/app/api/tools/agent-bridge/agents/[id]/dns/route.ts
index 368f8bcdeda..fa7f5f6abc7 100644
--- a/src/app/api/tools/agent-bridge/agents/[id]/dns/route.ts
+++ b/src/app/api/tools/agent-bridge/agents/[id]/dns/route.ts
@@ -13,10 +13,10 @@ import { sanitizeErrorMessage } from "@omniroute/open-sse/utils/error";
 import { createErrorResponse } from "@/lib/api/errorResponse";
 import { ALL_TARGETS } from "@/mitm/targets/index";
 
-type Params = { params: { id: string } };
+type Params = { params: Promise<{ id: string }> };
 
 export async function POST(request: Request, { params }: Params): Promise<Response> {
-  const { id } = params;
+  const { id } = await params;
 
   let body: unknown;
   try {
diff --git a/src/app/api/usage/codex-reset-credit/route.ts b/src/app/api/usage/codex-reset-credit/route.ts
index ecbae077823..c9840bb8aa0 100644
--- a/src/app/api/usage/codex-reset-credit/route.ts
+++ b/src/app/api/usage/codex-reset-credit/route.ts
@@ -1,13 +1,57 @@
 import { NextResponse } from "next/server";
 import { z } from "zod";
-import { CodexResetCreditError, consumeCodexResetCredit } from "@/lib/usage/codexResetCredits";
+import { requireManagementAuth } from "@/lib/api/requireManagementAuth";
+import { sanitizeErrorMessage } from "@omniroute/open-sse/utils/error.ts";
+import {
+  CodexResetCreditError,
+  consumeCodexResetCredit,
+  listCodexResetCredits,
+} from "@/lib/usage/codexResetCredits";
+
+const ConnectionIdSchema = z.string().trim().min(1).max(256);
 
 const CodexResetCreditBodySchema = z.object({
-  connectionId: z.string().optional(),
-  idempotencyKey: z.string().optional(),
+  connectionId: ConnectionIdSchema,
+  idempotencyKey: z.string().trim().min(1).max(256),
+  creditId: z.string().trim().min(1).max(512).optional(),
 });
 
+function buildErrorResponse(error: unknown) {
+  const status = error instanceof CodexResetCreditError ? error.status : 500;
+  const code = error instanceof CodexResetCreditError ? error.code : "codex_reset_credit_failed";
+  const message =
+    error instanceof CodexResetCreditError
+      ? sanitizeErrorMessage(error.message) || "Codex reset-credit request failed."
+      : "Codex reset-credit request failed.";
+  console.error("[API] /api/usage/codex-reset-credit error:", error);
+  return NextResponse.json({ ok: false, code, error: message }, { status });
+}
+
+export async function GET(request: Request) {
+  const authError = await requireManagementAuth(request);
+  if (authError) return authError;
+
+  try {
+    const parsed = ConnectionIdSchema.safeParse(
+      new URL(request.url).searchParams.get("connectionId")
+    );
+    if (!parsed.success) {
+      return NextResponse.json(
+        { ok: false, code: "invalid_connection_id", error: "Invalid connectionId." },
+        { status: 400 }
+      );
+    }
+    const result = await listCodexResetCredits(parsed.data);
+    return NextResponse.json({ ok: true, ...result });
+  } catch (error) {
+    return buildErrorResponse(error);
+  }
+}
+
 export async function POST(request: Request) {
+  const authError = await requireManagementAuth(request);
+  if (authError) return authError;
+
   try {
     const raw = await request.json().catch(() => ({}));
     const parsed = CodexResetCreditBodySchema.safeParse(raw);
@@ -18,16 +62,13 @@ export async function POST(request: Request) {
       );
     }
 
-    const connectionId = parsed.data.connectionId ?? "";
-    const idempotencyKey = parsed.data.idempotencyKey ?? "";
-
-    const result = await consumeCodexResetCredit(connectionId, idempotencyKey);
+    const result = await consumeCodexResetCredit(
+      parsed.data.connectionId,
+      parsed.data.idempotencyKey,
+      parsed.data.creditId
+    );
     return NextResponse.json({ ok: true, ...result });
   } catch (error) {
-    const status = error instanceof CodexResetCreditError ? error.status : 500;
-    const code = error instanceof CodexResetCreditError ? error.code : "codex_reset_credit_failed";
-    const message = error instanceof Error ? error.message : "Failed to redeem Codex reset credit.";
-    console.error("[API] POST /api/usage/codex-reset-credit error:", error);
-    return NextResponse.json({ ok: false, code, error: message }, { status });
+    return buildErrorResponse(error);
   }
 }
diff --git a/src/app/api/v1/images/generations/route.ts b/src/app/api/v1/images/generations/route.ts
index 162bacc1a51..857aeb3f047 100644
--- a/src/app/api/v1/images/generations/route.ts
+++ b/src/app/api/v1/images/generations/route.ts
@@ -10,6 +10,7 @@ import {
   parseImageModel,
   getImageProvider,
   getImageModelEntry,
+  modalitiesRequireImageInput,
 } from "@omniroute/open-sse/config/imageRegistry.ts";
 import { errorResponse, unavailableResponse } from "@omniroute/open-sse/utils/error.ts";
 import { HTTP_STATUS } from "@omniroute/open-sse/config/constants.ts";
@@ -151,7 +152,13 @@ async function postHandler(request, context) {
   const imageModelEntry = getImageModelEntry(body.model);
   const inputModalities = imageModelEntry?.inputModalities || ["text"];
   const requiresPrompt = inputModalities.includes("text");
-  const requiresImageInput = inputModalities.includes("image");
+  // imageRequired is an explicit registry override for models that list "text" among
+  // their modalities (they accept a prompt) but mechanically require an input image
+  // regardless — e.g. Stability AI's dedicated edit/control/upscale endpoints. Without
+  // it, modalitiesRequireImageInput() would infer "image optional" for any model that
+  // also lists "text", which is wrong for those.
+  const requiresImageInput =
+    Boolean(imageModelEntry?.imageRequired) || modalitiesRequireImageInput(inputModalities);
   const hasPrompt = typeof body.prompt === "string" && body.prompt.trim().length > 0;
   const hasImageInput = hasImageGenerationInput(body);
 
diff --git a/src/app/api/v1/models/catalog.ts b/src/app/api/v1/models/catalog.ts
index f4419da2825..41b7a994515 100644
--- a/src/app/api/v1/models/catalog.ts
+++ b/src/app/api/v1/models/catalog.ts
@@ -69,6 +69,8 @@ import {
   intersectStringArrays,
   minKnownNumber,
   maybeOmitCatalogModelName,
+  getThinkingCapabilityFields,
+  mergeComboCapabilities,
 } from "./catalogHelpers";
 import {
   qualifyOpenRouterModelId,
@@ -529,32 +531,25 @@ async function buildUnifiedModelsResponseCore(
             ? ["text"]
             : undefined;
 
-      const capabilities: Record<string, boolean> = {};
-      if (typeof synced?.tool_call === "boolean") {
-        capabilities.tool_calling = synced.tool_call;
-      } else if (typeof registryModel?.toolCalling === "boolean") {
-        capabilities.tool_calling = registryModel.toolCalling;
-      } else if (typeof spec?.supportsTools === "boolean") {
-        capabilities.tool_calling = spec.supportsTools;
+      const capabilities: Record<string, boolean | string[]> = {};
+      capabilities.tool_calling = canonical.capabilities.toolCalling;
+      capabilities.reasoning = canonical.capabilities.reasoning;
+      if (typeof canonical.capabilities.vision === "boolean") {
+        capabilities.vision = canonical.capabilities.vision;
       }
-      if (typeof synced?.reasoning === "boolean") {
-        capabilities.reasoning = synced.reasoning;
-      } else if (typeof registryModel?.supportsReasoning === "boolean") {
-        capabilities.reasoning = registryModel.supportsReasoning;
-      } else if (typeof spec?.supportsThinking === "boolean") {
-        capabilities.reasoning = spec.supportsThinking;
+      if (typeof canonical.capabilities.attachment === "boolean") {
+        capabilities.attachment = canonical.capabilities.attachment;
       }
-      if (typeof knownVision === "boolean") capabilities.vision = knownVision;
-      if (typeof synced?.attachment === "boolean") capabilities.attachment = synced.attachment;
-      if (typeof synced?.structured_output === "boolean") {
-        capabilities.structured_output = synced.structured_output;
+      if (typeof canonical.capabilities.structuredOutput === "boolean") {
+        capabilities.structured_output = canonical.capabilities.structuredOutput;
       }
-      if (typeof synced?.temperature === "boolean") capabilities.temperature = synced.temperature;
-      if (typeof synced?.reasoning === "boolean") {
-        capabilities.thinking = synced.reasoning;
-      } else if (typeof spec?.supportsThinking === "boolean") {
-        capabilities.thinking = spec.supportsThinking;
+      if (typeof canonical.capabilities.temperature === "boolean") {
+        capabilities.temperature = canonical.capabilities.temperature;
       }
+      Object.assign(
+        capabilities,
+        getThinkingCapabilityFields(providerId, modelId, canonical.capabilities.supportsThinking)
+      );
 
       return {
         ...(contextLength ? { contextLength } : {}),
@@ -606,22 +601,7 @@ async function buildUnifiedModelsResponseCore(
         ? intersectStringArrays(knownMetadata.map((metadata) => metadata.outputModalities || []))
         : [];
 
-      const capabilities: Record<string, boolean> = {};
-      for (const key of [
-        "tool_calling",
-        "reasoning",
-        "vision",
-        "attachment",
-        "structured_output",
-        "temperature",
-        "thinking",
-      ]) {
-        const values = knownMetadata.map((metadata) => metadata.capabilities[key]);
-        if (values.every((value): value is boolean => typeof value === "boolean")) {
-          const [first] = values;
-          if (values.every((value) => value === first)) capabilities[key] = first;
-        }
-      }
+      const capabilities = mergeComboCapabilities(knownMetadata);
 
       return {
         ...baseMetadata,
diff --git a/src/app/api/v1/models/catalogHelpers.ts b/src/app/api/v1/models/catalogHelpers.ts
index cde7d2ad0d4..1dcd106c368 100644
--- a/src/app/api/v1/models/catalogHelpers.ts
+++ b/src/app/api/v1/models/catalogHelpers.ts
@@ -3,6 +3,11 @@
 // catalog host shrinks toward the file-size cap without changing behavior — every
 // function body here is byte-identical to its previous in-catalog definition.
 
+import {
+  CANONICAL_EFFORT_VALUES,
+  extendCodexGpt56EffortValues,
+} from "@/shared/reasoning/effortStandardization";
+
 export interface CustomModelEntry {
   id?: string;
   name?: string;
@@ -30,7 +35,7 @@ export type ComboTargetCatalogMetadata = {
   maxOutputTokens?: number;
   inputModalities?: string[];
   outputModalities?: string[];
-  capabilities: Record<string, boolean>;
+  capabilities: Record<string, boolean | string[]>;
 };
 
 export function isPositiveFiniteNumber(value: unknown): value is number {
@@ -74,3 +79,53 @@ export function minKnownNumber(values: Array<number | undefined>): number | unde
   if (knownValues.length === 0) return undefined;
   return Math.min(...knownValues);
 }
+
+export function getThinkingCapabilityFields(
+  providerId: string,
+  modelId: string,
+  resolvedThinking?: boolean | null
+): Record<string, boolean | string[]> {
+  const supportsThinking = resolvedThinking;
+  if (typeof supportsThinking !== "boolean") return {};
+  return {
+    thinking: supportsThinking,
+    supportsThinking,
+    ...(supportsThinking
+      ? {
+          effort_tiers: extendCodexGpt56EffortValues(providerId, modelId, CANONICAL_EFFORT_VALUES),
+        }
+      : {}),
+  };
+}
+
+export function mergeComboCapabilities(
+  metadata: ComboTargetCatalogMetadata[]
+): Record<string, boolean | string[]> {
+  const capabilities: Record<string, boolean | string[]> = {};
+  for (const key of [
+    "tool_calling",
+    "reasoning",
+    "vision",
+    "attachment",
+    "structured_output",
+    "temperature",
+    "thinking",
+    "supportsThinking",
+  ]) {
+    const values = metadata.map((entry) => entry.capabilities[key]);
+    if (values.every((value): value is boolean => typeof value === "boolean")) {
+      const [first] = values;
+      if (values.every((value) => value === first)) capabilities[key] = first;
+    }
+  }
+  const effortTiers = metadata.map((entry) => entry.capabilities.effort_tiers);
+  if (
+    effortTiers.every(
+      (value): value is string[] =>
+        Array.isArray(value) && value.every((entry) => typeof entry === "string")
+    )
+  ) {
+    capabilities.effort_tiers = intersectStringArrays(effortTiers);
+  }
+  return capabilities;
+}
diff --git a/src/app/api/v1/provider-plugin-manifest/route.ts b/src/app/api/v1/provider-plugin-manifest/route.ts
index 9774b49f536..593693c9df3 100644
--- a/src/app/api/v1/provider-plugin-manifest/route.ts
+++ b/src/app/api/v1/provider-plugin-manifest/route.ts
@@ -1,12 +1,16 @@
+import { createHash } from "node:crypto";
+
 import { CORS_HEADERS } from "@/shared/utils/cors";
 import { generateProviderPluginManifest } from "@omniroute/open-sse/config/providerPluginManifestRegistry.ts";
 
-const JSON_HEADERS = {
+const CACHE_HEADERS = {
   ...CORS_HEADERS,
-  "Content-Type": "application/json",
   "Cache-Control": "public, max-age=60",
 } as const;
 
+// The provider registry is module-static; process reloads rebuild this snapshot.
+let cachedPayload: { body: string; etag: string } | null = null;
+
 export async function OPTIONS() {
   return new Response(null, {
     headers: {
@@ -17,8 +21,36 @@ export async function OPTIONS() {
   });
 }
 
-export async function GET() {
-  return new Response(JSON.stringify(generateProviderPluginManifest()), {
-    headers: JSON_HEADERS,
+function createEtag(body: string): string {
+  return `"${createHash("sha256").update(body).digest("base64url")}"`;
+}
+
+function matchesEtag(ifNoneMatch: string | null, etag: string): boolean {
+  return Boolean(
+    ifNoneMatch
+      ?.split(",")
+      .map((value) => value.trim())
+      .some((value) => value === "*" || value === etag || value === `W/${etag}`)
+  );
+}
+
+function getManifestPayload(): { body: string; etag: string } {
+  if (cachedPayload) return cachedPayload;
+
+  const body = JSON.stringify(generateProviderPluginManifest());
+  cachedPayload = { body, etag: createEtag(body) };
+  return cachedPayload;
+}
+
+export async function GET(request: Request) {
+  const { body, etag } = getManifestPayload();
+  const headers = { ...CACHE_HEADERS, ETag: etag };
+
+  if (matchesEtag(request.headers.get("If-None-Match"), etag)) {
+    return new Response(null, { status: 304, headers });
+  }
+
+  return new Response(body, {
+    headers: { ...headers, "Content-Type": "application/json" },
   });
 }
diff --git a/src/app/globals.css b/src/app/globals.css
index 4df43dff070..566fafb51d3 100644
--- a/src/app/globals.css
+++ b/src/app/globals.css
@@ -482,19 +482,15 @@ select option {
   color: var(--color-text-main);
 }
 
-/* ReactFlow Controls (zoom +/- buttons) dark mode theming */
-.react-flow__controls-button {
-  background-color: var(--color-surface);
-  border-bottom-color: var(--color-border);
-  fill: var(--color-text-main);
-}
-
-.react-flow__controls-button:hover {
-  background-color: var(--color-bg-subtle);
-}
-
-.react-flow__controls-button svg {
-  fill: var(--color-text-main);
+/* React Flow loads its component stylesheet after this global file in some
+   bundles, so direct background/fill declarations can lose to its shorthand
+   rules. Override the supported theme variables on our canvas instead. */
+.react-flow.omniroute-flow {
+  --xy-controls-button-background-color: var(--color-surface);
+  --xy-controls-button-background-color-hover: var(--color-bg-subtle);
+  --xy-controls-button-color: var(--color-text-main);
+  --xy-controls-button-color-hover: var(--color-text-main);
+  --xy-controls-button-border-color: var(--color-border);
 }
 
 /* Desktop sidebar visibility must not depend on Tailwind's hidden/lg:flex cascade
diff --git a/src/i18n/messages/de.json b/src/i18n/messages/de.json
index 319f06ff7aa..8ef8690a673 100644
--- a/src/i18n/messages/de.json
+++ b/src/i18n/messages/de.json
@@ -3837,6 +3837,10 @@
     "removeModel": "Modell entfernen",
     "rateLimitProtected": "Geschützt",
     "rateLimitUnprotected": "Ungeschützt",
+    "providerQuotaShort": "Quota",
+    "hideConnectionFromProviderQuota": "Diesen Account auf der Seite „Provider Quota“ ausblenden",
+    "showConnectionInProviderQuota": "Diesen Account auf der Seite „Provider Quota“ anzeigen",
+    "quotaVisibilityUpdateFailed": "Die Sichtbarkeit unter „Provider Quota“ konnte nicht aktualisiert werden",
     "enableRateLimitProtection": "Klicken Sie hier, um den Ratenbegrenzungsschutz zu aktivieren",
     "disableRateLimitProtection": "Klicken Sie hier, um den Ratenbegrenzungsschutz zu deaktivieren",
     "productionKey": "Produktionsschlüssel",
@@ -4530,7 +4534,7 @@
     "proxy": "Stellvertreter",
     "httpProxy": "HTTP Proxy",
     "1proxy": "1proxy",
-    "proxySubTabsAria": "Proxy sections",
+    "proxySubTabsAria": "Proxy-Bereiche",
     "requestBodyLimitTitle": "Request Body Limit",
     "requestBodyLimitDescription": "Maximum API payload size accepted before a request body is parsed. Dedicated upload routes keep at least their built-in 100 MB limit.",
     "requestBodyLimitInputLabel": "Request body limit in MB",
@@ -4582,6 +4586,30 @@
     "syncNow": "Sync Now",
     "syncing": "Syncing...",
     "lastSync": "Last sync",
+    "proxyGlobalConfigTab": "Globale Konfiguration",
+    "proxyPoolTab": "Proxy-Pool",
+    "freePoolTab": "Kostenloser Pool",
+    "proxyDocumentationTab": "Dokumentation",
+    "proxyFreePoolFilterProtocol": "Nach Protokoll filtern",
+    "proxyFreePoolProtocol": "Protokoll",
+    "proxyFreePoolCountryPlaceholder": "Land (z. B. DE)",
+    "proxyFreePoolFilterCountry": "Nach Land filtern",
+    "proxyFreePoolMinQualityPlaceholder": "Mindestqualität",
+    "proxyFreePoolMinQualityLabel": "Minimale Qualitätsbewertung",
+    "proxyFreePoolSyncAll": "Alle synchronisieren",
+    "proxyFreePoolTotal": "Gesamt",
+    "proxyFreePoolInPool": "Im Pool",
+    "proxyFreePoolAvgQuality": "Durchschnittliche Qualität",
+    "proxyFreePoolSelected": "{count} ausgewählt",
+    "proxyFreePoolAddSelected": "Ausgewählte zum Pool hinzufügen",
+    "proxyFreePoolAddVisible": "Alle sichtbaren zum Pool hinzufügen",
+    "proxyFreePoolSource": "Quelle",
+    "proxyFreePoolHostPort": "Host:Port",
+    "proxyFreePoolType": "Typ",
+    "proxyFreePoolCountry": "Land",
+    "proxyFreePoolQuality": "Qualität",
+    "proxyFreePoolLatency": "Latenz",
+    "proxyFreePoolEmpty": "Keine Proxys gefunden. Klicke auf „Alle synchronisieren“, um sie aus den Quellen abzurufen.",
     "never": "Never",
     "justNow": "gerade jetzt",
     "modelsDevStats": "Sync Statistics",
@@ -5163,6 +5191,8 @@
     "compressionAutoTrigger": "Auto-Trigger Threshold",
     "compressionCacheTTL": "Cache TTL",
     "compressionPreserveSystem": "Preserve System Prompt",
+    "compressionLiveZoneTitle": "Cache-ausgerichtete Live Zone",
+    "compressionLiveZoneDesc": "Hält den komprimierten Gesprächspräfix stabil und verarbeitet nur neu angehängte Einträge.",
     "compressionCavemanConfig": "Caveman Engine Configuration",
     "compressionCavemanConfigDesc": "Fine-tune the rule-based compression engine",
     "compressionRoles": "Compress Message Roles",
@@ -5607,7 +5637,26 @@
     "searchFilters": "Suchfilter...",
     "tooltipDedup": "Wie aggressiv wiederholte Zeilen entfernt werden.",
     "tooltipMaxChars": "Maximale Zeichen pro Ausgabeblock.",
-    "tooltipMaxLines": "Maximale Anzahl von Zeilen, die von der Befehlsausgabe ferngehalten werden sollen. Der Überschuss wird abgeschnitten."
+    "tooltipMaxLines": "Maximale Anzahl von Zeilen, die von der Befehlsausgabe ferngehalten werden sollen. Der Überschuss wird abgeschnitten.",
+    "tomlImportTitle": "RTK-TOML-Filter importieren",
+    "tomlImportDesc": "Filter im RTK-TOML-Schema v1 validieren und global installieren. Die Installation schreibt DATA_DIR/rtk/filters.toml; vorhandene Dateien werden nur nach ausdrücklicher Bestätigung ersetzt.",
+    "tomlChooseFile": "TOML-Datei auswählen",
+    "tomlImportPlaceholder": "Inhalt einer RTK-filters.toml hier einfügen ...",
+    "tomlValidate": "Validieren",
+    "tomlValidating": "Wird validiert …",
+    "tomlInstall": "Global installieren",
+    "tomlInstalling": "Wird installiert …",
+    "tomlConfirmOverwrite": "Vorhandene globale Datei ersetzen und Sicherung erstellen",
+    "tomlValidationPassed": "Validierung erfolgreich",
+    "tomlValidationFailed": "Validierung mit fehlgeschlagenen Inline-Tests abgeschlossen",
+    "tomlValidationSummary": "{filters} Filter, {tests} Inline-Tests",
+    "tomlInstalled": "Unter {path} installiert",
+    "tomlBackupCreated": "Eine Sicherung der vorherigen Datei wurde erstellt.",
+    "tomlTestCount": "{count} Tests",
+    "tomlTestPassed": "Bestanden",
+    "tomlTestFailed": "Fehlgeschlagen",
+    "tomlImportError": "Die RTK-TOML-Datei konnte nicht verarbeitet werden.",
+    "tomlFileReadError": "Die ausgewählte TOML-Datei konnte nicht gelesen werden."
   },
   "contextCombos": {
     "title": "Compression Combos",
@@ -6092,7 +6141,21 @@
     "forceRefresh": "Jetzt aktualisieren",
     "resetCreditsLabel": "Reset-Credits",
     "redeemResetCredit": "Reset einlösen",
-    "confirmRedeemResetCredit": "Einen Codex-Reset-Credit für dieses Konto einlösen? Dabei wird genau ein Reset-Credit verbraucht.",
+    "manageResetCredits": "Credits anzeigen",
+    "viewResetCredits": "Reset-Credits anzeigen",
+    "resetCreditsModalTitle": "Codex-Reset-Credits",
+    "resetCreditsModalExplainer": "Die Credits sind nach Ablaufdatum sortiert. Bei der automatischen Einlösung wird immer der Credit verwendet, der als Nächstes abläuft.",
+    "resetCreditsLoadFailed": "Reset-Credits konnten nicht geladen werden",
+    "resetCreditsDetailsUnavailable": "Die Details der Credits sind derzeit nicht verfügbar. Bitte aktualisieren und erneut versuchen.",
+    "noResetCreditsAvailable": "Es sind keine Reset-Credits verfügbar.",
+    "resetCreditDefaultTitle": "Vollständiger Reset",
+    "resetCreditExpiresFirst": "Läuft zuerst ab",
+    "resetCreditExpiresAt": "Läuft am {date} ab",
+    "resetCreditNoExpiry": "Kein Ablaufdatum",
+    "redeemThisResetCredit": "Einlösen",
+    "confirmRedeemResetCreditTitle": "Diesen Reset-Credit einlösen?",
+    "confirmRedeemResetCredit": "Durch das Einlösen werden die berechtigten Codex-Nutzungsfenster sofort zurückgesetzt und dieser Credit dauerhaft verbraucht.",
+    "confirmRedeemResetCreditButton": "Credit einlösen",
     "resetCreditRedeemed": "Reset eingelöst",
     "resetCreditRedeemFailed": "Reset-Credit konnte nicht eingelöst werden",
     "suiteBuilderSaveFailed": "Failed to save suite",
diff --git a/src/i18n/messages/en.json b/src/i18n/messages/en.json
index a8c7e32d57d..c23daa08559 100644
--- a/src/i18n/messages/en.json
+++ b/src/i18n/messages/en.json
@@ -4227,6 +4227,10 @@
     "removeModel": "Remove model",
     "rateLimitProtected": "Protected",
     "rateLimitUnprotected": "Unprotected",
+    "providerQuotaShort": "Quota",
+    "hideConnectionFromProviderQuota": "Hide this account from Provider Quota",
+    "showConnectionInProviderQuota": "Show this account in Provider Quota",
+    "quotaVisibilityUpdateFailed": "Failed to update Provider Quota visibility",
     "enableRateLimitProtection": "Click to enable rate limit protection",
     "disableRateLimitProtection": "Click to disable rate limit protection",
     "testAllProgress": "Tested {done}/{total}",
@@ -5804,6 +5808,8 @@
     "compressionPreserveSystemAlways": "Always",
     "compressionPreserveSystemWhenNoCache": "When no cache",
     "compressionPreserveSystemNever": "Never",
+    "compressionLiveZoneTitle": "Cache-aligned Live Zone",
+    "compressionLiveZoneDesc": "Keep the compressed conversation prefix stable and process only newly appended items.",
     "compressionCavemanConfig": "Caveman Engine Configuration",
     "compressionCavemanConfigDesc": "Fine-tune the rule-based compression engine",
     "compressionCavemanPanelHint": "Its on/off and level are set in the panel:",
@@ -6350,7 +6356,26 @@
     "learnButton": "Suggest filter",
     "learnEmpty": "No suggestion yet. Type a command you have run and scan.",
     "learnSamplesUsed": "Learned from {count} matching sample(s)",
-    "suggestionError": "Could not load suggestions. Check management auth and try again."
+    "suggestionError": "Could not load suggestions. Check management auth and try again.",
+    "tomlImportTitle": "Import RTK TOML filters",
+    "tomlImportDesc": "Validate RTK TOML schema v1 filters and install them globally. Installation writes DATA_DIR/rtk/filters.toml; existing files are replaced only after explicit confirmation.",
+    "tomlChooseFile": "Choose a TOML file",
+    "tomlImportPlaceholder": "Paste RTK filters.toml content here...",
+    "tomlValidate": "Validate",
+    "tomlValidating": "Validating…",
+    "tomlInstall": "Install globally",
+    "tomlInstalling": "Installing…",
+    "tomlConfirmOverwrite": "Replace the existing global file and create a backup",
+    "tomlValidationPassed": "Validation passed",
+    "tomlValidationFailed": "Validation completed with failing inline tests",
+    "tomlValidationSummary": "{filters} filter(s), {tests} inline test(s)",
+    "tomlInstalled": "Installed at {path}",
+    "tomlBackupCreated": "A backup of the previous file was created.",
+    "tomlTestCount": "{count} test(s)",
+    "tomlTestPassed": "Passed",
+    "tomlTestFailed": "Failed",
+    "tomlImportError": "Could not process the RTK TOML file.",
+    "tomlFileReadError": "Could not read the selected TOML file."
   },
   "contextCombos": {
     "title": "Compression Combos",
@@ -6912,7 +6937,21 @@
     "forceRefresh": "Refresh now",
     "resetCreditsLabel": "Reset credits",
     "redeemResetCredit": "Redeem reset",
-    "confirmRedeemResetCredit": "Redeem one Codex reset credit for this account? This consumes one reset credit.",
+    "manageResetCredits": "View credits",
+    "viewResetCredits": "View reset credits",
+    "resetCreditsModalTitle": "Codex reset credits",
+    "resetCreditsModalExplainer": "Credits are ordered by expiration. Automatic redemption always uses the credit that expires first.",
+    "resetCreditsLoadFailed": "Failed to load reset credits",
+    "resetCreditsDetailsUnavailable": "Credit details are currently unavailable. Refresh and try again.",
+    "noResetCreditsAvailable": "No reset credits are available.",
+    "resetCreditDefaultTitle": "Full reset",
+    "resetCreditExpiresFirst": "Expires first",
+    "resetCreditExpiresAt": "Expires {date}",
+    "resetCreditNoExpiry": "No expiration date",
+    "redeemThisResetCredit": "Redeem",
+    "confirmRedeemResetCreditTitle": "Redeem this reset credit?",
+    "confirmRedeemResetCredit": "Redeeming immediately resets the eligible Codex usage windows and permanently consumes this credit.",
+    "confirmRedeemResetCreditButton": "Redeem credit",
     "resetCreditRedeemed": "Reset redeemed",
     "resetCreditRedeemFailed": "Failed to redeem reset credit",
     "suiteBuilderSaveFailed": "Failed to save suite",
diff --git a/src/i18n/messages/pt-BR.json b/src/i18n/messages/pt-BR.json
index af7b8eb5f01..1838f761797 100644
--- a/src/i18n/messages/pt-BR.json
+++ b/src/i18n/messages/pt-BR.json
@@ -4227,6 +4227,10 @@
     "removeModel": "Remover modelo",
     "rateLimitProtected": "Protegido",
     "rateLimitUnprotected": "Desprotegido",
+    "providerQuotaShort": "Cota",
+    "hideConnectionFromProviderQuota": "Ocultar esta conta da página de Cotas dos Provedores",
+    "showConnectionInProviderQuota": "Mostrar esta conta na página de Cotas dos Provedores",
+    "quotaVisibilityUpdateFailed": "Não foi possível atualizar a visibilidade em Cotas dos Provedores",
     "enableRateLimitProtection": "Clique para habilitar proteção de limite de taxa",
     "disableRateLimitProtection": "Clique para desabilitar proteção de limite de taxa",
     "testAllProgress": "__MISSING__:Tested {done}/{total}",
@@ -5766,6 +5770,8 @@
     "compressionPreserveSystemAlways": "Sempre",
     "compressionPreserveSystemWhenNoCache": "Quando não há cache",
     "compressionPreserveSystemNever": "Nunca",
+    "compressionLiveZoneTitle": "Zona ativa alinhada ao cache",
+    "compressionLiveZoneDesc": "Mantém estável o prefixo comprimido da conversa e processa apenas os itens recém-adicionados.",
     "compressionCavemanConfig": "Caveman Engine Configuration",
     "compressionCavemanConfigDesc": "Fine-tune the rule-based compression engine",
     "compressionCavemanPanelHint": "Ativado/desativado e o nível são definidos no painel:",
@@ -6350,7 +6356,26 @@
     "learnButton": "__MISSING__:Suggest filter",
     "learnEmpty": "__MISSING__:No suggestion yet. Type a command you have run and scan.",
     "learnSamplesUsed": "__MISSING__:Learned from {count} matching sample(s)",
-    "suggestionError": "__MISSING__:Could not load suggestions. Check management auth and try again."
+    "suggestionError": "__MISSING__:Could not load suggestions. Check management auth and try again.",
+    "tomlImportTitle": "Importar filtros TOML do RTK",
+    "tomlImportDesc": "Valide os filtros do esquema TOML v1 do RTK e instale-os globalmente. A instalação grava em DATA_DIR/rtk/filters.toml; arquivos existentes só são substituídos após confirmação explícita.",
+    "tomlChooseFile": "Escolher um arquivo TOML",
+    "tomlImportPlaceholder": "Cole aqui o conteúdo do arquivo filters.toml do RTK...",
+    "tomlValidate": "Validar",
+    "tomlValidating": "Validando…",
+    "tomlInstall": "Instalar globalmente",
+    "tomlInstalling": "Instalando…",
+    "tomlConfirmOverwrite": "Substituir o arquivo global existente e criar um backup",
+    "tomlValidationPassed": "Validação bem-sucedida",
+    "tomlValidationFailed": "Validação concluída com falhas nos testes embutidos",
+    "tomlValidationSummary": "{filters} filtro(s), {tests} teste(s) embutido(s)",
+    "tomlInstalled": "Instalado em {path}",
+    "tomlBackupCreated": "Foi criado um backup do arquivo anterior.",
+    "tomlTestCount": "{count} teste(s)",
+    "tomlTestPassed": "Aprovado",
+    "tomlTestFailed": "Reprovado",
+    "tomlImportError": "Não foi possível processar o arquivo TOML do RTK.",
+    "tomlFileReadError": "Não foi possível ler o arquivo TOML selecionado."
   },
   "contextCombos": {
     "title": "Combos de Compressao",
@@ -6912,7 +6937,21 @@
     "forceRefresh": "Atualizar agora",
     "resetCreditsLabel": "Créditos de redefinição",
     "redeemResetCredit": "Resgatar redefinição",
-    "confirmRedeemResetCredit": "Resgatar um crédito de redefinição do Codex para esta conta? Isso consome um crédito de redefinição.",
+    "manageResetCredits": "Ver créditos",
+    "viewResetCredits": "Ver créditos de redefinição",
+    "resetCreditsModalTitle": "Créditos de redefinição do Codex",
+    "resetCreditsModalExplainer": "Os créditos estão ordenados por data de expiração. O resgate automático sempre usa primeiro o crédito que expira antes.",
+    "resetCreditsLoadFailed": "Falha ao carregar os créditos de redefinição",
+    "resetCreditsDetailsUnavailable": "Os detalhes dos créditos não estão disponíveis no momento. Atualize e tente novamente.",
+    "noResetCreditsAvailable": "Não há créditos de redefinição disponíveis.",
+    "resetCreditDefaultTitle": "Redefinição completa",
+    "resetCreditExpiresFirst": "Expira primeiro",
+    "resetCreditExpiresAt": "Expira em {date}",
+    "resetCreditNoExpiry": "Sem data de expiração",
+    "redeemThisResetCredit": "Resgatar",
+    "confirmRedeemResetCreditTitle": "Resgatar este crédito de redefinição?",
+    "confirmRedeemResetCredit": "O resgate redefine imediatamente as janelas de uso elegíveis do Codex e consome este crédito de forma permanente.",
+    "confirmRedeemResetCreditButton": "Resgatar crédito",
     "resetCreditRedeemed": "Redefinição resgatada",
     "resetCreditRedeemFailed": "Falha ao resgatar crédito de redefinição",
     "suiteBuilderSaveFailed": "Falha ao salvar a suíte customizada",
diff --git a/src/instrumentation-node.ts b/src/instrumentation-node.ts
index d32fd67f1ab..d971716f1e6 100755
--- a/src/instrumentation-node.ts
+++ b/src/instrumentation-node.ts
@@ -144,6 +144,59 @@ async function ensureSecrets(): Promise<void> {
   }
 }
 
+/**
+ * Warm the model catalog's durable, apiKey-independent sub-caches at startup
+ * so real /v1/models traffic (any client, any API key) avoids paying their
+ * cold-build cost on first use. Fire-and-forget from the caller, non-fatal.
+ *
+ * getUnifiedModelsResponse()'s own top-level Response cache (`catalogCache`
+ * in catalog.ts) is keyed by prefix/isCodex/apiKey AND has only a 1.5s TTL
+ * (CATALOG_CACHE_TTL_MS — a burst-dedup window added for #6408 to coalesce
+ * concurrent SDK/dashboard requests, not a startup-warm cache). Warming that
+ * cache with an unauthenticated dummy request has essentially no lasting
+ * effect: real traffic almost never arrives within 1.5s of this warmup
+ * completing, regardless of whether its cache key happens to match a real
+ * client's apiKey. The one genuinely durable, apiKey-independent cost in the
+ * catalog build is getOpenRouterCatalog()'s 24h disk-cached network fetch
+ * (src/lib/catalog/openrouterCatalog.ts) — buildUnifiedModelsResponseCore()
+ * calls it unconditionally whenever an OpenRouter connection is configured,
+ * decoupled from the per-key Response cache entirely, so warming it directly
+ * here benefits every subsequent /v1/models request regardless of that
+ * request's own apiKey. Only fetched when an OpenRouter connection actually
+ * exists, so deployments that never use OpenRouter don't pay an unconditional
+ * third-party network call at every boot.
+ *
+ * Exported (rather than left inline in registerNodejs()) so it can be unit
+ * tested directly without exercising the rest of the startup sequence.
+ */
+export async function warmModelCatalogCache(): Promise<void> {
+  try {
+    const { getUnifiedModelsResponse } = await import("@/app/api/v1/models/catalog");
+    await getUnifiedModelsResponse(new Request("http://127.0.0.1/v1/models"));
+    console.log("[STARTUP] Model catalog cache warmed");
+  } catch (err: unknown) {
+    const msg = err instanceof Error ? err.message : String(err);
+    console.warn("[STARTUP] Model catalog warmup failed (non-fatal):", msg);
+  }
+  try {
+    const [{ getProviderConnections }, { getOpenRouterCatalog }] = await Promise.all([
+      import("@/lib/db/providers"),
+      import("@/lib/catalog/openrouterCatalog"),
+    ]);
+    const openrouterConnections = await getProviderConnections({
+      provider: "openrouter",
+      isActive: true,
+    });
+    if (openrouterConnections.length > 0) {
+      await getOpenRouterCatalog();
+      console.log("[STARTUP] OpenRouter model catalog cache warmed");
+    }
+  } catch (err: unknown) {
+    const msg = err instanceof Error ? err.message : String(err);
+    console.warn("[STARTUP] OpenRouter catalog warmup failed (non-fatal):", msg);
+  }
+}
+
 export async function registerNodejs(): Promise<void> {
   // Rename the process title so OmniRoute is identifiable in ps/htop instead
   // of the generic "next-server" standalone server name.
@@ -358,6 +411,11 @@ export async function registerNodejs(): Promise<void> {
     console.warn("[STARTUP] Could not initialize vacuum scheduler (non-fatal):", msg);
   }
 
+  // Warm the model catalog's durable, apiKey-independent sub-caches at
+  // startup — see warmModelCatalogCache() for why the top-level Response
+  // cache alone doesn't deliver this. Fire-and-forget, non-fatal.
+  void warmModelCatalogCache();
+
   if (!isBackgroundServicesDisabled()) {
     try {
       const { bootstrapEmbeddedServices } = await import("@/lib/services/bootstrap");
diff --git a/src/lib/cliTools/batchStatusCache.ts b/src/lib/cliTools/batchStatusCache.ts
index 6fec549febf..5233c023317 100644
--- a/src/lib/cliTools/batchStatusCache.ts
+++ b/src/lib/cliTools/batchStatusCache.ts
@@ -1,14 +1,17 @@
 // DRY: shared between /api/cli-tools/status and /api/cli-tools/all-statuses (plan 14 F2)
 // In-memory mtime-based cache for batch CLI tool status results.
-// Cache invalidated when mtime changes. Lives until server restart (no TTL).
+// Cache invalidated when mtime changes or its short TTL expires.
 
 import type { ToolBatchStatus } from "@/shared/types/cliBatchStatus";
 
 export interface CacheEntry {
   mtimeMs: number;
+  cachedAt: number;
   result: ToolBatchStatus;
 }
 
+const CACHE_TTL_MS = 30_000;
+
 /** Singleton in-memory cache: toolId → { mtimeMs, result } */
 const _cache = new Map<string, CacheEntry>();
 
@@ -18,18 +21,31 @@ const _cache = new Map<string, CacheEntry>();
  *  - entry doesn't exist
  *  - stored mtimeMs !== provided mtimeMs (config file changed)
  */
-export function getCached(toolId: string, mtimeMs: number): ToolBatchStatus | null {
+export function getCached(
+  toolId: string,
+  mtimeMs: number,
+  now: number = Date.now()
+): ToolBatchStatus | null {
   const entry = _cache.get(toolId);
   if (!entry) return null;
   if (entry.mtimeMs !== mtimeMs) return null;
+  if (now - entry.cachedAt >= CACHE_TTL_MS) {
+    _cache.delete(toolId);
+    return null;
+  }
   return entry.result;
 }
 
 /**
  * Store a result in the cache for a toolId with its mtime.
  */
-export function setCached(toolId: string, mtimeMs: number, result: ToolBatchStatus): void {
-  _cache.set(toolId, { mtimeMs, result });
+export function setCached(
+  toolId: string,
+  mtimeMs: number,
+  result: ToolBatchStatus,
+  now: number = Date.now()
+): void {
+  _cache.set(toolId, { mtimeMs, cachedAt: now, result });
 }
 
 /**
diff --git a/src/lib/config/runtimeSettings.ts b/src/lib/config/runtimeSettings.ts
index 406a4bdbc9a..218846bfa0b 100644
--- a/src/lib/config/runtimeSettings.ts
+++ b/src/lib/config/runtimeSettings.ts
@@ -1,5 +1,6 @@
 import { clearHealthCheckLogCache } from "@/lib/tokenHealthCheck";
 import { setCustomBannedSignals } from "@omniroute/open-sse/services/accountFallback.ts";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 
 type JsonRecord = Record<string, unknown>;
 
@@ -86,14 +87,6 @@ function isTruthyEnvFlag(value: string | undefined): boolean {
   return new Set(["1", "true", "yes", "on"]).has(value.trim().toLowerCase());
 }
 
-function isAutomatedTestProcess(): boolean {
-  return (
-    typeof process !== "undefined" &&
-    (process.env.NODE_ENV === "test" ||
-      process.env.VITEST !== undefined ||
-      process.argv.some((arg) => arg.includes("test")))
-  );
-}
 
 function toRecord(value: unknown): JsonRecord {
   return value && typeof value === "object" && !Array.isArray(value) ? (value as JsonRecord) : {};
diff --git a/src/lib/credentialHealth/scheduler.ts b/src/lib/credentialHealth/scheduler.ts
index 0e264b63303..fb5d7a6d312 100644
--- a/src/lib/credentialHealth/scheduler.ts
+++ b/src/lib/credentialHealth/scheduler.ts
@@ -24,6 +24,7 @@ import {
   initCredentialCache,
 } from "@/lib/credentialHealth/cache";
 import { emit } from "@/lib/events/eventBus";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 
 // ── Config ────────────────────────────────────────────────────────────────
 
@@ -66,14 +67,6 @@ function isBuildProcess(): boolean {
   return typeof process !== "undefined" && process.env.NEXT_PHASE === "phase-production-build";
 }
 
-function isAutomatedTestProcess(): boolean {
-  return (
-    typeof process !== "undefined" &&
-    (process.env.NODE_ENV === "test" ||
-      process.env.VITEST !== undefined ||
-      process.argv.some((arg) => arg.includes("test")))
-  );
-}
 
 function isCredentialHealthCheckDisabled(): boolean {
   if (isBuildProcess() || isAutomatedTestProcess()) return true;
diff --git a/src/lib/db/backup.ts b/src/lib/db/backup.ts
index 7bcd1b6a37b..4de29920a1e 100644
--- a/src/lib/db/backup.ts
+++ b/src/lib/db/backup.ts
@@ -14,6 +14,7 @@ import {
   DATA_DIR,
 } from "./core";
 import { resetAllDbModuleState } from "./stateReset";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 
 type CountRow = { cnt?: number };
 
@@ -308,12 +309,7 @@ function isPlainObject(value: unknown): value is Record<string, unknown> {
 }
 
 function isSqliteAutoBackupDisabled() {
-  const isTest =
-    typeof process !== "undefined" &&
-    (process.env.NODE_ENV === "test" ||
-      process.env.VITEST !== undefined ||
-      process.argv.some((a) => a.includes("test")));
-  if (isTest) return true;
+  if (isAutomatedTestProcess()) return true;
 
   const value = process.env.DISABLE_SQLITE_AUTO_BACKUP;
   if (!value) return false;
diff --git a/src/lib/db/caseMapping.ts b/src/lib/db/caseMapping.ts
index 007a3d3d724..1cbb0e602bd 100644
--- a/src/lib/db/caseMapping.ts
+++ b/src/lib/db/caseMapping.ts
@@ -11,6 +11,14 @@
 
 type JsonRecord = Record<string, unknown>;
 
+const BOOLEAN_CAMEL_COLUMNS = new Set([
+  "isActive",
+  "rateLimitProtection",
+  "proxyEnabled",
+  "perKeyProxyEnabled",
+  "quotaVisible",
+]);
+
 export function toSnakeCase(str: string): string {
   return str.replace(/([A-Z])/g, "_$1").toLowerCase();
 }
@@ -33,12 +41,7 @@ export function rowToCamel(row: unknown): JsonRecord | null {
   const result: JsonRecord = {};
   for (const [k, v] of Object.entries(row as JsonRecord)) {
     const camelKey = toCamelCase(k);
-    if (
-      camelKey === "isActive" ||
-      camelKey === "rateLimitProtection" ||
-      camelKey === "proxyEnabled" ||
-      camelKey === "perKeyProxyEnabled"
-    ) {
+    if (BOOLEAN_CAMEL_COLUMNS.has(camelKey)) {
       result[camelKey] = v === 1 || v === true;
     } else if (camelKey === "providerSpecificData" && typeof v === "string") {
       try {
diff --git a/src/lib/db/cleanup.ts b/src/lib/db/cleanup.ts
index 5855c140bb4..c3cf716280f 100644
--- a/src/lib/db/cleanup.ts
+++ b/src/lib/db/cleanup.ts
@@ -244,6 +244,66 @@ export async function cleanupMemoryEntries(): Promise<CleanupResult> {
   return result;
 }
 
+/**
+ * Clean up old domain_cost_history based on retention settings. (#6848)
+ * Uses unix-epoch `timestamp` column (INTEGER).
+ */
+export async function cleanupDomainCostHistory(): Promise<CleanupResult> {
+  const db = getDbInstance();
+  const retention = getRetentionSettings();
+
+  const retentionDays = retention.domainCostHistory;
+  const cutoffEpoch = Math.floor(Date.now() / 1000) - retentionDays * 86_400;
+
+  const result: CleanupResult = { deleted: 0, errors: 0 };
+
+  try {
+    const stmt = db.prepare("DELETE FROM domain_cost_history WHERE timestamp < ?");
+    const runResult = stmt.run(cutoffEpoch);
+    result.deleted = runResult.changes;
+
+    console.log(
+      `[Cleanup] Deleted ${result.deleted} domain_cost_history older than ${retentionDays} days`
+    );
+  } catch (err: unknown) {
+    console.error("[Cleanup] Error cleaning domain_cost_history:", err);
+    result.errors++;
+  }
+
+  return result;
+}
+
+/**
+ * Clean up old compression_cache_stats based on retention settings. (#6848)
+ * Uses `created_at` column (DATETIME string).
+ */
+export async function cleanupCompressionCacheStats(): Promise<CleanupResult> {
+  const db = getDbInstance();
+  const retention = getRetentionSettings();
+
+  const retentionDays = retention.compressionCacheStats;
+  const cutoffDate = new Date();
+  cutoffDate.setDate(cutoffDate.getDate() - retentionDays);
+  const cutoffISO = cutoffDate.toISOString();
+
+  const result: CleanupResult = { deleted: 0, errors: 0 };
+
+  try {
+    const stmt = db.prepare("DELETE FROM compression_cache_stats WHERE created_at < ?");
+    const runResult = stmt.run(cutoffISO);
+    result.deleted = runResult.changes;
+
+    console.log(
+      `[Cleanup] Deleted ${result.deleted} compression_cache_stats older than ${retentionDays} days`
+    );
+  } catch (err: unknown) {
+    console.error("[Cleanup] Error cleaning compression_cache_stats:", err);
+    result.errors++;
+  }
+
+  return result;
+}
+
 /**
  * Clean up old xp_audit_log based on retention settings.
  */
@@ -274,6 +334,35 @@ export async function cleanupXpAuditLog(): Promise<CleanupResult> {
   return result;
 }
 
+/**
+ * Clean up old compression_run_telemetry based on retention settings. (#6848)
+ * Uses unix-epoch `timestamp` column (INTEGER).
+ */
+export async function cleanupCompressionRunTelemetry(): Promise<CleanupResult> {
+  const db = getDbInstance();
+  const retention = getRetentionSettings();
+
+  const retentionDays = retention.compressionRunTelemetry;
+  const cutoffEpoch = Math.floor(Date.now() / 1000) - retentionDays * 86_400;
+
+  const result: CleanupResult = { deleted: 0, errors: 0 };
+
+  try {
+    const stmt = db.prepare("DELETE FROM compression_run_telemetry WHERE timestamp < ?");
+    const runResult = stmt.run(cutoffEpoch);
+    result.deleted = runResult.changes;
+
+    console.log(
+      `[Cleanup] Deleted ${result.deleted} compression_run_telemetry older than ${retentionDays} days`
+    );
+  } catch (err: unknown) {
+    console.error("[Cleanup] Error cleaning compression_run_telemetry:", err);
+    result.errors++;
+  }
+
+  return result;
+}
+
 /**
  * Run all cleanup functions if auto-cleanup is enabled.
  */
@@ -300,7 +389,10 @@ export async function runAutoCleanup(): Promise<{
     mcpAudit: await cleanupMcpAudit(),
     a2aEvents: await cleanupA2aEvents(),
     memoryEntries: await cleanupMemoryEntries(),
+    domainCostHistory: await cleanupDomainCostHistory(),
+    compressionCacheStats: await cleanupCompressionCacheStats(),
     xpAuditLog: await cleanupXpAuditLog(),
+    compressionRunTelemetry: await cleanupCompressionRunTelemetry(),
     proxyLogs: await cleanupProxyLogs(),
   };
 
@@ -523,9 +615,7 @@ export async function cleanupProxyLogs(): Promise<CleanupResult> {
     const runResult = stmt.run(cutoffISO);
     result.deleted = runResult.changes;
 
-    console.log(
-      `[Cleanup] Deleted ${result.deleted} proxy_logs older than ${retentionDays} days`
-    );
+    console.log(`[Cleanup] Deleted ${result.deleted} proxy_logs older than ${retentionDays} days`);
   } catch (err: unknown) {
     console.error("[Cleanup] Error cleaning proxy_logs:", err);
     result.errors++;
@@ -557,9 +647,7 @@ export function startCleanupScheduler(): void {
       const proxyResult = await cleanupProxyLogs();
       const totalDeleted = result.totalDeleted + proxyResult.deleted;
       if (totalDeleted > 0) {
-        console.log(
-          `[Cleanup] Startup cleanup freed ${totalDeleted} rows. Running VACUUM...`
-        );
+        console.log(`[Cleanup] Startup cleanup freed ${totalDeleted} rows. Running VACUUM...`);
         try {
           const db = getDbInstance();
           db.exec("VACUUM");
@@ -580,9 +668,7 @@ export function startCleanupScheduler(): void {
       const proxyResult = await cleanupProxyLogs();
       const totalDeleted = result.totalDeleted + proxyResult.deleted;
       if (totalDeleted > 0) {
-        console.log(
-          `[Cleanup] Periodic cleanup freed ${totalDeleted} rows. Running VACUUM...`
-        );
+        console.log(`[Cleanup] Periodic cleanup freed ${totalDeleted} rows. Running VACUUM...`);
         try {
           const db = getDbInstance();
           db.exec("VACUUM");
diff --git a/src/lib/db/compression.ts b/src/lib/db/compression.ts
index 6bb7e0fd1bd..f015d3687c5 100644
--- a/src/lib/db/compression.ts
+++ b/src/lib/db/compression.ts
@@ -554,6 +554,7 @@ export async function getCompressionSettings(): Promise<CompressionConfig> {
     ultra: normalizeUltraConfig(undefined),
     contextBudget: normalizeContextBudgetConfig(undefined),
     contextEditing: { ...DEFAULT_CONTEXT_EDITING_CONFIG },
+    liveZone: { enabled: false },
     engines: {},
     activeComboId: null,
   };
@@ -661,6 +662,9 @@ export async function getCompressionSettings(): Promise<CompressionConfig> {
       case "contextEditing":
         config.contextEditing = normalizeContextEditingConfig(parsed);
         break;
+      case "liveZone":
+        config.liveZone = { enabled: toRecord(parsed).enabled === true };
+        break;
       case "engines":
         storedEngines = parseStoredEnginesMap(parsed);
         break;
diff --git a/src/lib/db/core.ts b/src/lib/db/core.ts
index 8207b09a8a8..165ecae4567 100644
--- a/src/lib/db/core.ts
+++ b/src/lib/db/core.ts
@@ -36,6 +36,7 @@ import {
 import { migrateLegacyEncryptedString } from "./encryption";
 import { invalidateDbCache } from "./readCache";
 import { rowToCamel } from "./caseMapping";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 // Re-exported so existing call sites that pull these helpers off the core module keep working.
 export { toSnakeCase, toCamelCase, objToSnake, rowToCamel, cleanNulls } from "./caseMapping";
 import {
@@ -240,6 +241,7 @@ const SCHEMA_SQL = `
     max_concurrent INTEGER,
     proxy_enabled INTEGER NOT NULL DEFAULT 1,
     per_key_proxy_enabled INTEGER NOT NULL DEFAULT 0,
+    quota_visible INTEGER NOT NULL DEFAULT 1,
     quota_window_thresholds_json TEXT,
     rate_limit_overrides_json TEXT,
     created_at TEXT NOT NULL,
@@ -248,6 +250,7 @@ const SCHEMA_SQL = `
   CREATE INDEX IF NOT EXISTS idx_pc_provider ON provider_connections(provider);
   CREATE INDEX IF NOT EXISTS idx_pc_active ON provider_connections(is_active);
   CREATE INDEX IF NOT EXISTS idx_pc_priority ON provider_connections(provider, priority);
+  CREATE INDEX IF NOT EXISTS idx_pc_auth_active_refresh ON provider_connections(auth_type, is_active, refresh_token);
 
   CREATE TABLE IF NOT EXISTS provider_nodes (
     id TEXT PRIMARY KEY,
@@ -808,14 +811,6 @@ function offloadLegacyCallLogDetails(db: SqliteDatabase) {
   }
 }
 
-function isAutomatedTestProcess(): boolean {
-  return (
-    typeof process !== "undefined" &&
-    (process.env.NODE_ENV === "test" ||
-      process.env.VITEST !== undefined ||
-      process.argv.some((arg) => arg.includes("test")))
-  );
-}
 
 function shouldRunStartupDbHealthCheck(): boolean {
   if (process.env.OMNIROUTE_FORCE_DB_HEALTHCHECK === "1") return true;
@@ -1041,8 +1036,7 @@ export function getDbInstance(): SqliteDatabase {
         let hasData = false;
         try {
           const count = probe.prepare("SELECT COUNT(*) as c FROM provider_connections").get() as
-            | { c: number }
-            | undefined;
+            { c: number } | undefined;
           hasData = Boolean(count && count.c > 0);
         } catch {
           // Table might not exist at all — truly incompatible
@@ -1096,7 +1090,11 @@ export function getDbInstance(): SqliteDatabase {
       // V8 heap (sql.js loads the whole file into WASM memory). Throwing
       // immediately gives the user a clear "increase --max-old-space-size"
       // signal instead of silently renaming a perfectly good DB.
-      if (/out of memory|allocation failure|Array buffer allocation failed|allocation failed/i.test(message)) {
+      if (
+        /out of memory|allocation failure|Array buffer allocation failed|allocation failed/i.test(
+          message
+        )
+      ) {
         // Cycle-breaker (#6835): the OOM path never renames the file away,
         // so it never trips the generic probe-failed/restore cap above. Cap
         // it independently after 3 consecutive OOM failures (same threshold
@@ -1163,6 +1161,7 @@ export function getDbInstance(): SqliteDatabase {
   db.pragma("busy_timeout = 2000");
   db.pragma("synchronous = NORMAL");
   db.pragma(`cache_size = -${DEFAULT_DATABASE_SETTINGS.optimization.cacheSize}`);
+  db.pragma("temp_store = MEMORY");
   db.exec(SCHEMA_SQL);
   ensureProviderConnectionsColumns(db);
   ensureUsageHistoryColumns(db);
diff --git a/src/lib/db/databaseSettings.ts b/src/lib/db/databaseSettings.ts
index 8d037865e39..f29ac628564 100644
--- a/src/lib/db/databaseSettings.ts
+++ b/src/lib/db/databaseSettings.ts
@@ -49,7 +49,10 @@ const LEGACY_FLAT_KEYS: {
     callLogs: ["callLogs"],
     usageHistory: ["usageHistory"],
     memoryEntries: ["memoryEntries"],
+    domainCostHistory: ["domainCostHistory"],
+    compressionCacheStats: ["compressionCacheStats"],
     xpAuditLog: ["xpAuditLog"],
+    compressionRunTelemetry: ["compressionRunTelemetry"],
     autoCleanupEnabled: ["autoCleanupEnabled"],
   },
   aggregation: {
diff --git a/src/lib/db/migrationRunner.ts b/src/lib/db/migrationRunner.ts
index 982eb1a3222..4d0efa8dd97 100644
--- a/src/lib/db/migrationRunner.ts
+++ b/src/lib/db/migrationRunner.ts
@@ -19,6 +19,7 @@ import path from "path";
 import { fileURLToPath } from "url";
 import type { SqliteAdapter } from "./adapters/types";
 import { DEFAULT_DATABASE_SETTINGS } from "@/types/databaseSettings";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 import {
   RENAMED_MIGRATION_COMPATIBILITY,
   LEGACY_VERSION_SLOT_MIGRATIONS,
@@ -883,10 +884,7 @@ export function runMigrations(db: SqliteAdapter, options?: { isNewDb?: boolean }
 
   // ── Safety Check 2: Mass-migration detection (abort if existing DB + many migrations) ──
   // Skip in test environments where fresh DBs legitimately have many pending migrations.
-  const isTestEnvironment =
-    process.env.NODE_ENV === "test" ||
-    process.env.VITEST !== undefined ||
-    (typeof process.argv !== "undefined" && process.argv.some((arg) => arg.includes("test")));
+  const isTestEnvironment = isAutomatedTestProcess();
 
   // #3416: resolve the threshold at call time so OMNIROUTE_MAX_PENDING_MIGRATIONS
   // can override the default (0 disables the check). The abort message below
diff --git a/src/lib/db/migrations/125_provider_connection_quota_visibility.sql b/src/lib/db/migrations/125_provider_connection_quota_visibility.sql
new file mode 100644
index 00000000000..26bf92df36a
--- /dev/null
+++ b/src/lib/db/migrations/125_provider_connection_quota_visibility.sql
@@ -0,0 +1,5 @@
+-- 125_provider_connection_quota_visibility.sql
+-- Controls whether a provider connection is shown in quota-focused dashboard views.
+-- This is presentation-only and does not affect routing or provider-limit collection.
+
+ALTER TABLE provider_connections ADD COLUMN quota_visible INTEGER NOT NULL DEFAULT 1;
diff --git a/src/lib/db/providers.ts b/src/lib/db/providers.ts
index 1217c860b3c..fdaa5f6a37b 100644
--- a/src/lib/db/providers.ts
+++ b/src/lib/db/providers.ts
@@ -39,11 +39,75 @@ interface DbLike {
   prepare: <TRow = unknown>(sql: string) => StatementLike<TRow>;
 }
 
+// Real column set for provider_connections (must match the CREATE TABLE in
+// core.ts's SCHEMA_SQL). getProviderConnections()'s optional `columns`
+// projection is interpolated directly into the SELECT clause, so every
+// requested name must be validated against this allowlist before use —
+// there is no current caller that passes untrusted input, but the
+// projection API itself must never accept an arbitrary string.
+const PROVIDER_CONNECTIONS_COLUMNS = new Set([
+  "id",
+  "provider",
+  "auth_type",
+  "name",
+  "email",
+  "priority",
+  "is_active",
+  "access_token",
+  "refresh_token",
+  "expires_at",
+  "token_expires_at",
+  "scope",
+  "project_id",
+  "test_status",
+  "error_code",
+  "last_error",
+  "last_error_at",
+  "last_error_type",
+  "last_error_source",
+  "backoff_level",
+  "rate_limited_until",
+  "health_check_interval",
+  "last_health_check_at",
+  "last_tested",
+  "api_key",
+  "id_token",
+  "provider_specific_data",
+  "expires_in",
+  "display_name",
+  "global_priority",
+  "default_model",
+  "token_type",
+  "consecutive_use_count",
+  "rate_limit_protection",
+  "last_used_at",
+  "group",
+  "max_concurrent",
+  "proxy_enabled",
+  "per_key_proxy_enabled",
+  "quota_window_thresholds_json",
+  "rate_limit_overrides_json",
+  "created_at",
+  "updated_at",
+]);
+
 // ──────────────── Provider Connections ────────────────
 
-export async function getProviderConnections(filter: JsonRecord = {}) {
+export async function getProviderConnections(filter: JsonRecord = {}, columns?: string[]) {
   const db = getDbInstance() as unknown as DbLike;
-  let sql = "SELECT * FROM provider_connections";
+  let selectCols = "*";
+  if (columns?.length) {
+    const invalidColumns = columns.filter((col) => !PROVIDER_CONNECTIONS_COLUMNS.has(col));
+    if (invalidColumns.length > 0) {
+      throw new Error(
+        `getProviderConnections: invalid column(s) requested: ${invalidColumns.join(", ")}`
+      );
+    }
+    // "group" is a reserved SQL keyword — the schema declares it quoted, so
+    // it must be re-quoted here too or the generated SELECT is a syntax error.
+    selectCols = columns.map((col) => (col === "group" ? `"group"` : col)).join(", ");
+  }
+  let sql = `SELECT ${selectCols} FROM provider_connections`;
   const conditions: string[] = [];
   const params: Record<string, unknown> = {};
 
@@ -316,6 +380,7 @@ export async function createProviderConnection(data: JsonRecord) {
     updatedAt: now,
     proxyEnabled: normalizeBooleanColumn(data.proxyEnabled, true),
     perKeyProxyEnabled: normalizeBooleanColumn(data.perKeyProxyEnabled, false),
+    quotaVisible: normalizeBooleanColumn(data.quotaVisible, true),
   };
 
   // Optional fields
@@ -347,6 +412,7 @@ export async function createProviderConnection(data: JsonRecord) {
     "maxConcurrent",
     "proxyEnabled",
     "perKeyProxyEnabled",
+    "quotaVisible",
     "quotaWindowThresholds",
     "rateLimitOverrides",
     "healthCheckInterval",
@@ -405,7 +471,7 @@ function _insertConnectionRow(db: DbLike, conn: JsonRecord) {
       last_tested, api_key, id_token, provider_specific_data,
       expires_in, display_name, global_priority, default_model,
       token_type, consecutive_use_count, rate_limit_protection, last_used_at, "group", max_concurrent,
-      proxy_enabled, per_key_proxy_enabled, quota_window_thresholds_json, rate_limit_overrides_json,
+      proxy_enabled, per_key_proxy_enabled, quota_visible, quota_window_thresholds_json, rate_limit_overrides_json,
       created_at, updated_at
     ) VALUES (
       @id, @provider, @authType, @name, @email, @priority, @isActive,
@@ -416,7 +482,7 @@ function _insertConnectionRow(db: DbLike, conn: JsonRecord) {
       @lastTested, @apiKey, @idToken, @providerSpecificData,
       @expiresIn, @displayName, @globalPriority, @defaultModel,
       @tokenType, @consecutiveUseCount, @rateLimitProtection, @lastUsedAt, @group, @maxConcurrent,
-      @proxyEnabled, @perKeyProxyEnabled, @quotaWindowThresholdsJson, @rateLimitOverridesJson,
+      @proxyEnabled, @perKeyProxyEnabled, @quotaVisible, @quotaWindowThresholdsJson, @rateLimitOverridesJson,
       @createdAt, @updatedAt
     )
   `
@@ -463,6 +529,7 @@ function _insertConnectionRow(db: DbLike, conn: JsonRecord) {
     maxConcurrent: conn.maxConcurrent ?? null,
     proxyEnabled: normalizeBooleanColumn(conn.proxyEnabled, true) ? 1 : 0,
     perKeyProxyEnabled: normalizeBooleanColumn(conn.perKeyProxyEnabled, false) ? 1 : 0,
+    quotaVisible: normalizeBooleanColumn(conn.quotaVisible, true) ? 1 : 0,
     quotaWindowThresholdsJson: serializeJsonField(conn.quotaWindowThresholds),
     rateLimitOverridesJson: serializeJsonField(conn.rateLimitOverrides),
     createdAt: conn.createdAt,
@@ -470,37 +537,11 @@ function _insertConnectionRow(db: DbLike, conn: JsonRecord) {
   });
 }
 
-function _updateConnectionRow(db: DbLike, id: string, data: JsonRecord) {
-  const now = data.updatedAt || new Date().toISOString();
-  db.prepare(
-    `
-    UPDATE provider_connections SET
-      provider = @provider, auth_type = @authType, name = @name, email = @email,
-      priority = @priority, is_active = @isActive, access_token = @accessToken,
-      refresh_token = @refreshToken, expires_at = @expiresAt, token_expires_at = @tokenExpiresAt,
-      scope = @scope, project_id = @projectId, test_status = @testStatus, error_code = @errorCode,
-      last_error = @lastError, last_error_at = @lastErrorAt, last_error_type = @lastErrorType,
-      last_error_source = @lastErrorSource, backoff_level = @backoffLevel,
-      rate_limited_until = @rateLimitedUntil, health_check_interval = @healthCheckInterval,
-      last_health_check_at = @lastHealthCheckAt, last_tested = @lastTested, api_key = @apiKey,
-      id_token = @idToken, provider_specific_data = @providerSpecificData,
-      expires_in = @expiresIn, display_name = @displayName, global_priority = @globalPriority,
-      default_model = @defaultModel, token_type = @tokenType,
-      consecutive_use_count = @consecutiveUseCount,
-      rate_limit_protection = @rateLimitProtection,
-      last_used_at = @lastUsedAt,
-      "group" = @group,
-      max_concurrent = @maxConcurrent,
-      quota_window_thresholds_json = @quotaWindowThresholdsJson,
-      proxy_enabled = @proxyEnabled,
-      per_key_proxy_enabled = @perKeyProxyEnabled,
-      rate_limit_overrides_json = @rateLimitOverridesJson,
-      last_ping_at = @lastPingAt,
-      last_pinged_reset_key = @lastPingedResetKey,
-      updated_at = @updatedAt
-    WHERE id = @id
-  `
-  ).run({
+// Assembles the `.run()` params for _updateConnectionRow's UPDATE statement.
+// Split out purely to keep _updateConnectionRow under the max-lines-per-function
+// gate — same field mapping/normalization as before, just relocated.
+function _buildUpdateConnectionRowParams(id: string, data: JsonRecord, now: unknown) {
+  return {
     id,
     provider: data.provider,
     authType: data.authType || null,
@@ -544,11 +585,46 @@ function _updateConnectionRow(db: DbLike, id: string, data: JsonRecord) {
     quotaWindowThresholdsJson: serializeJsonField(data.quotaWindowThresholds),
     proxyEnabled: normalizeBooleanColumn(data.proxyEnabled, true) ? 1 : 0,
     perKeyProxyEnabled: normalizeBooleanColumn(data.perKeyProxyEnabled, false) ? 1 : 0,
+    quotaVisible: normalizeBooleanColumn(data.quotaVisible, true) ? 1 : 0,
     rateLimitOverridesJson: serializeJsonField(data.rateLimitOverrides),
     lastPingAt: data.lastPingAt || null,
     lastPingedResetKey: data.lastPingedResetKey || null,
     updatedAt: now,
-  });
+  };
+}
+
+function _updateConnectionRow(db: DbLike, id: string, data: JsonRecord) {
+  const now = data.updatedAt || new Date().toISOString();
+  db.prepare(
+    `
+    UPDATE provider_connections SET
+      provider = @provider, auth_type = @authType, name = @name, email = @email,
+      priority = @priority, is_active = @isActive, access_token = @accessToken,
+      refresh_token = @refreshToken, expires_at = @expiresAt, token_expires_at = @tokenExpiresAt,
+      scope = @scope, project_id = @projectId, test_status = @testStatus, error_code = @errorCode,
+      last_error = @lastError, last_error_at = @lastErrorAt, last_error_type = @lastErrorType,
+      last_error_source = @lastErrorSource, backoff_level = @backoffLevel,
+      rate_limited_until = @rateLimitedUntil, health_check_interval = @healthCheckInterval,
+      last_health_check_at = @lastHealthCheckAt, last_tested = @lastTested, api_key = @apiKey,
+      id_token = @idToken, provider_specific_data = @providerSpecificData,
+      expires_in = @expiresIn, display_name = @displayName, global_priority = @globalPriority,
+      default_model = @defaultModel, token_type = @tokenType,
+      consecutive_use_count = @consecutiveUseCount,
+      rate_limit_protection = @rateLimitProtection,
+      last_used_at = @lastUsedAt,
+      "group" = @group,
+      max_concurrent = @maxConcurrent,
+      quota_window_thresholds_json = @quotaWindowThresholdsJson,
+      proxy_enabled = @proxyEnabled,
+      per_key_proxy_enabled = @perKeyProxyEnabled,
+      quota_visible = @quotaVisible,
+      rate_limit_overrides_json = @rateLimitOverridesJson,
+      last_ping_at = @lastPingAt,
+      last_pinged_reset_key = @lastPingedResetKey,
+      updated_at = @updatedAt
+    WHERE id = @id
+  `
+  ).run(_buildUpdateConnectionRowParams(id, data, now));
 }
 
 export async function updateProviderConnection(id: string, data: JsonRecord) {
@@ -621,8 +697,9 @@ export async function clearConnectionErrorIfUnchanged(
   }
 ): Promise<boolean> {
   const db = getDbInstance() as unknown as DbLike;
-  const result = db.prepare(
-    `
+  const result = db
+    .prepare(
+      `
     UPDATE provider_connections SET
       test_status = 'active',
       last_error = NULL,
@@ -638,13 +715,14 @@ export async function clearConnectionErrorIfUnchanged(
       AND IFNULL(last_error_at, '') = ?
       AND IFNULL(rate_limited_until, '') = ?
     `
-  ).run(
-    new Date().toISOString(),
-    id,
-    expected.testStatus ?? "",
-    expected.lastErrorAt ?? "",
-    expected.rateLimitedUntil ?? ""
-  );
+    )
+    .run(
+      new Date().toISOString(),
+      id,
+      expected.testStatus ?? "",
+      expected.lastErrorAt ?? "",
+      expected.rateLimitedUntil ?? ""
+    );
   const applied = (result.changes ?? 0) > 0;
   if (applied) {
     backupDbFile("pre-write");
diff --git a/src/lib/db/proxies.ts b/src/lib/db/proxies.ts
index 1f051280455..d0742dfd172 100755
--- a/src/lib/db/proxies.ts
+++ b/src/lib/db/proxies.ts
@@ -31,6 +31,11 @@ import {
   coerceProxyPayload,
   redactProxySecrets,
 } from "./proxies/mappers";
+import { isGlobalProxyEnabled, PROXY_ALIVE_PREDICATE } from "./proxies/guards";
+export {
+  hasBlockingProxyAssignment,
+  hasBlockingProxyAssignmentForProvider,
+} from "./proxies/guards";
 export { extractRelayAuth, redactProxySecrets } from "./proxies/mappers";
 
 let proxyRegistryGeneration = 0;
@@ -291,9 +296,14 @@ export async function createProxy(payload: ProxyPayload) {
 }
 
 /**
- * Upsert a proxy by host+port.
- * If a proxy with the same host and port already exists, update it.
- * Otherwise, create a new one. Used by the bulk import feature.
+ * Upsert a proxy by its credential tuple (host+port+username+password).
+ * If a proxy with the same host, port, username AND password already exists,
+ * update it. Otherwise, create a new one. Used by the bulk import feature.
+ *
+ * #7594: host+port alone is NOT a stable identity. Rotating residential/gateway
+ * proxies route every credential through one shared host:port, so keying only on
+ * host+port collapsed distinct-credential imports onto the first existing row
+ * (the same entry got "updated" N times instead of N entries being created).
  */
 export async function upsertProxy(payload: ProxyPayload): Promise<{
   proxy: ProxyRegistryRecord | null;
@@ -302,10 +312,14 @@ export async function upsertProxy(payload: ProxyPayload): Promise<{
   const db = getDbInstance();
   const host = (payload.host || "").trim();
   const port = Number(payload.port);
+  const username = (payload.username || "").trim();
+  const password = (payload.password || "").trim();
 
   const existing = db
-    .prepare("SELECT id FROM proxy_registry WHERE host = ? AND port = ? LIMIT 1")
-    .get(host, port) as { id?: string } | undefined;
+    .prepare(
+      "SELECT id FROM proxy_registry WHERE host = ? AND port = ? AND username = ? AND password = ? LIMIT 1"
+    )
+    .get(host, port, username, password) as { id?: string } | undefined;
 
   if (existing?.id) {
     const updated = await updateProxy(existing.id, payload);
@@ -822,9 +836,6 @@ export async function deleteProxyById(id: string, options?: { force?: boolean })
 // usable so a working proxy is never stranded; only known-dead states are
 // excluded so a dead proxy stops being handed out (every request would
 // otherwise pay the timeout or leak out the host IP).
-const PROXY_ALIVE_PREDICATE =
-  "(p.status IS NULL OR LOWER(p.status) NOT IN ('inactive','error','disabled','dead','down'))";
-
 // Resolve one scope's alive pool to a single proxy via its rotation strategy.
 // Returns the standard registry resolution shape, or null when the pool is empty
 // or every member is dead (preserving the #6246 fail-closed contract — a dead
@@ -908,59 +919,6 @@ export async function resolveProxyForScopeFromRegistry(scope: string, scopeId?:
   }
 }
 
-/**
- * #6246 fail-closed guard. Returns true when a connection would egress DIRECTLY
- * ONLY because its ASSIGNED proxy (account/provider/global scope) is dead/inactive
- * — i.e. the request must be BLOCKED, not silently sent on the real IP.
- *
- * Callers use this after `resolveProxyForConnection` returns a direct result: if
- * the operator assigned a proxy but every assigned proxy is dead, leaking the IP
- * is worse than failing the request. An explicit "proxy off" (global or per
- * connection) is a deliberate direct choice and is NOT treated as a leak. Read-only
- * and best-effort: any DB error fails OPEN (returns false) so a guard never breaks
- * the request path.
- */
-export function hasBlockingProxyAssignment(connectionId: string): boolean {
-  try {
-    const db = getDbInstance();
-
-    // Explicit global "proxy off" → direct is intended, never a leak.
-    const globalRow = db
-      .prepare("SELECT value FROM key_value WHERE namespace = 'settings' AND key = 'proxyEnabled'")
-      .get() as { value?: string } | undefined;
-    if (globalRow?.value) {
-      try {
-        if (JSON.parse(globalRow.value) === false) return false;
-      } catch {
-        /* malformed → treat as enabled */
-      }
-    }
-
-    // Explicit per-connection "proxy off" → direct is intended.
-    const conn = db
-      .prepare("SELECT provider, proxy_enabled FROM provider_connections WHERE id = ?")
-      .get(connectionId) as { provider?: string | null; proxy_enabled?: number } | undefined;
-    if (conn && conn.proxy_enabled === 0) return false;
-    const provider = conn?.provider ?? null;
-
-    // A proxy is assigned to this connection at some scope, but every assigned
-    // proxy is dead (the alive filter would have resolved a live one already).
-    const dead = db
-      .prepare(
-        `SELECT 1 FROM proxy_assignments a JOIN proxy_registry p ON p.id = a.proxy_id
-           WHERE ((a.scope = 'account' AND a.scope_id = ?)
-               OR (a.scope = 'provider' AND a.scope_id = ?)
-               OR (a.scope = 'global'))
-             AND NOT ${PROXY_ALIVE_PREDICATE}
-           LIMIT 1`
-      )
-      .get(connectionId, provider);
-    return !!dead;
-  } catch {
-    return false;
-  }
-}
-
 export async function migrateLegacyProxyConfigToRegistry(options?: { force?: boolean }) {
   const force = options?.force === true;
   const db = getDbInstance();
@@ -1127,6 +1085,9 @@ export async function bulkAssignProxyToScope(
  */
 export async function resolveProxyForProvider(providerId: string) {
   try {
+    const db = getDbInstance();
+    if (!isGlobalProxyEnabled(db)) return null;
+
     // Resolve by specificity across both storage backends. The GUI Custom tab
     // still writes provider/global proxies to the legacy config, while Saved
     // Proxy uses the registry. A registry-global fallback must not shadow a
diff --git a/src/lib/db/proxies/guards.ts b/src/lib/db/proxies/guards.ts
new file mode 100644
index 00000000000..d0fd7c5069f
--- /dev/null
+++ b/src/lib/db/proxies/guards.ts
@@ -0,0 +1,81 @@
+import { getDbInstance } from "../core";
+
+export const PROXY_ALIVE_PREDICATE =
+  "(p.status IS NULL OR LOWER(p.status) NOT IN ('inactive','error','disabled','dead','down'))";
+
+export function isGlobalProxyEnabled(db: ReturnType<typeof getDbInstance>): boolean {
+  try {
+    const row = db
+      .prepare("SELECT value FROM key_value WHERE namespace = 'settings' AND key = 'proxyEnabled'")
+      .get() as { value?: string } | undefined;
+    if (!row?.value) return true;
+    try {
+      return JSON.parse(row.value) !== false;
+    } catch {
+      return true;
+    }
+  } catch {
+    return true;
+  }
+}
+
+/**
+ * #6246 fail-closed guard for a connection with an assigned dead proxy pool.
+ * Explicitly disabling proxying globally or for the connection allows direct egress.
+ */
+export function hasBlockingProxyAssignment(connectionId: string): boolean {
+  try {
+    const db = getDbInstance();
+    if (!isGlobalProxyEnabled(db)) return false;
+
+    const conn = db
+      .prepare("SELECT provider, proxy_enabled FROM provider_connections WHERE id = ?")
+      .get(connectionId) as { provider?: string | null; proxy_enabled?: number } | undefined;
+    if (conn && conn.proxy_enabled === 0) return false;
+    const provider = conn?.provider ?? null;
+    const dead = db
+      .prepare(
+        `SELECT 1 FROM proxy_assignments a JOIN proxy_registry p ON p.id = a.proxy_id
+           WHERE ((a.scope = 'account' AND a.scope_id = ?)
+               OR (a.scope = 'provider' AND a.scope_id = ?)
+               OR (a.scope = 'global'))
+             AND NOT ${PROXY_ALIVE_PREDICATE}
+           LIMIT 1`
+      )
+      .get(connectionId, provider);
+    return !!dead;
+  } catch {
+    return false;
+  }
+}
+
+/**
+ * #7380 fail-closed guard for providers without a connection row. Returns true
+ * when a provider/global proxy assignment exists but all assigned proxies are known dead.
+ */
+export function hasBlockingProxyAssignmentForProvider(providerId: string): boolean {
+  try {
+    const db = getDbInstance();
+    if (!isGlobalProxyEnabled(db)) return false;
+
+    const assignments = db
+      .prepare(
+        `SELECT
+           EXISTS(
+             SELECT 1 FROM proxy_assignments a
+             WHERE ((a.scope = 'provider' AND a.scope_id = ?)
+                 OR a.scope = 'global')
+           ) AS assigned,
+           EXISTS(
+             SELECT 1 FROM proxy_assignments a JOIN proxy_registry p ON p.id = a.proxy_id
+             WHERE ((a.scope = 'provider' AND a.scope_id = ?)
+                 OR a.scope = 'global')
+               AND ${PROXY_ALIVE_PREDICATE}
+           ) AS alive`
+      )
+      .get(providerId, providerId) as { assigned?: number; alive?: number } | undefined;
+    return assignments?.assigned === 1 && assignments.alive === 0;
+  } catch {
+    return false;
+  }
+}
diff --git a/src/lib/db/schemaColumns.ts b/src/lib/db/schemaColumns.ts
index ab8ec16525f..2951e214e83 100644
--- a/src/lib/db/schemaColumns.ts
+++ b/src/lib/db/schemaColumns.ts
@@ -48,6 +48,12 @@ export function ensureProviderConnectionsColumns(db: SqliteDatabase) {
       );
       console.log("[DB] Added provider_connections.per_key_proxy_enabled column");
     }
+    if (!columnNames.has("quota_visible")) {
+      db.exec(
+        "ALTER TABLE provider_connections ADD COLUMN quota_visible INTEGER NOT NULL DEFAULT 1"
+      );
+      console.log("[DB] Added provider_connections.quota_visible column");
+    }
     if (!columnNames.has("quota_window_thresholds_json")) {
       db.exec("ALTER TABLE provider_connections ADD COLUMN quota_window_thresholds_json TEXT");
       console.log("[DB] Added provider_connections.quota_window_thresholds_json column");
diff --git a/src/lib/initCloudSync.ts b/src/lib/initCloudSync.ts
index 78b26e89bf0..bc061042c77 100644
--- a/src/lib/initCloudSync.ts
+++ b/src/lib/initCloudSync.ts
@@ -1,18 +1,11 @@
 import initializeCloudSync from "@/shared/services/initializeCloudSync";
 import { startBudgetResetJob } from "@/lib/jobs/budgetResetJob";
 import { startModelSyncScheduler } from "@/shared/services/modelSyncScheduler";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 
 // Initialize runtime background sync services once per server process.
 let initialized = false;
 
-function isAutomatedTestProcess(
-  env: NodeJS.ProcessEnv = process.env,
-  argv: string[] = process.argv
-): boolean {
-  return (
-    env.NODE_ENV === "test" || env.VITEST !== undefined || argv.some((arg) => arg.includes("test"))
-  );
-}
 
 export function shouldSkipCloudSyncInitialization(
   env: NodeJS.ProcessEnv = process.env,
diff --git a/src/lib/issueAgent/audit.ts b/src/lib/issueAgent/audit.ts
new file mode 100644
index 00000000000..9abbff516ea
--- /dev/null
+++ b/src/lib/issueAgent/audit.ts
@@ -0,0 +1,43 @@
+import { mkdir, appendFile } from "node:fs/promises";
+import { join } from "node:path";
+import { homedir } from "node:os";
+
+import type { RecordedTriageRun } from "./recordedTriage";
+
+export interface IssueAgentAuditOptions {
+  dataDir?: string;
+  now?: Date;
+}
+
+export interface IssueAgentAuditResult {
+  path: string;
+}
+
+function defaultDataDir(): string {
+  return process.env.DATA_DIR || join(homedir(), ".omniroute");
+}
+
+export async function appendIssueAgentAuditRecord(
+  run: RecordedTriageRun,
+  options: IssueAgentAuditOptions = {}
+): Promise<IssueAgentAuditResult> {
+  const dataDir = options.dataDir || defaultDataDir();
+  const auditDir = join(dataDir, "issue-agent");
+  const auditPath = join(auditDir, "audit.jsonl");
+  const row = {
+    ts: (options.now ?? new Date()).toISOString(),
+    runId: run.runId,
+    mode: run.mode,
+    repository: run.repository,
+    issueNumber: run.issueNumber,
+    issueUrl: run.issueUrl,
+    dryRun: run.dryRun,
+    runner: run.runner,
+    context: run.context,
+    steps: run.steps,
+  };
+
+  await mkdir(auditDir, { recursive: true });
+  await appendFile(auditPath, `${JSON.stringify(row)}\n`, "utf8");
+  return { path: auditPath };
+}
diff --git a/src/lib/issueAgent/execution.ts b/src/lib/issueAgent/execution.ts
new file mode 100644
index 00000000000..85243de010f
--- /dev/null
+++ b/src/lib/issueAgent/execution.ts
@@ -0,0 +1,114 @@
+import type { RecordedTriageRun } from "@/lib/issueAgent/recordedTriage";
+
+export interface RecordedTriageExecutionInput {
+  run: RecordedTriageRun;
+  model?: string;
+  provider?: string;
+  routingPolicy?: string;
+  timeoutMs?: number;
+}
+
+export type ChatCompletionsPost = (request: Request) => Promise<Response>;
+
+export interface RecordedTriageChatCompletion {
+  status: number;
+  body: unknown;
+}
+
+export class RecordedTriageTimeoutError extends Error {
+  constructor(timeoutMs: number) {
+    super(`Issue Agent triage timed out after ${timeoutMs}ms`);
+    this.name = "RecordedTriageTimeoutError";
+  }
+}
+
+const DEFAULT_MODEL = "auto/quality";
+const DEFAULT_TIMEOUT_MS = 60_000;
+const MAX_TIMEOUT_MS = 120_000;
+
+function configuredString(value: string | undefined, envName: string): string | undefined {
+  const configured = value ?? process.env[envName];
+  const normalized = configured?.trim();
+  return normalized || undefined;
+}
+
+function resolveModel(input: RecordedTriageExecutionInput): string {
+  const model = configuredString(input.model, "OMNIROUTE_ISSUE_AGENT_MODEL") ?? DEFAULT_MODEL;
+  const provider = configuredString(input.provider, "OMNIROUTE_ISSUE_AGENT_PROVIDER");
+
+  if (!provider || model.includes("/")) return model;
+  return `${provider}/${model}`;
+}
+
+function resolveTimeoutMs(input: RecordedTriageExecutionInput): number {
+  const configured = input.timeoutMs ?? Number(process.env.OMNIROUTE_ISSUE_AGENT_TIMEOUT_MS);
+  if (!Number.isFinite(configured) || configured <= 0) return DEFAULT_TIMEOUT_MS;
+  return Math.min(Math.floor(configured), MAX_TIMEOUT_MS);
+}
+
+function buildMessages(run: RecordedTriageRun) {
+  return [
+    {
+      role: "system",
+      content:
+        "You are the OmniRoute Issue Agent. Analyze only the recorded GitHub context and produce a concise, actionable triage response. Do not claim to have accessed external state.",
+    },
+    {
+      role: "user",
+      content: [
+        `Issue: ${run.repository}#${run.issueNumber}`,
+        `URL: ${run.issueUrl}`,
+        `Intent: ${run.context.intent}`,
+        `Title: ${run.context.issueTitle ?? "(untitled)"}`,
+        "Recorded context:",
+        run.context.redactedDigestSource || "(none)",
+      ].join("\n"),
+    },
+  ];
+}
+
+/**
+ * Route recorded-triage work through the same in-process chat endpoint used by
+ * clients, retaining its initialization, admission, guardrails, and policy path.
+ */
+export async function executeRecordedTriageChatCompletion(
+  input: RecordedTriageExecutionInput,
+  post: ChatCompletionsPost
+): Promise<RecordedTriageChatCompletion> {
+  const controller = new AbortController();
+  const timeoutMs = resolveTimeoutMs(input);
+  const timeout = setTimeout(() => controller.abort(), timeoutMs);
+  const routingPolicy = configuredString(
+    input.routingPolicy,
+    "OMNIROUTE_ISSUE_AGENT_ROUTING_POLICY"
+  );
+
+  try {
+    const response = await post(
+      new Request("http://localhost/api/v1/chat/completions", {
+        method: "POST",
+        signal: controller.signal,
+        headers: {
+          "Content-Type": "application/json",
+          ...(routingPolicy ? { "X-OmniRoute-Mode": routingPolicy } : {}),
+        },
+        body: JSON.stringify({
+          model: resolveModel(input),
+          messages: buildMessages(input.run),
+          max_tokens: 1200,
+          temperature: 0,
+          stream: false,
+        }),
+      })
+    );
+
+    return { status: response.status, body: await response.json() };
+  } catch (error) {
+    if (error instanceof Error && error.name === "AbortError") {
+      throw new RecordedTriageTimeoutError(timeoutMs);
+    }
+    throw error;
+  } finally {
+    clearTimeout(timeout);
+  }
+}
diff --git a/src/lib/issueAgent/githubExport.ts b/src/lib/issueAgent/githubExport.ts
new file mode 100644
index 00000000000..e478ad7f4af
--- /dev/null
+++ b/src/lib/issueAgent/githubExport.ts
@@ -0,0 +1,57 @@
+import type { RecordedTriageComment, RecordedTriageContextInput } from "./recordedTriage";
+
+export interface NormalizedGitHubIssueExport {
+  issueUrl: string;
+  recordedContext: RecordedTriageContextInput;
+}
+
+function asRecord(value: unknown): Record<string, unknown> {
+  return value && typeof value === "object" ? (value as Record<string, unknown>) : {};
+}
+
+function readString(value: unknown): string | undefined {
+  return typeof value === "string" && value.trim().length > 0 ? value : undefined;
+}
+
+function isBotUser(user: Record<string, unknown>, fallbackAuthor: string | undefined): boolean {
+  const type = readString(user.type);
+  if (type?.toLowerCase() === "bot") return true;
+  return /\[bot\]$/i.test(fallbackAuthor ?? "");
+}
+
+function normalizeComment(value: unknown): RecordedTriageComment {
+  const row = asRecord(value);
+  const user = asRecord(row.user);
+  const author = readString(row.author) ?? readString(row.user_login) ?? readString(user.login);
+  const body = readString(row.body) ?? "";
+  const explicitBot = typeof row.isBot === "boolean" ? row.isBot : undefined;
+  return {
+    author,
+    body,
+    isBot: explicitBot ?? isBotUser(user, author),
+  };
+}
+
+export function normalizeGitHubIssueExport(value: unknown): NormalizedGitHubIssueExport {
+  const root = asRecord(value);
+  const issue = asRecord(root.issue ?? root.pull_request ?? root.pr ?? root);
+  const issueUrl = readString(issue.html_url) ?? readString(issue.url) ?? readString(root.html_url) ?? readString(root.url);
+  if (!issueUrl) {
+    throw new Error("Recorded GitHub export must include html_url or url");
+  }
+
+  const rawComments = Array.isArray(root.comments)
+    ? root.comments
+    : Array.isArray(issue.comments)
+      ? issue.comments
+      : [];
+
+  return {
+    issueUrl,
+    recordedContext: {
+      title: readString(issue.title),
+      body: readString(issue.body),
+      comments: rawComments.slice(0, 100).map(normalizeComment),
+    },
+  };
+}
diff --git a/src/lib/issueAgent/recordedTriage.ts b/src/lib/issueAgent/recordedTriage.ts
new file mode 100644
index 00000000000..661ab58c48b
--- /dev/null
+++ b/src/lib/issueAgent/recordedTriage.ts
@@ -0,0 +1,124 @@
+import { createHash } from "node:crypto";
+import { redactSecrets } from "@/shared/utils/logRedaction";
+
+export interface RecordedTriageComment {
+  author?: string;
+  body?: string;
+  isBot?: boolean;
+}
+
+export interface RecordedTriageContextInput {
+  title?: string;
+  body?: string;
+  comments?: RecordedTriageComment[];
+}
+
+export interface RecordedTriageInput {
+  issueUrl?: string;
+  dryRun?: boolean;
+  recordedContext?: RecordedTriageContextInput;
+}
+
+export interface RecordedTriageContextSummary {
+  issueTitle: string | null;
+  commentCount: number;
+  humanCommentCount: number;
+  botCommentCount: number;
+  intent: "bugfix" | "review" | "question" | "unknown";
+  redactedDigestSource: string;
+}
+
+export interface RecordedTriageRun {
+  accepted: true;
+  mode: "recorded-triage";
+  runner: "deterministic-recorded-triage" | "omniroute-chat-completions";
+  runId: string;
+  issueUrl: string;
+  repository: string;
+  issueNumber: number;
+  dryRun: boolean;
+  context: RecordedTriageContextSummary;
+  steps: string[];
+}
+
+const GITHUB_ISSUE_URL =
+  /^https:\/\/(?:[^@/\s]+@)?github\.com\/([^/\s]+)\/([^/\s]+)\/(?:issues|pull)\/(\d+)(?:[/?#].*)?$/i;
+
+const RECORDED_TRIAGE_STEPS = [
+  "load-recorded-github-context",
+  "classify-mention-intent",
+  "draft-safe-response-plan",
+  "emit-audit-record",
+];
+
+function redactUrlCredentials(issueUrl: string): string {
+  return issueUrl.replace(/^https:\/\/[^@/\s]+@github\.com\//i, "https://[REDACTED]@github.com/");
+}
+
+function buildRunId(repository: string, issueNumber: number): string {
+  const digest = createHash("sha256")
+    .update(`${repository}#${issueNumber}`)
+    .digest("hex")
+    .slice(0, 16);
+  return `issue-agent-recorded-triage-${digest}`;
+}
+
+function normalizeText(value: unknown, maxLength = 400): string {
+  if (typeof value !== "string") return "";
+  return value.trim().replace(/\s+/g, " ").slice(0, maxLength);
+}
+
+function classifyIntent(text: string): RecordedTriageContextSummary["intent"] {
+  const lower = text.toLowerCase();
+  if (/\b(bug|fix|failing|regression|broken|patch)\b/.test(lower)) return "bugfix";
+  if (/\b(review|pr|pull request|approve|merge)\b/.test(lower)) return "review";
+  if (/\b(question|why|how|what|help)\b/.test(lower)) return "question";
+  return "unknown";
+}
+
+function summarizeContext(
+  input: RecordedTriageContextInput | undefined
+): RecordedTriageContextSummary {
+  const comments = Array.isArray(input?.comments) ? input.comments.slice(0, 50) : [];
+  const title = normalizeText(input?.title, 160);
+  const body = normalizeText(input?.body);
+  const commentTexts = comments.map((comment) => normalizeText(comment.body, 200)).filter(Boolean);
+  const digestSource = redactSecrets([title, body, ...commentTexts].filter(Boolean).join("\n"));
+  const botCommentCount = comments.filter((comment) => comment.isBot === true).length;
+  const humanCommentCount = comments.length - botCommentCount;
+
+  return {
+    issueTitle: title || null,
+    commentCount: comments.length,
+    humanCommentCount,
+    botCommentCount,
+    intent: classifyIntent(digestSource),
+    redactedDigestSource: digestSource.slice(0, 1200),
+  };
+}
+
+export function createRecordedTriageRun(input: RecordedTriageInput): RecordedTriageRun {
+  const issueUrl = typeof input.issueUrl === "string" ? input.issueUrl.trim() : "";
+  const match = GITHUB_ISSUE_URL.exec(issueUrl);
+  if (!match) {
+    throw new Error("Expected a GitHub issue or pull request URL");
+  }
+
+  const owner = match[1]!;
+  const repo = match[2]!;
+  const issueNumber = Number(match[3]);
+  const repository = `${owner}/${repo}`;
+
+  return {
+    accepted: true,
+    mode: "recorded-triage",
+    runner: "deterministic-recorded-triage",
+    runId: buildRunId(repository, issueNumber),
+    issueUrl: redactUrlCredentials(issueUrl),
+    repository,
+    issueNumber,
+    dryRun: input.dryRun !== false,
+    context: summarizeContext(input.recordedContext),
+    steps: [...RECORDED_TRIAGE_STEPS],
+  };
+}
diff --git a/src/lib/localHealthCheck.ts b/src/lib/localHealthCheck.ts
index e61a35aac44..938ece9e54b 100644
--- a/src/lib/localHealthCheck.ts
+++ b/src/lib/localHealthCheck.ts
@@ -12,6 +12,7 @@
  */
 
 import { getProviderNodes } from "@/lib/localDb";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 
 // ── Types ────────────────────────────────────────────────────────────────
 
@@ -37,14 +38,6 @@ function isBuildProcess(): boolean {
   return typeof process !== "undefined" && process.env.NEXT_PHASE === "phase-production-build";
 }
 
-function isAutomatedTestProcess(): boolean {
-  return (
-    typeof process !== "undefined" &&
-    (process.env.NODE_ENV === "test" ||
-      process.env.VITEST !== undefined ||
-      process.argv.some((arg) => arg.includes("test")))
-  );
-}
 
 // ── State (globalThis survives HMR re-evaluation) ───────────────────────
 
diff --git a/src/lib/oauth/constants/oauth.ts b/src/lib/oauth/constants/oauth.ts
index b9c36adc26b..b5012102fc9 100644
--- a/src/lib/oauth/constants/oauth.ts
+++ b/src/lib/oauth/constants/oauth.ts
@@ -74,8 +74,10 @@ export const CODEX_CONFIG = {
 // Qwen OAuth Configuration (Device Code Flow with PKCE)
 export const QWEN_CONFIG = {
   clientId: resolvePublicCred("qwen_id", "QWEN_OAUTH_CLIENT_ID"),
-  deviceCodeUrl: "https://qwen.ai/api/v1/oauth2/device/code",
-  tokenUrl: "https://qwen.ai/api/v1/oauth2/token",
+  // Host is chat.qwen.ai — the bare qwen.ai host 404s on these paths (verified: the qwen-code
+  // device flow lives at chat.qwen.ai, returning a valid device_code; qwen.ai returns Not Found).
+  deviceCodeUrl: "https://chat.qwen.ai/api/v1/oauth2/device/code",
+  tokenUrl: "https://chat.qwen.ai/api/v1/oauth2/token",
   scope: "openid profile email model.completion",
   codeChallengeMethod: "S256",
 };
diff --git a/src/lib/oauth/providers/codebuddy-cn.ts b/src/lib/oauth/providers/codebuddy-cn.ts
index 3a4fa935441..f24a4de0dba 100644
--- a/src/lib/oauth/providers/codebuddy-cn.ts
+++ b/src/lib/oauth/providers/codebuddy-cn.ts
@@ -38,7 +38,10 @@ export const codebuddyCn = {
   flowType: "device_code" as const,
 
   requestDeviceCode: async (config: CodeBuddyConfig): Promise<CodeBuddyDeviceCodeResponse> => {
-    const response = await fetch(config.stateUrl, {
+    // CodeBuddy reads `platform` from the QUERY string, not the JSON body — sending it only in the
+    // body returns 400 "platform is empty" (verified). Pass it as a query param; body kept as-is.
+    const stateUrl = `${config.stateUrl}?platform=${encodeURIComponent(config.platform)}`;
+    const response = await fetch(stateUrl, {
       method: "POST",
       headers: {
         "Content-Type": "application/json",
diff --git a/src/lib/providers/validation/webProvidersB.ts b/src/lib/providers/validation/webProvidersB.ts
index 3d9ddd84d4c..6f1069f9ff8 100644
--- a/src/lib/providers/validation/webProvidersB.ts
+++ b/src/lib/providers/validation/webProvidersB.ts
@@ -292,7 +292,7 @@ export async function validateCopilotWebProvider({ apiKey, providerSpecificData
   }
 }
 
-function extractM365CredentialParts(raw: string, providerSpecificData: Record<string, unknown>) {
+export function extractM365CredentialParts(raw: string, providerSpecificData: Record<string, unknown>) {
   const text = raw.trim();
   const parts: Record<string, string> = {};
 
@@ -304,13 +304,22 @@ function extractM365CredentialParts(raw: string, providerSpecificData: Record<st
     if (key && value) parts[key] = value;
   }
 
-  if (/^wss:\/\/substrate\.office\.com\/m365Copilot\/Chathub\//i.test(text)) {
+  // Accept the current M365 web endpoint (m365.cloud.microsoft, including
+  // regional subdomains) plus the two legacy hosts (substrate.office.com,
+  // copilot.microsoft.com). The path still carries /m365Copilot/Chathub/<tenant>,
+  // so extraction is unchanged. (OmniRoute issue #7078)
+  if (/^wss:\/\//i.test(text)) {
     try {
       const url = new URL(text);
-      parts.access_token ||= url.searchParams.get("access_token") || "";
-      parts.chathubPath ||= decodeURIComponent(
-        url.pathname.split("/m365Copilot/Chathub/")[1] || ""
+      const hostOk = /^(?:[\w-]+\.)*(?:m365\.cloud\.microsoft|copilot\.microsoft\.com|substrate\.office\.com)$/i.test(
+        url.hostname
       );
+      if (hostOk && url.pathname.startsWith("/m365Copilot/Chathub/")) {
+        parts.access_token ||= url.searchParams.get("access_token") || "";
+        parts.chathubPath ||= decodeURIComponent(
+          url.pathname.split("/m365Copilot/Chathub/")[1] || ""
+        );
+      }
     } catch {
       // Fall through to the structured key/value parser result.
     }
diff --git a/src/lib/quota/connectionRecovery.ts b/src/lib/quota/connectionRecovery.ts
index 1e0e8d30343..46d8ee46fb5 100644
--- a/src/lib/quota/connectionRecovery.ts
+++ b/src/lib/quota/connectionRecovery.ts
@@ -22,6 +22,7 @@
  */
 
 import { cooldownUntilMs } from "@omniroute/open-sse/services/accountFallback.ts";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 
 /**
  * The transient-cooldown status written by `markAccountUnavailable()` for a
@@ -223,14 +224,6 @@ function isBuildProcess(): boolean {
   return typeof process !== "undefined" && process.env.NEXT_PHASE === "phase-production-build";
 }
 
-function isAutomatedTestProcess(): boolean {
-  return (
-    typeof process !== "undefined" &&
-    (process.env.NODE_ENV === "test" ||
-      process.env.VITEST !== undefined ||
-      process.argv.some((arg) => arg.includes("test")))
-  );
-}
 
 function isRecoverySchedulerDisabled(): boolean {
   return (
diff --git a/src/lib/resilience/modelLockoutSettings.ts b/src/lib/resilience/modelLockoutSettings.ts
index 39e961e724e..205863288aa 100644
--- a/src/lib/resilience/modelLockoutSettings.ts
+++ b/src/lib/resilience/modelLockoutSettings.ts
@@ -1,3 +1,5 @@
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
+
 export interface ModelLockoutSettings {
   enabled: boolean;
   errorCodes: number[];
@@ -56,10 +58,7 @@ export function resolveModelLockoutSettings(
 ): ModelLockoutSettings {
   const record = asRecord(settings);
   const raw = asRecord(record.modelLockout);
-  const isTest =
-    process.env.NODE_ENV === "test" ||
-    process.execArgv.includes("--test") ||
-    process.argv.some((arg) => typeof arg === "string" && arg.includes("test"));
+  const isTest = isAutomatedTestProcess();
 
   const baseCooldownMs = toInteger(
     raw.baseCooldownMs,
diff --git a/src/lib/routerEval/index.ts b/src/lib/routerEval/index.ts
new file mode 100644
index 00000000000..9368ae4f78f
--- /dev/null
+++ b/src/lib/routerEval/index.ts
@@ -0,0 +1,428 @@
+type JsonRecord = Record<string, unknown>;
+
+export type RouterObservation = {
+  sampleId: string;
+  routeInput: JsonRecord;
+  configId: string;
+  selectedModel: string | null;
+  expectedModel: string | null;
+  latencyMs: number;
+  costUsd: number;
+  success: boolean;
+  timestamp: string;
+  metadata: JsonRecord;
+};
+
+export type RouterObservationInput = {
+  sampleId?: unknown;
+  id?: unknown;
+  routeInput?: unknown;
+  configId?: unknown;
+  selectedModel?: unknown;
+  model?: unknown;
+  expectedModel?: unknown;
+  requestedModel?: unknown;
+  latencyMs?: unknown;
+  latency?: unknown;
+  durationMs?: unknown;
+  duration?: unknown;
+  costUsd?: unknown;
+  cost?: unknown;
+  success?: unknown;
+  status?: unknown;
+  error?: unknown;
+  timestamp?: unknown;
+  tokens?: {
+    input?: unknown;
+    output?: unknown;
+  };
+  metadata?: unknown;
+};
+
+export type RouterConfigAggregate = {
+  configId: string;
+  samples: number;
+  successRate: number;
+  avgLatencyMs: number;
+  p50LatencyMs: number;
+  p95LatencyMs: number;
+  avgCostUsd: number;
+  aiq: number;
+};
+
+export type RouterEvalReport = {
+  evaluatedAt: string;
+  summary: {
+    totalSamples: number;
+    validSamples: number;
+    droppedSamples: number;
+    uniqueConfigs: number;
+  };
+  configurations: RouterConfigAggregate[];
+  frontier: RouterConfigAggregate[];
+  top: RouterConfigAggregate[];
+};
+
+export type RouterEvalComparison = {
+  baseline: RouterEvalReport;
+  candidate: RouterEvalReport;
+  delta: {
+    aiq: number;
+    costUsd: number;
+  };
+  regressions: string[];
+};
+
+export type RouterEvalArtifact = {
+  schemaVersion: 1;
+  kind: "router-eval-report" | "router-eval-comparison";
+  generatedAt: string;
+  metadata?: RouterEvalArtifactMetadata;
+  report?: RouterEvalReport;
+  comparison?: RouterEvalComparison;
+};
+
+export type RouterEvalArtifactMetadata = {
+  candidate?: {
+    source: string;
+    path?: string;
+    dbSource?: string;
+  };
+  baseline?: {
+    source: string;
+    path?: string;
+    dbSource?: string;
+  };
+  window?: {
+    since?: string;
+    limit?: number;
+  };
+  thresholds?: {
+    maxAiqDrop: number;
+    maxCostIncrease: number;
+  };
+  outputs?: {
+    markdown?: string;
+    json?: string;
+    corpus?: string;
+  };
+};
+
+function asRecord(value: unknown): JsonRecord {
+  return value && typeof value === "object" && !Array.isArray(value) ? (value as JsonRecord) : {};
+}
+
+function asString(value: unknown, fallback = ""): string {
+  return typeof value === "string" && value.trim().length > 0 ? value : fallback;
+}
+
+function asNumber(value: unknown, fallback = 0): number {
+  if (typeof value === "number" && Number.isFinite(value)) return value;
+  if (typeof value === "string") {
+    const parsed = Number.parseFloat(value);
+    return Number.isFinite(parsed) ? parsed : fallback;
+  }
+  return fallback;
+}
+
+function asBoolean(value: unknown, fallback = false): boolean {
+  if (typeof value === "boolean") return value;
+  if (typeof value === "string") {
+    if (value.toLowerCase() === "true") return true;
+    if (value.toLowerCase() === "false") return false;
+  }
+  if (typeof value === "number") return value !== 0;
+  return fallback;
+}
+
+function percentile(sortedValues: number[], p: number): number {
+  if (sortedValues.length === 0) return 0;
+  if (sortedValues.length === 1) return sortedValues[0];
+  const bounded = Math.min(1, Math.max(0, p));
+  const index = Math.floor((sortedValues.length - 1) * bounded);
+  return sortedValues[index] ?? 0;
+}
+
+function computeAiq(successRate: number, avgLatencyMs: number, avgCostUsd: number): number {
+  const latencyPenalty = avgLatencyMs / 1_000;
+  const costPenalty = avgCostUsd * 1_000;
+  return Number((successRate * 100 - latencyPenalty - costPenalty).toFixed(3));
+}
+
+function aggregateValues(values: number[]) {
+  if (values.length === 0) return { avg: 0, p50: 0, p95: 0 };
+  const sorted = [...values].sort((a, b) => a - b);
+  const sum = sorted.reduce((acc, value) => acc + value, 0);
+  return {
+    avg: sum / sorted.length,
+    p50: percentile(sorted, 0.5),
+    p95: percentile(sorted, 0.95),
+  };
+}
+
+function resolveObservationSampleId(value: RouterObservationInput): string {
+  return asString(value.sampleId ?? value.id, "").trim();
+}
+
+function resolveObservationModelFields(value: RouterObservationInput): {
+  selectedModel: string | null;
+  expectedModel: string | null;
+} {
+  const selectedModel = asString(value.selectedModel ?? value.model);
+  const expectedModel = asString(value.expectedModel ?? value.requestedModel, null);
+  return { selectedModel: selectedModel || null, expectedModel };
+}
+
+function resolveObservationLatencyMs(value: RouterObservationInput): number {
+  return asNumber(value.latencyMs ?? value.latency ?? value.durationMs ?? value.duration, 0);
+}
+
+function resolveObservationCostUsd(value: RouterObservationInput): number {
+  const explicitCost = asNumber(value.costUsd ?? value.cost, NaN);
+  if (!Number.isNaN(explicitCost)) return explicitCost;
+
+  const promptTokens = asNumber(value.tokens?.input);
+  const completionTokens = asNumber(value.tokens?.output);
+  return Number(((promptTokens + completionTokens) * 0.000001).toFixed(6));
+}
+
+function resolveObservationSuccess(value: RouterObservationInput): boolean {
+  const status = asNumber(value.status, 500);
+  const statusIndicatesSuccess = status >= 200 && status < 400;
+  return asBoolean(value.success, statusIndicatesSuccess && !value.error);
+}
+
+export function toRouterObservation(input: unknown): RouterObservation | null {
+  if (!input || typeof input !== "object" || Array.isArray(input)) return null;
+
+  const value = input as RouterObservationInput;
+  const sampleId = resolveObservationSampleId(value);
+  if (!sampleId) return null;
+
+  const { selectedModel, expectedModel } = resolveObservationModelFields(value);
+
+  return {
+    sampleId,
+    routeInput: asRecord(value.routeInput, {}),
+    configId: asString(value.configId, "default"),
+    selectedModel,
+    expectedModel,
+    latencyMs: resolveObservationLatencyMs(value),
+    costUsd: resolveObservationCostUsd(value),
+    success: resolveObservationSuccess(value),
+    timestamp: asString(value.timestamp, new Date(0).toISOString()),
+    metadata: asRecord(value.metadata),
+  };
+}
+
+export function summarizeRouterObservations(observations: RouterObservation[]) {
+  const byConfig = new Map<string, RouterObservation[]>();
+  for (const obs of observations) {
+    if (!byConfig.has(obs.configId)) byConfig.set(obs.configId, []);
+    byConfig.get(obs.configId)?.push(obs);
+  }
+
+  const configurations: RouterConfigAggregate[] = [];
+  for (const [configId, rows] of byConfig) {
+    const latencies = rows.map((row) => Math.max(0, row.latencyMs));
+    const costs = rows.map((row) => Math.max(0, row.costUsd));
+    const latencyStats = aggregateValues(latencies);
+    const costStats = aggregateValues(costs);
+    const successCount = rows.filter((row) => row.success).length;
+    const successRate = rows.length === 0 ? 0 : successCount / rows.length;
+    configurations.push({
+      configId,
+      samples: rows.length,
+      successRate,
+      avgLatencyMs: latencyStats.avg,
+      p50LatencyMs: latencyStats.p50,
+      p95LatencyMs: latencyStats.p95,
+      avgCostUsd: costStats.avg,
+      aiq: computeAiq(successRate, latencyStats.avg, costStats.avg),
+    });
+  }
+
+  const top = [...configurations].sort(
+    (a, b) =>
+      b.aiq - a.aiq ||
+      b.successRate - a.successRate ||
+      a.avgLatencyMs - b.avgLatencyMs ||
+      a.avgCostUsd - b.avgCostUsd
+  );
+  const frontier = computeParetoFrontier(configurations);
+
+  return {
+    configurations: configurations.sort((a, b) => b.aiq - a.aiq),
+    frontier,
+    top,
+  };
+}
+
+export function runRouterEval(observations: RouterObservation[]): RouterEvalReport {
+  const valid = observations.filter((value) => Boolean(value.selectedModel));
+  const droppedSamples = observations.length - valid.length;
+  const { configurations, frontier, top } = summarizeRouterObservations(valid);
+
+  return {
+    evaluatedAt: new Date().toISOString(),
+    summary: {
+      totalSamples: observations.length,
+      validSamples: valid.length,
+      droppedSamples,
+      uniqueConfigs: configurations.length,
+    },
+    configurations,
+    frontier,
+    top,
+  };
+}
+
+export function computeParetoFrontier(configs: RouterConfigAggregate[]) {
+  const ordered = [...configs].sort(
+    (a, b) => a.avgCostUsd - b.avgCostUsd || a.avgLatencyMs - b.avgLatencyMs
+  );
+  const frontier: RouterConfigAggregate[] = [];
+
+  for (const config of ordered) {
+    const dominated = frontier.some(
+      (candidate) =>
+        candidate.aiq >= config.aiq &&
+        candidate.avgCostUsd <= config.avgCostUsd &&
+        candidate.avgLatencyMs <= config.avgLatencyMs
+    );
+    if (!dominated) frontier.push(config);
+  }
+
+  return frontier.sort((a, b) => b.aiq - a.aiq || a.avgCostUsd - b.avgCostUsd);
+}
+
+export function compareRouterEvalRuns(
+  baseline: RouterEvalReport,
+  candidate: RouterEvalReport,
+  thresholds: {
+    aiqDrop: number;
+    relativeCostIncrease: number;
+  } = { aiqDrop: 0, relativeCostIncrease: 0 }
+) {
+  const baselineBest = baseline.top[0];
+  const candidateBest = candidate.top[0];
+  const regressions: string[] = [];
+  if (!baselineBest || !candidateBest) {
+    return { baseline, candidate, delta: { aiq: 0, costUsd: 0 }, regressions };
+  }
+
+  const aiq = candidateBest.aiq - baselineBest.aiq;
+  const baselineCost = baselineBest.avgCostUsd;
+  const cost = candidateBest.avgCostUsd;
+  const costIncrease =
+    baselineCost === 0 ? (cost > 0 ? Infinity : 0) : (cost - baselineCost) / baselineCost;
+
+  if (aiq < -Math.abs(thresholds.aiqDrop)) {
+    regressions.push(`AIQ dropped by ${(-aiq).toFixed(3)} below threshold`);
+  }
+
+  if (costIncrease > thresholds.relativeCostIncrease) {
+    regressions.push(`cost increased by ${(costIncrease * 100).toFixed(2)}% above threshold`);
+  }
+
+  return {
+    baseline,
+    candidate,
+    delta: {
+      aiq,
+      costUsd: costIncrease * baselineCost,
+    },
+    regressions,
+  };
+}
+
+export function formatRouterEvalReport(report: RouterEvalReport): string {
+  const best = report.top[0];
+  const lines = [
+    "# Router Eval Report",
+    `Generated: ${report.evaluatedAt}`,
+    `Samples: ${report.summary.validSamples}/${report.summary.totalSamples} valid, ${report.summary.uniqueConfigs} config(s)`,
+    "",
+    "## Top Configurations",
+    "",
+    "| Config | Samples | AIQ | Success | Avg Latency | p50 | p95 | Avg Cost |",
+    "| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |",
+  ];
+
+  for (const config of report.top.slice(0, 10)) {
+    lines.push(
+      `| ${config.configId} | ${config.samples} | ${config.aiq.toFixed(3)} | ${(config.successRate * 100).toFixed(2)}% | ${config.avgLatencyMs.toFixed(2)}ms | ${config.p50LatencyMs.toFixed(2)}ms | ${config.p95LatencyMs.toFixed(2)}ms | $${config.avgCostUsd.toFixed(6)} |`
+    );
+  }
+
+  if (best) {
+    lines.push("");
+    lines.push(`Best: **${best.configId}** (AIQ ${best.aiq.toFixed(3)})`);
+  }
+
+  if (report.frontier.length > 0) {
+    lines.push("");
+    lines.push("## Pareto Frontier");
+    lines.push("");
+    lines.push("| Config | AIQ | Cost | Latency |");
+    lines.push("| --- | ---: | ---: | ---: |");
+    for (const config of report.frontier) {
+      lines.push(
+        `| ${config.configId} | ${config.aiq.toFixed(3)} | $${config.avgCostUsd.toFixed(6)} | ${config.avgLatencyMs.toFixed(2)}ms |`
+      );
+    }
+  }
+
+  return `${lines.join("\n")}\n`;
+}
+
+export function formatRouterEvalComparison(comparison: RouterEvalComparison): string {
+  const bestCandidate = comparison.candidate.top[0];
+  const bestBaseline = comparison.baseline.top[0];
+  const passes = comparison.regressions.length === 0;
+  const lines = [
+    "# Router Eval Comparison",
+    `Passed: ${passes ? "✅" : "❌"}`,
+    `AIQ delta: ${comparison.delta.aiq.toFixed(3)}`,
+    `Cost delta: $${comparison.delta.costUsd.toFixed(6)}`,
+    "",
+  ];
+
+  if (bestBaseline && bestCandidate) {
+    lines.push(`Baseline best: ${bestBaseline.configId} (AIQ ${bestBaseline.aiq.toFixed(3)})`);
+    lines.push(`Candidate best: ${bestCandidate.configId} (AIQ ${bestCandidate.aiq.toFixed(3)})`);
+    lines.push("");
+  }
+
+  if (comparison.regressions.length > 0) {
+    lines.push("## Regressions");
+    for (const reason of comparison.regressions) {
+      lines.push(`- ${reason}`);
+    }
+  }
+
+  return `${lines.join("\n")}\n`;
+}
+
+export function createRouterEvalArtifact(
+  value: RouterEvalReport | RouterEvalComparison,
+  metadata?: RouterEvalArtifactMetadata
+): RouterEvalArtifact {
+  if ("candidate" in value && "baseline" in value) {
+    return {
+      schemaVersion: 1,
+      kind: "router-eval-comparison",
+      generatedAt: value.candidate.evaluatedAt,
+      metadata,
+      comparison: value,
+    };
+  }
+
+  return {
+    schemaVersion: 1,
+    kind: "router-eval-report",
+    generatedAt: value.evaluatedAt,
+    metadata,
+    report: value,
+  };
+}
diff --git a/src/lib/tokenHealthCheck.ts b/src/lib/tokenHealthCheck.ts
index b21b91005c2..5593edfdba6 100644
--- a/src/lib/tokenHealthCheck.ts
+++ b/src/lib/tokenHealthCheck.ts
@@ -25,6 +25,7 @@ import {
   refreshCopilotToken,
 } from "@omniroute/open-sse/services/tokenRefresh.ts";
 import { pickMaskedDisplayValue } from "@/shared/utils/maskEmail";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
 
 // ── Constants ────────────────────────────────────────────────────────────────
 const TICK_MS = 60 * 1000; // sweep interval: every 60 seconds
@@ -38,15 +39,6 @@ function isBuildProcess(): boolean {
   return typeof process !== "undefined" && process.env.NEXT_PHASE === "phase-production-build";
 }
 
-function isAutomatedTestProcess(): boolean {
-  return (
-    typeof process !== "undefined" &&
-    (process.env.NODE_ENV === "test" ||
-      process.env.VITEST !== undefined ||
-      process.argv.some((arg) => arg.includes("test")))
-  );
-}
-
 function getConnectionLogLabel(conn: { name?: string; email?: string; id?: string }): string {
   return pickMaskedDisplayValue([conn.name, conn.email], conn.id || "-");
 }
@@ -185,13 +177,10 @@ function isHealthCheckDisabled(): boolean {
 }
 
 /**
- * Providers excluded from the PROACTIVE refresh sweep, comma-separated and
- * case-insensitive (e.g. "codex,openai"). A targeted alternative to the blunt
- * OMNIROUTE_DISABLE_TOKEN_HEALTHCHECK switch: it lets an operator keep the
- * rotating-token cascade providers (Codex/OpenAI share one Auth0 family) off the
- * proactive sweep — leaving their refresh to the reactive, serialized 401 path —
- * WITHOUT also starving short-TTL providers like Kimi-coding, whose tokens expire
- * while idle when the whole sweep is disabled.
+ * Providers excluded from the PROACTIVE sweep, comma-separated, case-insensitive
+ * (e.g. "codex,openai"). Targeted alternative to OMNIROUTE_DISABLE_TOKEN_HEALTHCHECK:
+ * keeps rotating-token cascade providers (Codex/OpenAI share one Auth0 family) on the
+ * reactive 401 path WITHOUT starving short-TTL providers (Kimi-coding) sweep-wide.
  */
 function getHealthCheckSkipProviders(): Set<string> {
   const raw = process.env.OMNIROUTE_HEALTHCHECK_SKIP_PROVIDERS || "";
@@ -277,12 +266,12 @@ export function clearHealthCheckLogCache() {
 
 declare global {
   var __omnirouteTokenHC:
-    { initialized: boolean; interval: ReturnType<typeof setInterval> | null } | undefined;
+    | { initialized: boolean; interval: ReturnType<typeof setInterval> | null; sweeping: boolean }
+    | undefined;
 }
-
 function getHCState() {
   if (!globalThis.__omnirouteTokenHC) {
-    globalThis.__omnirouteTokenHC = { initialized: false, interval: null };
+    globalThis.__omnirouteTokenHC = { initialized: false, interval: null, sweeping: false };
   }
   return globalThis.__omnirouteTokenHC;
 }
@@ -322,7 +311,12 @@ export function stopTokenHealthCheck() {
 }
 
 // ── Core sweep ───────────────────────────────────────────────────────────────
-async function sweep() {
+export async function sweep() {
+  const state = getHCState();
+  if (state.sweeping) {
+    return log(`${LOG_PREFIX} Sweep skipped — previous sweep still in progress`);
+  }
+  state.sweeping = true;
   try {
     const connections = await getProviderConnections({ authType: "oauth" });
 
@@ -339,13 +333,18 @@ async function sweep() {
         logError(`${LOG_PREFIX} Error checking ${conn.name || conn.id}:`, err.message);
       }
 
-      // Stagger delay between checks to prevent bursting (Issue #1220)
+      // Stagger + randomized jitter between checks to prevent bursting (Issue #1220)
       if (staggerMs > 0 && i < connections.length - 1) {
-        await new Promise((resolve) => setTimeout(resolve, staggerMs));
+        const jitterMin = parseInt(process.env.HEALTHCHECK_JITTER_MIN_MS || "500", 10);
+        const jitterMax = parseInt(process.env.HEALTHCHECK_JITTER_MAX_MS || "5000", 10);
+        const jitter = jitterMin + Math.random() * Math.max(0, jitterMax - jitterMin);
+        await new Promise((resolve) => setTimeout(resolve, staggerMs + jitter));
       }
     }
   } catch (err) {
     logError(`${LOG_PREFIX} Sweep error:`, err.message);
+  } finally {
+    state.sweeping = false;
   }
 }
 
diff --git a/src/lib/usage/codexResetCredits.ts b/src/lib/usage/codexResetCredits.ts
index 1e770270c04..ca99d72c29c 100644
--- a/src/lib/usage/codexResetCredits.ts
+++ b/src/lib/usage/codexResetCredits.ts
@@ -26,6 +26,25 @@ type CodexConnectionLike = JsonRecord & {
 
 export type CodexResetCreditOutcome = "reset" | "alreadyRedeemed";
 
+export interface CodexResetCredit {
+  id: string;
+  resetType?: string;
+  status?: string;
+  grantedAt?: string;
+  expiresAt?: string | null;
+  title?: string;
+  description?: string;
+}
+
+export type PublicCodexResetCredit = Omit<CodexResetCredit, "id"> & {
+  selectionToken: string;
+};
+
+export interface CodexResetCreditList {
+  credits: PublicCodexResetCredit[];
+  availableCount: number;
+}
+
 export class CodexResetCreditError extends Error {
   status: number;
   code: string;
@@ -116,16 +135,56 @@ function isUnavailableResetCredit(record: JsonRecord): boolean {
   const status = normalizeOutcome(
     record.status ?? record.state ?? record.outcome ?? record.result ?? record.code
   );
-  if (status && ["consumed", "redeemed", "used", "expired", "unavailable"].includes(status)) {
+  if (
+    status &&
+    ["consumed", "redeeming", "redeemed", "used", "expired", "unavailable"].includes(status)
+  ) {
     return true;
   }
   return record.consumed === true || record.redeemed === true || record.available === false;
 }
 
-function extractCreditIdFromRecord(value: unknown): string | null {
+function parseOptionalTimestamp(record: JsonRecord, keys: string[]): string | null | undefined {
+  for (const key of keys) {
+    const value = record[key];
+    if (value === null) return null;
+    if (typeof value === "string" && value.trim().length > 0) return value.trim();
+  }
+  return undefined;
+}
+
+function parseResetCredit(value: unknown): CodexResetCredit | null {
   const record = toRecord(value);
   if (!record || Object.keys(record).length === 0 || isUnavailableResetCredit(record)) return null;
-  return extractStringField(record, ["credit_id", "creditId", "id"]);
+  const id = extractStringField(record, ["credit_id", "creditId", "id"]);
+  if (!id) return null;
+
+  const expiresAt = parseOptionalTimestamp(record, [
+    "expires_at",
+    "expiresAt",
+    "expiration_at",
+    "expirationAt",
+  ]);
+  if (expiresAt) {
+    const expiresAtMs = Date.parse(expiresAt);
+    if (Number.isFinite(expiresAtMs) && expiresAtMs <= Date.now()) return null;
+  }
+
+  const resetType = extractStringField(record, ["reset_type", "resetType"]);
+  const status = extractStringField(record, ["status", "state"]);
+  const grantedAt = parseOptionalTimestamp(record, ["granted_at", "grantedAt"]);
+  const title = extractStringField(record, ["title"]);
+  const description = extractStringField(record, ["description"]);
+
+  return {
+    id,
+    ...(resetType ? { resetType } : {}),
+    ...(status ? { status } : {}),
+    ...(grantedAt !== undefined ? { grantedAt } : {}),
+    ...(expiresAt !== undefined ? { expiresAt } : {}),
+    ...(title ? { title } : {}),
+    ...(description ? { description } : {}),
+  };
 }
 
 function getResetCreditCandidates(payload: unknown): unknown[] {
@@ -148,12 +207,55 @@ function getResetCreditCandidates(payload: unknown): unknown[] {
   return [];
 }
 
-function parseAvailableResetCreditId(payload: unknown): string {
-  for (const candidate of getResetCreditCandidates(payload)) {
-    const creditId = extractCreditIdFromRecord(candidate);
-    if (creditId) return creditId;
+function getExpirySortValue(credit: CodexResetCredit): number {
+  if (!credit.expiresAt) return Number.POSITIVE_INFINITY;
+  const expiresAtMs = Date.parse(credit.expiresAt);
+  return Number.isFinite(expiresAtMs) ? expiresAtMs : Number.POSITIVE_INFINITY;
+}
+
+function parseAvailableResetCredits(payload: unknown): {
+  credits: CodexResetCredit[];
+  availableCount: number;
+} {
+  const credits = getResetCreditCandidates(payload)
+    .map(parseResetCredit)
+    .filter((credit): credit is CodexResetCredit => credit !== null)
+    .map((credit, index) => ({ credit, index }))
+    .sort(
+      (a, b) => getExpirySortValue(a.credit) - getExpirySortValue(b.credit) || a.index - b.index
+    )
+    .map(({ credit }) => credit);
+
+  const payloadRecord = toRecord(payload);
+  const reportedCount = Number(payloadRecord.available_count ?? payloadRecord.availableCount);
+
+  return {
+    credits,
+    availableCount: Number.isFinite(reportedCount)
+      ? Math.max(0, Math.trunc(reportedCount))
+      : credits.length,
+  };
+}
+
+function toPublicResetCredit(credit: CodexResetCredit): PublicCodexResetCredit {
+  const { id, ...metadata } = credit;
+  return { ...metadata, selectionToken: id };
+}
+
+function selectResetCredit(payload: unknown, requestedCreditId?: string): CodexResetCredit {
+  const { credits } = parseAvailableResetCredits(payload);
+  if (requestedCreditId) {
+    const selected = credits.find((credit) => credit.id === requestedCreditId);
+    if (selected) return selected;
+    throw new CodexResetCreditError(
+      409,
+      "selected_credit_unavailable",
+      "The selected Codex reset credit is no longer available."
+    );
   }
 
+  if (credits[0]) return credits[0];
+
   throw new CodexResetCreditError(409, "no_credit", "No Codex reset credits are available.");
 }
 
@@ -266,7 +368,8 @@ async function fetchResetCredits(connection: CodexConnectionLike): Promise<Respo
 
 async function consumeWithAuthRetry(
   connection: CodexConnectionLike,
-  idempotencyKey: string
+  idempotencyKey: string,
+  requestedCreditId?: string
 ): Promise<{ connection: CodexConnectionLike; response: Response }> {
   let refreshedConnection = await refreshCodexConnectionIfNeeded(connection);
   let creditsResponse = await fetchResetCredits(refreshedConnection);
@@ -286,8 +389,8 @@ async function consumeWithAuthRetry(
     );
   }
 
-  const creditId = parseAvailableResetCreditId(creditsPayload);
-  let response = await postConsumeResetCredit(refreshedConnection, idempotencyKey, creditId);
+  const credit = selectResetCredit(creditsPayload, requestedCreditId);
+  let response = await postConsumeResetCredit(refreshedConnection, idempotencyKey, credit.id);
 
   if (response.status === 401 || response.status === 403) {
     refreshedConnection = await refreshCodexConnectionIfNeeded(refreshedConnection, true);
@@ -301,16 +404,61 @@ async function consumeWithAuthRetry(
         `Codex reset-credit API returned HTTP ${refreshedCreditsResponse.status}.`
       );
     }
-    const refreshedCreditId = parseAvailableResetCreditId(refreshedCreditsPayload);
-    response = await postConsumeResetCredit(refreshedConnection, idempotencyKey, refreshedCreditId);
+    const refreshedCredit = selectResetCredit(refreshedCreditsPayload, requestedCreditId);
+    response = await postConsumeResetCredit(
+      refreshedConnection,
+      idempotencyKey,
+      refreshedCredit.id
+    );
   }
 
   return { connection: refreshedConnection, response };
 }
 
+export async function listCodexResetCredits(connectionId: string): Promise<CodexResetCreditList> {
+  if (!connectionId || typeof connectionId !== "string") {
+    throw new CodexResetCreditError(400, "connection_id_required", "connectionId is required.");
+  }
+
+  try {
+    let connection = await loadCodexConnection(connectionId);
+    connection = await refreshCodexConnectionIfNeeded(connection);
+    let response = await fetchResetCredits(connection);
+
+    if (response.status === 401 || response.status === 403) {
+      connection = await refreshCodexConnectionIfNeeded(connection, true);
+      response = await fetchResetCredits(connection);
+    }
+
+    const payload = await readResponsePayload(response);
+    if (!response.ok) {
+      throwKnownConsumeError(payload);
+      throw new CodexResetCreditError(
+        response.status,
+        "codex_reset_credit_upstream_error",
+        `Codex reset-credit API returned HTTP ${response.status}.`
+      );
+    }
+
+    const result = parseAvailableResetCredits(payload);
+    return {
+      availableCount: result.availableCount,
+      credits: result.credits.map(toPublicResetCredit),
+    };
+  } catch (error) {
+    if (error instanceof CodexResetCreditError) throw error;
+    throw new CodexResetCreditError(
+      500,
+      "codex_reset_credit_list_failed",
+      sanitizeErrorMessage(error) || "Failed to load Codex reset credits."
+    );
+  }
+}
+
 export async function consumeCodexResetCredit(
   connectionId: string,
-  idempotencyKey: string
+  idempotencyKey: string,
+  creditId?: string
 ): Promise<{
   outcome: CodexResetCreditOutcome;
   usage: JsonRecord;
@@ -324,7 +472,13 @@ export async function consumeCodexResetCredit(
 
   try {
     const connection = await loadCodexConnection(connectionId);
-    const { response } = await consumeWithAuthRetry(connection, idempotencyKey.trim());
+    const normalizedCreditId =
+      typeof creditId === "string" && creditId.trim().length > 0 ? creditId.trim() : undefined;
+    const { response } = await consumeWithAuthRetry(
+      connection,
+      idempotencyKey.trim(),
+      normalizedCreditId
+    );
     const payload = await readResponsePayload(response);
 
     if (!response.ok) {
diff --git a/src/lib/usage/comboScoringInspector.ts b/src/lib/usage/comboScoringInspector.ts
index c61b3f7d122..9f6c9eb0b3f 100644
--- a/src/lib/usage/comboScoringInspector.ts
+++ b/src/lib/usage/comboScoringInspector.ts
@@ -1,4 +1,5 @@
 import { buildComboHealthAutopilotReport } from "@/lib/monitoring/comboHealthAutopilot";
+import { getCombos } from "@/lib/db/combos";
 import { getProviderConnections } from "@/lib/db/providers";
 import { buildComboForecastResponse } from "@/lib/usage/comboForecast";
 import { buildComboHealthResponse } from "@/lib/usage/comboHealth";
@@ -14,7 +15,9 @@ import {
   type ProviderCandidate,
   type ScoringFactors,
   type ScoringWeights,
+  validateWeights,
 } from "@omniroute/open-sse/services/autoCombo/scoring.ts";
+import { getModePack } from "@omniroute/open-sse/services/autoCombo/modePacks.ts";
 import { getTaskFitness } from "@omniroute/open-sse/services/autoCombo/taskFitness.ts";
 import type {
   ComboAutopilotCombo,
@@ -32,6 +35,7 @@ import type {
   ComboScoringInspectorResponse,
   ComboScoringInspectorSource,
   ComboScoringInspectorTarget,
+  ComboScoringInspectorWeightSource,
   UtilizationTimeRange,
 } from "@/shared/types/utilization";
 
@@ -58,6 +62,13 @@ type CandidateContext = {
   notes: Partial<Record<ComboScoringInspectorFactorKey, string>>;
 };
 
+type InspectorWeights = {
+  weights: ScoringWeights;
+  source: ComboScoringInspectorWeightSource;
+  modePack: string | null;
+  warning?: string;
+};
+
 const FACTOR_KEYS: ComboScoringInspectorFactorKey[] = [
   "quota",
   "health",
@@ -81,6 +92,82 @@ function normalizeTaskType(taskType: string | undefined): string {
   return typeof taskType === "string" && taskType.trim().length > 0 ? taskType.trim() : "default";
 }
 
+function isRecord(value: unknown): value is Record<string, unknown> {
+  return value !== null && typeof value === "object" && !Array.isArray(value);
+}
+
+/**
+ * Resolves the auto-strategy config object a combo's weights should be read from,
+ * honoring the same precedence the runtime auto-combo strategy uses: a dedicated
+ * `autoConfig`, then a nested `config.auto`, then the raw `config`, then an empty
+ * fallback so callers can probe optional fields without further null-checks.
+ */
+function resolveInspectorAutoConfig(combo: ComboRecord | undefined): Record<string, unknown> {
+  if (isRecord(combo?.autoConfig)) return combo.autoConfig;
+  if (isRecord(combo?.config)) {
+    if (isRecord(combo.config.auto)) return combo.config.auto;
+    return combo.config;
+  }
+  return {};
+}
+
+/** Extracts the mode-pack name referenced by the config, if any. */
+function resolveModePackName(config: Record<string, unknown>): string | null {
+  return typeof config.modePack === "string" ? config.modePack : null;
+}
+
+/** Resolves an explicit, validated `weights` object from the config, if present. */
+function resolveExplicitWeights(config: Record<string, unknown>): ScoringWeights | undefined {
+  const explicitWeights = isRecord(config.weights) ? (config.weights as ScoringWeights) : undefined;
+  return explicitWeights && validateWeights(explicitWeights) ? explicitWeights : undefined;
+}
+
+function resolveInspectorWeights(combo: ComboRecord | undefined): InspectorWeights {
+  const config = resolveInspectorAutoConfig(combo);
+
+  const modePack = resolveModePackName(config);
+  const resolvedModePack = modePack ? getModePack(modePack) : undefined;
+  if (resolvedModePack) {
+    return { weights: resolvedModePack, source: "mode_pack", modePack };
+  }
+
+  const explicitWeights = resolveExplicitWeights(config);
+  if (explicitWeights) {
+    return { weights: explicitWeights, source: "explicit", modePack: null };
+  }
+
+  return {
+    weights: DEFAULT_WEIGHTS,
+    source: "default",
+    modePack: null,
+    warning:
+      modePack || isRecord(config.weights)
+        ? "Configured auto weights are invalid or unavailable; default weights were used."
+        : undefined,
+  };
+}
+
+async function resolveConfiguredCombos(
+  options: ComboScoringInspectorOptions
+): Promise<ComboRecord[]> {
+  // #7087: `options.combos`, when supplied, must always win — regardless of the
+  // health/forecast/autopilot short-circuit below. The dashboard integration
+  // (comboHealthDashboard.ts::buildComboHealthDashboardResponse) always calls this
+  // with `combos` AND `healthResponse`/`forecastResponse`/`autopilotReport` set
+  // together (it already fetched everything once upstream to avoid redundant work),
+  // which used to hit the short-circuit FIRST and silently discard the caller's
+  // `combos` — so `combosById`/`combosByName` (used by resolveInspectorWeights() to
+  // report the combo's real configured modePack/weights) came back empty and every
+  // combo's weightSource fell back to "default" through that call path.
+  if (options.combos) return options.combos;
+  const alreadyHaveHealthSignals =
+    options.healthResponse &&
+    options.forecastResponse &&
+    (options.skipAutopilot || options.autopilotReport);
+  if (alreadyHaveHealthSignals) return [];
+  return (await getCombos()) as ComboRecord[];
+}
+
 function buildEmptyAutopilotReport(options: ComboScoringInspectorOptions): ComboAutopilotReport {
   return {
     status: "healthy",
@@ -260,9 +347,11 @@ async function buildInspectorCombo(
   forecastTargets: Map<string, ComboForecastTarget>,
   autopilotCombo: ComboAutopilotCombo | undefined,
   taskType: string,
-  weights: ScoringWeights
+  inspectorWeights: InspectorWeights
 ): Promise<ComboScoringInspectorCombo> {
   const warnings: string[] = [];
+  const { weights } = inspectorWeights;
+  if (inspectorWeights.warning) warnings.push(inspectorWeights.warning);
   const targets = combo.targetHealth ?? [];
   if (combo.strategy !== "auto") {
     warnings.push(
@@ -349,6 +438,8 @@ async function buildInspectorCombo(
     strategy: combo.strategy,
     taskType,
     weights: weights as Record<ComboScoringInspectorFactorKey, number>,
+    weightSource: inspectorWeights.source,
+    modePack: inspectorWeights.modePack,
     selectedExecutionKey: inspectorTargets[0]?.executionKey ?? null,
     targets: inspectorTargets,
     warnings,
@@ -359,14 +450,14 @@ export async function buildComboScoringInspectorResponse(
   options: ComboScoringInspectorOptions
 ): Promise<ComboScoringInspectorResponse> {
   const taskType = normalizeTaskType(options.taskType);
-  const weights = DEFAULT_WEIGHTS;
+  const configuredCombos = await resolveConfiguredCombos(options);
   const [health, forecast] = await Promise.all([
     options.healthResponse ??
       buildComboHealthResponse({
         range: options.range,
         comboId: options.comboId,
         now: options.now,
-        combos: options.combos,
+        combos: configuredCombos,
       }),
     options.forecastResponse ??
       buildComboForecastResponse({
@@ -374,7 +465,7 @@ export async function buildComboScoringInspectorResponse(
         horizon: options.horizon,
         comboId: options.comboId,
         now: options.now,
-        combos: options.combos,
+        combos: configuredCombos,
       }),
   ]);
   const autopilot =
@@ -388,13 +479,19 @@ export async function buildComboScoringInspectorResponse(
           includeHealthy: true,
           includeActions: false,
           now: options.now,
-          combos: options.combos,
+          combos: configuredCombos,
           healthResponse: health,
           forecastResponse: forecast,
         }));
 
   const forecastByComboId = new Map(forecast.combos.map((combo) => [combo.comboId, combo]));
   const autopilotByComboId = new Map(autopilot.combos.map((combo) => [combo.comboId, combo]));
+  const combosById = new Map(
+    configuredCombos.flatMap((combo) => (combo.id ? [[combo.id, combo]] : []))
+  );
+  const combosByName = new Map(
+    configuredCombos.flatMap((combo) => (combo.name ? [[combo.name, combo]] : []))
+  );
 
   return {
     asOf: new Date(options.now ?? Date.now()).toISOString(),
@@ -408,7 +505,9 @@ export async function buildComboScoringInspectorResponse(
           targetForecastMap(forecastByComboId.get(combo.comboId)?.targets ?? []),
           autopilotByComboId.get(combo.comboId),
           taskType,
-          weights
+          resolveInspectorWeights(
+            combosById.get(combo.comboId) ?? combosByName.get(combo.comboName)
+          )
         )
       )
     ),
diff --git a/src/mitm/dns/provision.ts b/src/mitm/dns/provision.ts
index 544207dbb8c..8679e37ffb4 100644
--- a/src/mitm/dns/provision.ts
+++ b/src/mitm/dns/provision.ts
@@ -3,7 +3,8 @@
  * is guarded and unit-testable without spawning the MITM server (#6127 / #6198).
  */
 
-import { addDNSEntry, addDNSEntries } from "./dnsConfig.ts";
+import { addDNSEntry, addDNSEntries, isSudoAvailable } from "./dnsConfig.ts";
+import { isRoot } from "../systemCommands.ts";
 import { ALL_TARGETS } from "../targets/index.ts";
 import { getAllAgentBridgeStates } from "@/lib/db/agentBridgeState.ts";
 import { listCustomHosts } from "@/lib/db/inspectorCustomHosts.ts";
@@ -23,45 +24,39 @@ export interface DnsProvisionDeps {
   addHostsDns?: (hosts: string[], sudoPassword: string) => Promise<void>;
   getAgentStates?: () => ReturnType<typeof getAllAgentBridgeStates>;
   listEnabledCustomHosts?: () => ReturnType<typeof listCustomHosts>;
+  /** Return true if privileged host-file writes are possible (sudo or root). */
+  canElevate?: () => boolean;
   logger?: DnsProvisionLogger;
 }
 
-/**
- * Provision every AgentBridge DNS entry (Antigravity defaults + agents with
- * `dns_enabled=true` + enabled custom hosts). **Every step is best-effort**: a failure
- * is logged with the full `err` — which carries the privileged command's stderr
- * (`systemCommands.ts` folds stderr into the Error message) — and never aborts the
- * bridge start.
- *
- * Previously the default step (`addDNSEntry`) was called unguarded while the two
- * sibling steps and cert install were wrapped, so in containers/headless (Docker
- * `USER node`, no `sudo`, read-only /etc/hosts) it threw out of `startMitmInternal`
- * and killed the whole start (#6127); its stderr also never reached app.log — only a
- * bare exit code hit the toast (#6198). Extracting + guarding all three steps here
- * restores the symmetry and makes the behavior unit-testable without spawning the
- * MITM server.
- */
-export async function provisionDnsEntries(
+/** Fully-resolved dependency set used by the per-step provisioning helpers below. */
+type ResolvedDnsProvisionDeps = {
+  addDefaultDns: (sudoPassword: string) => Promise<void>;
+  addHostsDns: (hosts: string[], sudoPassword: string) => Promise<void>;
+  getAgentStates: () => ReturnType<typeof getAllAgentBridgeStates>;
+  listEnabledCustomHosts: () => ReturnType<typeof listCustomHosts>;
+  logger: DnsProvisionLogger;
+};
+
+/** Antigravity default hosts — best-effort, never throws. */
+async function provisionDefaultDns(
   sudoPassword: string,
-  deps: DnsProvisionDeps = {}
+  deps: ResolvedDnsProvisionDeps
 ): Promise<void> {
-  const addDefaultDns = deps.addDefaultDns ?? addDNSEntry;
-  const addHostsDns = deps.addHostsDns ?? addDNSEntries;
-  const getAgentStates = deps.getAgentStates ?? getAllAgentBridgeStates;
-  const listEnabledCustomHosts =
-    deps.listEnabledCustomHosts ?? (() => listCustomHosts({ enabledOnly: true }));
-  const logger = deps.logger ?? defaultLog;
-
-  // Antigravity default hosts.
   try {
-    await addDefaultDns(sudoPassword);
+    await deps.addDefaultDns(sudoPassword);
   } catch (err) {
-    logger.error({ err }, "Failed to add default DNS entries (continuing)");
+    deps.logger.error({ err }, "Failed to add default DNS entries (continuing)");
   }
+}
 
-  // Collect hosts from agents that have dns_enabled=true in the DB.
+/** Hosts for agents with `dns_enabled=true` in the DB — best-effort, never throws. */
+async function provisionAgentDns(
+  sudoPassword: string,
+  deps: ResolvedDnsProvisionDeps
+): Promise<void> {
   try {
-    const agentStates = getAgentStates();
+    const agentStates = deps.getAgentStates();
     const agentHostsToAdd: string[] = [];
     for (const state of agentStates) {
       if (!state.dns_enabled) continue;
@@ -71,22 +66,79 @@ export async function provisionDnsEntries(
       }
     }
     if (agentHostsToAdd.length > 0) {
-      logger.info({ count: agentHostsToAdd.length }, "Adding DNS for agent host(s)...");
-      await addHostsDns(agentHostsToAdd, sudoPassword);
+      deps.logger.info({ count: agentHostsToAdd.length }, "Adding DNS for agent host(s)...");
+      await deps.addHostsDns(agentHostsToAdd, sudoPassword);
     }
   } catch (err) {
-    logger.error({ err }, "Failed to add agent DNS entries (continuing)");
+    deps.logger.error({ err }, "Failed to add agent DNS entries (continuing)");
   }
+}
 
-  // Collect enabled custom hosts.
+/** Enabled custom hosts — best-effort, never throws. */
+async function provisionCustomHostsDns(
+  sudoPassword: string,
+  deps: ResolvedDnsProvisionDeps
+): Promise<void> {
   try {
-    const customHosts = listEnabledCustomHosts();
+    const customHosts = deps.listEnabledCustomHosts();
     const customHostNames = customHosts.map((h) => h.host);
     if (customHostNames.length > 0) {
-      logger.info({ count: customHostNames.length }, "Adding DNS for custom host(s)...");
-      await addHostsDns(customHostNames, sudoPassword);
+      deps.logger.info({ count: customHostNames.length }, "Adding DNS for custom host(s)...");
+      await deps.addHostsDns(customHostNames, sudoPassword);
     }
   } catch (err) {
-    logger.error({ err }, "Failed to add custom host DNS entries (continuing)");
+    deps.logger.error({ err }, "Failed to add custom host DNS entries (continuing)");
+  }
+}
+
+/**
+ * Provision every AgentBridge DNS entry (Antigravity defaults + agents with
+ * `dns_enabled=true` + enabled custom hosts). **Every step is best-effort**: a failure
+ * is logged with the full `err` — which carries the privileged command's stderr
+ * (`systemCommands.ts` folds stderr into the Error message) — and never aborts the
+ * bridge start.
+ *
+ * Previously the default step (`addDNSEntry`) was called unguarded while the two
+ * sibling steps and cert install were wrapped, so in containers/headless (Docker
+ * `USER node`, no `sudo`, read-only /etc/hosts) it threw out of `startMitmInternal`
+ * and killed the whole start (#6127); its stderr also never reached app.log — only a
+ * bare exit code hit the toast (#6198). Extracting + guarding all three steps here
+ * restores the symmetry and makes the behavior unit-testable without spawning the
+ * MITM server.
+ */
+export async function provisionDnsEntries(
+  sudoPassword: string,
+  deps: DnsProvisionDeps = {}
+): Promise<void> {
+  const canElevate = deps.canElevate ?? (() => isSudoAvailable() || isRoot());
+  const logger = deps.logger ?? defaultLog;
+
+  // Explicit opt-out: skip all DNS modification when the env var is set.
+  if (process.env.SKIP_ANTIGRAVITY_DNS === "true") {
+    logger.info("Skipping DNS entries - SKIP_ANTIGRAVITY_DNS=true");
+    return;
+  }
+
+  // In containers (USER node, no sudo installed, not root) we cannot write
+  // to /etc/hosts. Rather than attempting sudo and swallowing the error,
+  // detect the condition up-front and bail out with a clear message.
+  if (!canElevate()) {
+    logger.info(
+      "Skipping DNS entries - sudo not available and not running as root (likely a container)"
+    );
+    return;
   }
+
+  const resolvedDeps: ResolvedDnsProvisionDeps = {
+    addDefaultDns: deps.addDefaultDns ?? addDNSEntry,
+    addHostsDns: deps.addHostsDns ?? addDNSEntries,
+    getAgentStates: deps.getAgentStates ?? getAllAgentBridgeStates,
+    listEnabledCustomHosts:
+      deps.listEnabledCustomHosts ?? (() => listCustomHosts({ enabledOnly: true })),
+    logger,
+  };
+
+  await provisionDefaultDns(sudoPassword, resolvedDeps);
+  await provisionAgentDns(sudoPassword, resolvedDeps);
+  await provisionCustomHostsDns(sudoPassword, resolvedDeps);
 }
diff --git a/src/server/authz/routeGuard.ts b/src/server/authz/routeGuard.ts
index ff2ec2c7e41..dbb4dadf001 100644
--- a/src/server/authz/routeGuard.ts
+++ b/src/server/authz/routeGuard.ts
@@ -37,6 +37,7 @@ export const LOCAL_ONLY_API_PREFIXES: ReadonlyArray<string> = [
   "/api/copilot/", // unauthenticated LLM driver — CLI-only by default; admins can opt-in to remote access via manage-scope bypass
   "/api/tools/agent-bridge/", // AgentBridge: spawns MITM server + DNS edits (Hard Rules #15 + #17)
   "/api/tools/traffic-inspector/", // Traffic Inspector: http-proxy listener + system proxy (Hard Rules #15 + #17)
+  "/api/issue-agent/", // Issue Agent: recorded/local triage executor surface; keep loopback/LAN until sandbox + audit hardening is complete
   "/api/plugins/", // plugins: load/execute via worker_threads + child_process (Hard Rules #15 + #17)
   "/api/plugins", // bare path: GET list + POST install also trigger plugin loading
   "/api/middleware/", // SECURITY_AUDIT M8: middleware hooks compile+run arbitrary JS via new vm.Script (src/lib/middleware/registry.ts) on the request hot path — same code-exec class as /api/plugins/, so loopback-gate it for parity (Hard Rules #15 + #17)
diff --git a/src/server/ws/liveServer.ts b/src/server/ws/liveServer.ts
index a4a910a8fb6..3c5f91c4e11 100644
--- a/src/server/ws/liveServer.ts
+++ b/src/server/ws/liveServer.ts
@@ -27,6 +27,7 @@ import { emit, on, onAny, getEventHistory, type HistoryEntry } from "@/lib/event
 import type { DashboardEventName, DashboardEventMap, DashboardChannel } from "@/lib/events/types";
 
 import { CHANNEL_EVENTS, getChannelForEvent } from "@/lib/events/types";
+import { isAutomatedTestProcess, isBuildProcess } from "@/shared/utils/testProcess";
 
 import {
   buildAllowedOrigins,
@@ -608,10 +609,7 @@ export async function startLiveDashboardServer(
 
 function isBuildOrTest(): boolean {
   return (
-    process.env.NEXT_PHASE === "phase-production-build" ||
-    process.env.NODE_ENV === "test" ||
-    process.env.VITEST !== undefined ||
-    process.argv.some((arg) => arg.includes("test"))
+    isBuildProcess() || isAutomatedTestProcess()
   );
 }
 
diff --git a/src/shared/components/Sidebar.tsx b/src/shared/components/Sidebar.tsx
index ead99f02244..34dafb09988 100644
--- a/src/shared/components/Sidebar.tsx
+++ b/src/shared/components/Sidebar.tsx
@@ -6,6 +6,11 @@ import { usePathname } from "next/navigation";
 import { cn } from "@/shared/utils/cn";
 import { getActiveSidebarHref } from "@/shared/utils/sidebarRouteMatch";
 import { filterSidebarSectionsByQuery } from "@/shared/utils/sidebarSearch";
+import {
+  expandActiveSection,
+  hydrateExpandedSections,
+  toggleExpandedSection,
+} from "@/shared/utils/sidebarExpansionState";
 import { APP_CONFIG } from "@/shared/constants/appConfig";
 import OmniRouteLogo from "./OmniRouteLogo";
 import Button from "./Button";
@@ -102,10 +107,12 @@ export default function Sidebar({
     new Set([DEFAULT_EXPANDED])
   );
   const [pinnedSections, setPinnedSections] = useState<Set<SidebarSectionId>>(new Set());
+  const [sidebarExpansionLoaded, setSidebarExpansionLoaded] = useState(false);
+  const skipInitialActiveExpansion = useRef(false);
   const [hoveredItem, setHoveredItem] = useState<HoveredItem>(null);
   const [searchQuery, setSearchQuery] = useState("");
 
-  // Load persisted state on mount; OmniProxy is pinned by default on first visit
+  // Load persisted state on mount. A stored [] intentionally means "all sections collapsed".
   useEffect(() => {
     const storedExpanded = loadFromStorage<SidebarSectionId[]>(EXPANDED_SECTIONS_KEY, [
       DEFAULT_EXPANDED,
@@ -122,15 +129,13 @@ export default function Sidebar({
         ? (JSON.parse(pinnedRaw) as SidebarSectionId[])
         : (SIDEBAR_SECTIONS.filter((s) => s.defaultPinned).map((s) => s.id) as SidebarSectionId[]);
 
-    const initialExpanded = new Set<SidebarSectionId>(
-      storedExpanded.length > 0 ? storedExpanded : [DEFAULT_EXPANDED]
-    );
     const initialPinned = new Set<SidebarSectionId>(storedPinned);
-    // Pinned sections must also be expanded
-    for (const id of initialPinned) initialExpanded.add(id);
+    const initialExpanded = hydrateExpandedSections(storedExpanded, initialPinned);
 
+    skipInitialActiveExpansion.current = storedExpanded.length === 0;
     setExpandedSections(initialExpanded);
     setPinnedSections(initialPinned);
+    setSidebarExpansionLoaded(true);
   }, []);
 
   useEffect(() => {
@@ -274,18 +279,21 @@ export default function Sidebar({
     ? filterSidebarSectionsByQuery(visibleSections, searchQuery)
     : visibleSections;
 
-  // Auto-expand the section containing the active page (without closing others)
+  // Keep the active page visible while preserving accordion semantics for unpinned sections.
   useEffect(() => {
-    if (collapsed) return;
+    if (collapsed || !sidebarExpansionLoaded) return;
+    if (skipInitialActiveExpansion.current) {
+      skipInitialActiveExpansion.current = false;
+      return;
+    }
     for (const section of visibleSections) {
       const sectionItems = section.children.flatMap((child: any) =>
         child.type === "group" ? child.items : [child]
       );
       if (sectionItems.some((item: any) => !item.external && item.href === activeHref)) {
         setExpandedSections((prev) => {
-          if (prev.has(section.id as SidebarSectionId)) return prev;
-          const next = new Set(prev);
-          next.add(section.id as SidebarSectionId);
+          const next = expandActiveSection(pinnedSections, section.id as SidebarSectionId);
+          if ([...next].every((id) => prev.has(id)) && next.size === prev.size) return prev;
           saveToStorage(EXPANDED_SECTIONS_KEY, [...next]);
           return next;
         });
@@ -293,26 +301,13 @@ export default function Sidebar({
       }
     }
     // eslint-disable-next-line react-hooks/exhaustive-deps
-  }, [activeHref, collapsed]);
+  }, [activeHref, collapsed, pinnedSections, sidebarExpansionLoaded]);
 
   // Accordion toggle: opening a section closes all non-pinned sections
   const toggleSection = useCallback(
     (sectionId: SidebarSectionId) => {
       setExpandedSections((prev) => {
-        const isOpen = prev.has(sectionId);
-        let next: Set<SidebarSectionId>;
-        if (isOpen) {
-          // Close this section
-          next = new Set(prev);
-          next.delete(sectionId);
-        } else {
-          // Accordion: keep only pinned sections + the new one
-          next = new Set<SidebarSectionId>();
-          for (const id of pinnedSections) {
-            next.add(id);
-          }
-          next.add(sectionId);
-        }
+        const next = toggleExpandedSection(prev, pinnedSections, sectionId);
         saveToStorage(EXPANDED_SECTIONS_KEY, [...next]);
         return next;
       });
diff --git a/src/shared/components/flow/FlowCanvas.tsx b/src/shared/components/flow/FlowCanvas.tsx
index d27c7ed9230..f106e67d863 100644
--- a/src/shared/components/flow/FlowCanvas.tsx
+++ b/src/shared/components/flow/FlowCanvas.tsx
@@ -121,6 +121,7 @@ export function FlowCanvas({
     <div ref={containerRef} className={className}>
       <ReactFlow
         key={fitKey}
+        className="omniroute-flow"
         nodes={nodes}
         edges={edges}
         nodeTypes={nodeTypes}
diff --git a/src/shared/components/flow/edgeStyles.ts b/src/shared/components/flow/edgeStyles.ts
index 839e2983ebc..3aeb4f17fff 100644
--- a/src/shared/components/flow/edgeStyles.ts
+++ b/src/shared/components/flow/edgeStyles.ts
@@ -10,7 +10,7 @@ export const FLOW_EDGE_COLORS = {
   active: STATUS_HEX.success,
   error: STATUS_HEX.error,
   last: STATUS_HEX.warning,
-  idle: "var(--color-border)",
+  idle: "var(--color-text-muted)",
 } as const;
 
 export interface FlowEdgeStyle {
@@ -28,5 +28,5 @@ export function edgeStyle(active: boolean, last: boolean, error: boolean): FlowE
   if (error) return { stroke: FLOW_EDGE_COLORS.error, strokeWidth: 2, opacity: 0.85 };
   if (active) return { stroke: FLOW_EDGE_COLORS.active, strokeWidth: 2.5, opacity: 1 };
   if (last) return { stroke: FLOW_EDGE_COLORS.last, strokeWidth: 1.5, opacity: 0.6 };
-  return { stroke: FLOW_EDGE_COLORS.idle, strokeWidth: 1, opacity: 0.2 };
+  return { stroke: FLOW_EDGE_COLORS.idle, strokeWidth: 1, opacity: 0.3 };
 }
diff --git a/src/shared/constants/mcpScopes.ts b/src/shared/constants/mcpScopes.ts
index 62bda5a99c0..d5babbb63c4 100644
--- a/src/shared/constants/mcpScopes.ts
+++ b/src/shared/constants/mcpScopes.ts
@@ -63,6 +63,12 @@ export const MCP_TOOL_SCOPES: Record<string, readonly McpScope[]> = {
   omniroute_set_compression_engine: ["write:compression"],
   omniroute_list_compression_combos: ["read:compression"],
   omniroute_compression_combo_stats: ["read:compression"],
+  omniroute_ccr_store: ["write:compression"],
+  omniroute_ccr_retrieve: ["read:compression"],
+  omniroute_ccr_inspect: ["read:compression"],
+  omniroute_ccr_list: ["read:compression"],
+  omniroute_ccr_delete: ["write:compression"],
+  omniroute_ccr_stats: ["read:compression"],
   omniroute_oneproxy_fetch: ["read:proxies"],
   omniroute_oneproxy_rotate: ["read:proxies"],
   omniroute_oneproxy_stats: ["read:proxies"],
diff --git a/src/shared/constants/providers/apikey/regional.ts b/src/shared/constants/providers/apikey/regional.ts
index b6708bd2685..5b893cb2dc6 100644
--- a/src/shared/constants/providers/apikey/regional.ts
+++ b/src/shared/constants/providers/apikey/regional.ts
@@ -351,4 +351,16 @@ export const APIKEY_PROVIDERS_REGIONAL = {
     passthroughModels: true,
     authHint: "Get API key at api.hcnsec.cn",
   },
+  agnes: {
+    id: "agnes",
+    alias: "agnes",
+    name: "Agnes AI",
+    icon: "auto_awesome",
+    color: "#10B981",
+    textIcon: "AG",
+    website: "https://agnes-ai.com",
+    hasFree: true,
+    freeNote: "Permanently free API - no credit card required.",
+    authHint: "Get API key at agnes-ai.com",
+  },
 };
diff --git a/src/shared/constants/sidebarVisibility.ts b/src/shared/constants/sidebarVisibility.ts
index 3d24d91343f..5de6d26425b 100644
--- a/src/shared/constants/sidebarVisibility.ts
+++ b/src/shared/constants/sidebarVisibility.ts
@@ -4,6 +4,7 @@ export { COMPRESSION_CONTEXT_GROUP, SIDEBAR_SECTIONS } from "./sidebarVisibility
 import { HIDEABLE_SIDEBAR_ITEM_IDS } from "./sidebarVisibility/types";
 import type {
   HideableSidebarItemId,
+  SidebarItemId,
   SidebarSectionId,
   SidebarItemDefinition,
   SidebarSectionChild,
@@ -11,7 +12,7 @@ import type {
   SidebarPresetDefinition,
 } from "./sidebarVisibility/types";
 
-export const SIDEBAR_ICON_ACCENTS: Partial<Record<HideableSidebarItemId, string>> = {
+export const SIDEBAR_ICON_ACCENTS: Partial<Record<SidebarItemId, string>> = {
   home: "#60A5FA",
   "api-manager": "#F59E0B",
   endpoints: "#38BDF8",
@@ -117,7 +118,7 @@ function getDeterministicIconAccent(id: string): string {
 
 export function getSidebarIconAccent(id: string): string {
   return (
-    SIDEBAR_ICON_ACCENTS[id as HideableSidebarItemId] ||
+    SIDEBAR_ICON_ACCENTS[id as SidebarItemId] ||
     SIDEBAR_SUBITEM_ICON_ACCENTS[id] ||
     getDeterministicIconAccent(id)
   );
diff --git a/src/shared/constants/sidebarVisibility/sections.ts b/src/shared/constants/sidebarVisibility/sections.ts
index 67f12721165..3ec67d3f755 100644
--- a/src/shared/constants/sidebarVisibility/sections.ts
+++ b/src/shared/constants/sidebarVisibility/sections.ts
@@ -731,7 +731,6 @@ export const SIDEBAR_SECTIONS: readonly SidebarSectionDefinition[] = [
       INTEGRATIONS_GROUP,
       PROXY_ITEM,
     ],
-    defaultPinned: true,
   },
   {
     id: "analytics",
diff --git a/src/shared/constants/sidebarVisibility/types.ts b/src/shared/constants/sidebarVisibility/types.ts
index 1a4f2cb2b9d..c5172189d52 100644
--- a/src/shared/constants/sidebarVisibility/types.ts
+++ b/src/shared/constants/sidebarVisibility/types.ts
@@ -34,8 +34,7 @@ export const HIDEABLE_SIDEBAR_ITEM_IDS = [
   // OmniProxy > Integrations
   "api-endpoints",
   "webhooks",
-  // OmniProxy — proxy
-  "proxy",
+  // OmniProxy — proxy tools
   "mitm-proxy",
   "1proxy",
   // Analytics
@@ -106,6 +105,11 @@ export const HIDEABLE_SIDEBAR_ITEM_IDS = [
 
 export type HideableSidebarItemId = (typeof HIDEABLE_SIDEBAR_ITEM_IDS)[number];
 
+/** Sidebar entries that are intentionally always available and cannot be hidden by presets. */
+export type AlwaysVisibleSidebarItemId = "proxy";
+
+export type SidebarItemId = HideableSidebarItemId | AlwaysVisibleSidebarItemId;
+
 export type SidebarSectionId =
   | "home"
   | "omni-proxy"
@@ -119,7 +123,7 @@ export type SidebarSectionId =
   | "help";
 
 export interface SidebarItemDefinition {
-  id: HideableSidebarItemId;
+  id: SidebarItemId;
   href: string;
   i18nKey: string;
   subtitleKey?: string;
diff --git a/src/shared/hooks/__tests__/useDisplayBaseUrl.test.tsx b/src/shared/hooks/__tests__/useDisplayBaseUrl.test.tsx
index c6dd92b1374..1750942d7fc 100644
--- a/src/shared/hooks/__tests__/useDisplayBaseUrl.test.tsx
+++ b/src/shared/hooks/__tests__/useDisplayBaseUrl.test.tsx
@@ -2,7 +2,11 @@
 import React, { act } from "react";
 import { createRoot } from "react-dom/client";
 import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
-import { DEFAULT_DISPLAY_BASE_URL } from "../useDisplayBaseUrl";
+import {
+  DEFAULT_DISPLAY_BASE_URL,
+  isPublicDisplayBaseUrl,
+  resolveDisplayBaseUrl,
+} from "../useDisplayBaseUrl";
 
 const cleanupCallbacks: Array<() => void> = [];
 
@@ -64,6 +68,108 @@ describe("useDisplayBaseUrl", () => {
     );
   });
 
+  it("classifies public domains separately from local and private addresses", () => {
+    expect(isPublicDisplayBaseUrl("https://api.example.com")).toBe(true);
+    expect(isPublicDisplayBaseUrl("http://localhost:20128")).toBe(false);
+    expect(isPublicDisplayBaseUrl("http://192.168.1.25:20128")).toBe(false);
+    expect(isPublicDisplayBaseUrl("http://100.88.4.55:20128")).toBe(false);
+    expect(isPublicDisplayBaseUrl("http://[::1]:20128")).toBe(false);
+  });
+
+  it("classifies IPv4 private ranges at their exact boundaries", () => {
+    const cases: Array<[host: string, expectedPublic: boolean]> = [
+      // 0.0.0.0/8 ("this" network) vs just above it
+      ["0.0.0.1", false],
+      ["1.0.0.1", true],
+      // RFC1918 10.0.0.0/8 vs just outside
+      ["9.255.255.255", true],
+      ["10.0.0.1", false],
+      ["11.0.0.0", true],
+      // loopback 127.0.0.0/8 vs just outside
+      ["126.255.255.255", true],
+      ["127.0.0.1", false],
+      ["128.0.0.0", true],
+      // multicast/reserved 224.0.0.0+ vs just below
+      ["223.255.255.255", true],
+      ["224.0.0.1", false],
+      // CGNAT RFC6598 100.64.0.0/10 vs just outside
+      ["100.63.255.255", true],
+      ["100.64.0.0", false],
+      ["100.127.255.255", false],
+      ["100.128.0.0", true],
+      // link-local 169.254.0.0/16 vs just outside
+      ["169.253.255.255", true],
+      ["169.254.0.1", false],
+      ["169.255.0.0", true],
+      // RFC1918 172.16.0.0/12 vs just outside
+      ["172.15.255.255", true],
+      ["172.16.0.0", false],
+      ["172.31.255.255", false],
+      ["172.32.0.0", true],
+      // RFC1918 192.168.0.0/16 vs just outside
+      ["192.167.255.255", true],
+      ["192.168.0.1", false],
+      ["192.169.0.0", true],
+    ];
+
+    for (const [host, expectedPublic] of cases) {
+      expect(isPublicDisplayBaseUrl(`http://${host}:20128`)).toBe(expectedPublic);
+    }
+  });
+
+  it("classifies IPv6 special ranges while keeping the check scoped to actual IPv6 hosts", () => {
+    expect(isPublicDisplayBaseUrl("http://[::]:20128")).toBe(false);
+    expect(isPublicDisplayBaseUrl("http://[::1]:20128")).toBe(false);
+    expect(isPublicDisplayBaseUrl("http://[fc00::1]:20128")).toBe(false);
+    expect(isPublicDisplayBaseUrl("http://[fd12::1]:20128")).toBe(false);
+    expect(isPublicDisplayBaseUrl("http://[fe80::1]:20128")).toBe(false);
+    expect(isPublicDisplayBaseUrl("http://[2001:db8::1]:20128")).toBe(true);
+    // A hostname that merely STARTS WITH "fd"/"fc" must stay public — the ULA/
+    // link-local checks are IPv6-only and must not leak into hostname matching.
+    expect(isPublicDisplayBaseUrl("http://fdroid.example.com:20128")).toBe(true);
+    expect(isPublicDisplayBaseUrl("http://fcbar.example.com:20128")).toBe(true);
+  });
+
+  it("keeps a configured public URL when the browser is on a local address", () => {
+    expect(resolveDisplayBaseUrl("https://api.example.com/", "http://localhost:20128")).toBe(
+      "https://api.example.com"
+    );
+  });
+
+  it("prefers the currently reachable public origin over another configured URL", () => {
+    expect(resolveDisplayBaseUrl("https://old.example.com", "https://api.example.com/")).toBe(
+      "https://api.example.com"
+    );
+  });
+
+  it("prefers a public browser origin over a loopback build-time value", async () => {
+    vi.stubEnv("NEXT_PUBLIC_BASE_URL", "http://localhost:20128");
+    vi.stubGlobal("location", { origin: "https://api.example.com" });
+
+    const { useDisplayBaseUrl } = await import("../useDisplayBaseUrl");
+    const container = makeContainer();
+    const root = createRoot(container);
+
+    function C() {
+      const url = useDisplayBaseUrl();
+      return <span data-testid="value">{url}</span>;
+    }
+
+    act(() => {
+      root.render(<C />);
+    });
+
+    expect(container.querySelector('[data-testid="value"]')?.textContent).toBe(
+      DEFAULT_DISPLAY_BASE_URL
+    );
+
+    await act(async () => {});
+
+    expect(container.querySelector('[data-testid="value"]')?.textContent).toBe(
+      "https://api.example.com"
+    );
+  });
+
   it("returns DEFAULT_DISPLAY_BASE_URL on first render and origin after mount when env unset", async () => {
     vi.stubEnv("NEXT_PUBLIC_BASE_URL", "");
 
diff --git a/src/shared/hooks/cli/useToolBatchStatuses.ts b/src/shared/hooks/cli/useToolBatchStatuses.ts
index 932dae00c74..692574c086b 100644
--- a/src/shared/hooks/cli/useToolBatchStatuses.ts
+++ b/src/shared/hooks/cli/useToolBatchStatuses.ts
@@ -15,11 +15,12 @@ export function useToolBatchStatuses(): UseToolBatchStatusesResult {
   const [loading, setLoading] = useState<boolean>(true);
   const [error, setError] = useState<string | null>(null);
 
-  const fetchStatuses = useCallback(async () => {
+  const fetchStatuses = useCallback(async (forceRefresh = false) => {
     setLoading(true);
     setError(null);
     try {
-      const res = await fetch("/api/cli-tools/all-statuses");
+      const url = `/api/cli-tools/all-statuses${forceRefresh ? "?refresh=true" : ""}`;
+      const res = await fetch(url);
       if (!res.ok) {
         const text = await res.text().catch(() => String(res.status));
         setError(`HTTP ${res.status}: ${text.slice(0, 200)}`);
@@ -37,6 +38,10 @@ export function useToolBatchStatuses(): UseToolBatchStatusesResult {
     }
   }, []);
 
+  const refetch = useCallback(() => {
+    void fetchStatuses(true);
+  }, [fetchStatuses]);
+
   useEffect(() => {
     void fetchStatuses();
 
@@ -50,5 +55,5 @@ export function useToolBatchStatuses(): UseToolBatchStatusesResult {
     };
   }, [fetchStatuses]);
 
-  return { statuses, loading, error, refetch: fetchStatuses };
+  return { statuses, loading, error, refetch };
 }
diff --git a/src/shared/hooks/index.ts b/src/shared/hooks/index.ts
index 896a06fc7e2..4c7c0594153 100644
--- a/src/shared/hooks/index.ts
+++ b/src/shared/hooks/index.ts
@@ -1,3 +1,8 @@
 // Shared Hooks - Export all
 export { useTheme } from "./useTheme";
-export { useDisplayBaseUrl, DEFAULT_DISPLAY_BASE_URL } from "./useDisplayBaseUrl";
+export {
+  useDisplayBaseUrl,
+  DEFAULT_DISPLAY_BASE_URL,
+  isPublicDisplayBaseUrl,
+  resolveDisplayBaseUrl,
+} from "./useDisplayBaseUrl";
diff --git a/src/shared/hooks/useDisplayBaseUrl.ts b/src/shared/hooks/useDisplayBaseUrl.ts
index 96081ba5831..9bb997dc863 100644
--- a/src/shared/hooks/useDisplayBaseUrl.ts
+++ b/src/shared/hooks/useDisplayBaseUrl.ts
@@ -10,13 +10,122 @@ function normalizeUrl(value?: string): string | null {
   return trimmed.replace(/\/+$/, "");
 }
 
+/**
+ * One RFC1918 / special-use IPv4 range, expressed as closed intervals on the
+ * first two octets. Unbounded ends use +/-Infinity so a single numeric
+ * comparison covers them without an extra branch.
+ */
+interface Ipv4Range {
+  readonly firstMin: number;
+  readonly firstMax: number;
+  readonly secondMin: number;
+  readonly secondMax: number;
+}
+
+/** RFC1918 + special-use IPv4 ranges treated as non-public for display purposes. */
+const PRIVATE_IPV4_RANGES: readonly Ipv4Range[] = [
+  { firstMin: 0, firstMax: 0, secondMin: -Infinity, secondMax: Infinity }, // 0.0.0.0/8 ("this" network)
+  { firstMin: 10, firstMax: 10, secondMin: -Infinity, secondMax: Infinity }, // RFC1918 10.0.0.0/8
+  { firstMin: 127, firstMax: 127, secondMin: -Infinity, secondMax: Infinity }, // loopback 127.0.0.0/8
+  { firstMin: 224, firstMax: Infinity, secondMin: -Infinity, secondMax: Infinity }, // multicast/reserved/broadcast
+  { firstMin: 100, firstMax: 100, secondMin: 64, secondMax: 127 }, // CGNAT RFC6598 100.64.0.0/10
+  { firstMin: 169, firstMax: 169, secondMin: 254, secondMax: 254 }, // link-local 169.254.0.0/16
+  { firstMin: 172, firstMax: 172, secondMin: 16, secondMax: 31 }, // RFC1918 172.16.0.0/12
+  { firstMin: 192, firstMax: 192, secondMin: 168, secondMax: 168 }, // RFC1918 192.168.0.0/16
+];
+
+function isInIpv4Range(first: number, second: number, range: Ipv4Range): boolean {
+  return (
+    first >= range.firstMin &&
+    first <= range.firstMax &&
+    second >= range.secondMin &&
+    second <= range.secondMax
+  );
+}
+
+function isPrivateIpv4(hostname: string): boolean {
+  const octets = hostname.split(".").map(Number);
+  if (octets.length !== 4 || octets.some((octet) => !Number.isInteger(octet))) return false;
+  const [first, second] = octets;
+  return PRIVATE_IPV4_RANGES.some((range) => isInIpv4Range(first, second, range));
+}
+
+function isSupportedProtocol(protocol: string): boolean {
+  return protocol === "http:" || protocol === "https:";
+}
+
+function isLoopbackHostname(hostname: string): boolean {
+  return !hostname || hostname === "localhost" || hostname.endsWith(".localhost");
+}
+
+function isMulticastDnsHostname(hostname: string): boolean {
+  return hostname.endsWith(".local");
+}
+
+function isIpv6LoopbackOrUnspecified(hostname: string): boolean {
+  return hostname === "::" || hostname === "::1";
+}
+
+function isIpv6UniqueLocal(hostname: string): boolean {
+  // RFC 4193 Unique Local Addresses: fc00::/7 (prefixes "fc" and "fd").
+  return hostname.startsWith("fc") || hostname.startsWith("fd");
+}
+
+const IPV6_LINK_LOCAL_PATTERN = /^fe[89ab]/;
+
+function isIpv6LinkLocal(hostname: string): boolean {
+  // RFC 4291 link-local: fe80::/10.
+  return IPV6_LINK_LOCAL_PATTERN.test(hostname);
+}
+
+/** Combines the IPv6-specific non-public checks the caller gates on `isIpv6`. */
+function isNonPublicIpv6(hostname: string): boolean {
+  return (
+    isIpv6LoopbackOrUnspecified(hostname) ||
+    isIpv6UniqueLocal(hostname) ||
+    isIpv6LinkLocal(hostname)
+  );
+}
+
+export function isPublicDisplayBaseUrl(value?: string): boolean {
+  const normalized = normalizeUrl(value);
+  if (!normalized) return false;
+
+  try {
+    const parsed = new URL(normalized);
+    if (!isSupportedProtocol(parsed.protocol)) return false;
+
+    const hostname = parsed.hostname.toLowerCase().replace(/^\[|\]$/g, "");
+    if (isLoopbackHostname(hostname)) return false;
+    if (isMulticastDnsHostname(hostname) || isPrivateIpv4(hostname)) return false;
+
+    // IPv6-only checks stay gated on isIpv6 — hostnames like "fdroid.example.com"
+    // legitimately start with "fd" and must not be misclassified as ULA addresses.
+    const isIpv6 = hostname.includes(":");
+    if (isIpv6 && isNonPublicIpv6(hostname)) return false;
+
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+export function resolveDisplayBaseUrl(envValue?: string, browserOrigin?: string): string {
+  const configuredUrl = normalizeUrl(envValue);
+  const currentOrigin = normalizeUrl(browserOrigin);
+
+  if (currentOrigin && isPublicDisplayBaseUrl(currentOrigin)) return currentOrigin;
+  if (configuredUrl && isPublicDisplayBaseUrl(configuredUrl)) return configuredUrl;
+  return currentOrigin ?? configuredUrl ?? DEFAULT_DISPLAY_BASE_URL;
+}
+
 /**
  * Returns the public base URL to display in the dashboard.
  *
- * Resolution chain:
- *   1. NEXT_PUBLIC_BASE_URL (env, trimmed + slash-normalized) — wins if set.
- *   2. window.location.origin after client mount — when env is unset.
- *   3. DEFAULT_DISPLAY_BASE_URL ("http://localhost:20128") — SSR / first render fallback.
+ * Resolution chain after client mount:
+ *   1. Public browser origin — proves the current tunnel/domain is reachable.
+ *   2. Public NEXT_PUBLIC_BASE_URL — keeps a configured public URL when opened locally.
+ *   3. Current browser origin, configured URL, then localhost as local fallbacks.
  *
  * DISPLAY ONLY — do NOT use this hook for OAuth `redirect_uri`.
  * OAuth callers must read `process.env.NEXT_PUBLIC_BASE_URL` directly to avoid
@@ -29,8 +138,7 @@ export function useDisplayBaseUrl(): string {
   const [url, setUrl] = useState<string>(envValue ?? DEFAULT_DISPLAY_BASE_URL);
 
   useEffect(() => {
-    if (envValue) return;
-    const origin = normalizeUrl(window.location.origin) ?? DEFAULT_DISPLAY_BASE_URL;
+    const resolvedUrl = resolveDisplayBaseUrl(envValue ?? undefined, window.location.origin);
     // Schedule via queueMicrotask so setState is called inside a callback,
     // not synchronously in the effect body (react-hooks/set-state-in-effect).
     // The unmounted guard prevents a stale setState on a torn-down root
@@ -38,7 +146,7 @@ export function useDisplayBaseUrl(): string {
     // before the microtask fires on the first effect invocation).
     let unmounted = false;
     queueMicrotask(() => {
-      if (!unmounted) setUrl(origin);
+      if (!unmounted) setUrl(resolvedUrl);
     });
     return () => {
       unmounted = true;
diff --git a/src/shared/services/cliRuntime.ts b/src/shared/services/cliRuntime.ts
index 371076b0c6b..279b8a22c48 100644
--- a/src/shared/services/cliRuntime.ts
+++ b/src/shared/services/cliRuntime.ts
@@ -114,13 +114,13 @@ const CLI_TOOLS: Record<string, any> = {
     },
   },
   continue: {
-    defaultCommand: null,
+    defaultCommand: "cn",
     envBinKey: "CLI_CONTINUE_BIN",
-    requiresBinary: false,
+    requiresBinary: true,
     // opencode and continue may take up to 15s on first run / cold start on VPS
     healthcheckTimeoutMs: 15000,
     paths: {
-      settings: ".continue/config.json",
+      settings: ".continue/config.yaml",
     },
   },
   opencode: {
@@ -571,6 +571,7 @@ export const getKnownToolPaths = (toolId: string): string[] => {
     ],
     cline: [["cline.cmd", "cline"]],
     kilo: [["kilocode.cmd", "kilocode"]],
+    continue: [["cn.cmd", "cn"]],
     opencode: [["opencode.cmd", "opencode"]],
     qoder: [
       ["qodercli.cmd", "qodercli"],
@@ -1112,6 +1113,7 @@ export const getCliRuntimeStatus = async (toolId: string) => {
   return {
     installed: true,
     runnable: healthcheck.runnable,
+    version: healthcheck.version,
     command,
     commandPath: located.commandPath,
     reason: healthcheck.reason,
diff --git a/src/shared/types/utilization.ts b/src/shared/types/utilization.ts
index 2a8e3cfeb77..7a257e5470b 100644
--- a/src/shared/types/utilization.ts
+++ b/src/shared/types/utilization.ts
@@ -82,6 +82,8 @@ export interface ComboRecord {
   name?: string;
   strategy?: string;
   models?: unknown[];
+  autoConfig?: unknown;
+  config?: unknown;
 }
 
 export type UtilizationTimeRange = "1h" | "24h" | "7d" | "30d";
@@ -270,11 +272,9 @@ export type ComboScoringInspectorFactorKey =
   | "resetWindowAffinity";
 
 export type ComboScoringInspectorSource =
-  | "combo_health"
-  | "combo_forecast"
-  | "combo_autopilot"
-  | "runtime"
-  | "default";
+  "combo_health" | "combo_forecast" | "combo_autopilot" | "runtime" | "default";
+
+export type ComboScoringInspectorWeightSource = "default" | "explicit" | "mode_pack";
 
 export interface ComboScoringInspectorFactor {
   key: ComboScoringInspectorFactorKey;
@@ -376,6 +376,8 @@ export interface ComboScoringInspectorCombo {
   strategy: string;
   taskType: string;
   weights: Record<ComboScoringInspectorFactorKey, number>;
+  weightSource: ComboScoringInspectorWeightSource;
+  modePack: string | null;
   selectedExecutionKey: string | null;
   targets: ComboScoringInspectorTarget[];
   warnings: string[];
diff --git a/src/shared/utils/providerQuotaVisibility.ts b/src/shared/utils/providerQuotaVisibility.ts
new file mode 100644
index 00000000000..39680052046
--- /dev/null
+++ b/src/shared/utils/providerQuotaVisibility.ts
@@ -0,0 +1,13 @@
+import { USAGE_SUPPORTED_PROVIDERS } from "@/shared/constants/providers";
+
+export interface ProviderQuotaVisibilityConnection {
+  quotaVisible?: boolean;
+}
+
+export function isProviderQuotaVisible(connection: ProviderQuotaVisibilityConnection): boolean {
+  return connection.quotaVisible !== false;
+}
+
+export function supportsProviderQuota(providerId: string): boolean {
+  return USAGE_SUPPORTED_PROVIDERS.includes(providerId);
+}
diff --git a/src/shared/utils/sidebarExpansionState.ts b/src/shared/utils/sidebarExpansionState.ts
new file mode 100644
index 00000000000..778e546798f
--- /dev/null
+++ b/src/shared/utils/sidebarExpansionState.ts
@@ -0,0 +1,29 @@
+import type { SidebarSectionId } from "@/shared/constants/sidebarVisibility";
+
+export function hydrateExpandedSections(
+  storedExpanded: readonly SidebarSectionId[],
+  pinned: ReadonlySet<SidebarSectionId>
+): Set<SidebarSectionId> {
+  return new Set([...storedExpanded, ...pinned]);
+}
+
+export function toggleExpandedSection(
+  expanded: ReadonlySet<SidebarSectionId>,
+  pinned: ReadonlySet<SidebarSectionId>,
+  sectionId: SidebarSectionId
+): Set<SidebarSectionId> {
+  if (expanded.has(sectionId)) {
+    const next = new Set(expanded);
+    next.delete(sectionId);
+    return next;
+  }
+
+  return new Set([...pinned, sectionId]);
+}
+
+export function expandActiveSection(
+  pinned: ReadonlySet<SidebarSectionId>,
+  sectionId: SidebarSectionId
+): Set<SidebarSectionId> {
+  return new Set([...pinned, sectionId]);
+}
diff --git a/src/shared/utils/testProcess.ts b/src/shared/utils/testProcess.ts
new file mode 100644
index 00000000000..26cf5493ed1
--- /dev/null
+++ b/src/shared/utils/testProcess.ts
@@ -0,0 +1,44 @@
+/**
+ * "Am I running under an automated test runner?" — one answer, one place.
+ *
+ * Eleven subsystems each carried their own copy of this check, and every copy decided it with
+ * `process.argv.some((arg) => arg.includes("test"))`. That substring test is wrong in a way that is
+ * easy to miss and expensive to hit: `'latest'.includes('test') === true`. Any argv carrying a
+ * `latest` segment — a release symlink (`/opt/app/latest/server.js`), an npm/npx cache path, a
+ * `--model=latest` flag — silently flipped the process into "test mode", and the subsystems gated on
+ * it turned themselves off without a word. The worst of them is `src/lib/db/backup.ts`:
+ * `isSqliteAutoBackupDisabled()` returns true, so SQLite auto-backup simply never runs. Same class of
+ * silent disable for migrations, cloud sync, the health checks and quota recovery.
+ *
+ * Matching rules:
+ *  - env first (`NODE_ENV=test`, `VITEST`) — unchanged behaviour.
+ *  - argv: `test`/`tests` only as a WHOLE token delimited by a path separator, dot or dash — so
+ *    `--test`, `tests/unit/x.test.ts` and `src/x.test.ts` match, while `latest`, `protest`,
+ *    `contest` and `attestation` do not.
+ *  - argv: known runner binaries (vitest/jest/mocha/ava/tap), which have no delimiter before "test".
+ *
+ * argv and env are parameters rather than globals so the behaviour is testable: under a test runner
+ * the globals always say "test", which is exactly why the substring bug could never be caught.
+ */
+
+/** `test`/`tests` as a whole token: `--test`, `/tests/`, `.test.ts` — but never `latest`/`protest`. */
+const TEST_TOKEN_RE = /(^|[\\/.-])tests?([\\/.-]|$)/i;
+
+/** Runner binaries: no delimiter precedes "test" in "vitest", so they need their own rule. */
+const TEST_RUNNER_RE = /(^|[\\/])(vitest|jest|mocha|ava|tap)(\.[cm]?js)?$/i;
+
+export function isAutomatedTestProcess(
+  // execArgv too: `node --test x.js` puts `--test` in execArgv, not argv (modelLockoutSettings was the
+  // only copy that remembered to look there — now every caller gets it).
+  argv: readonly string[] = typeof process !== "undefined" ? [...process.argv, ...process.execArgv] : [],
+  env: NodeJS.ProcessEnv = typeof process !== "undefined" ? process.env : ({} as NodeJS.ProcessEnv)
+): boolean {
+  if (env.NODE_ENV === "test") return true;
+  if (env.VITEST !== undefined) return true;
+  return argv.some((arg) => typeof arg === "string" && (TEST_TOKEN_RE.test(arg) || TEST_RUNNER_RE.test(arg)));
+}
+
+/** Next.js production build phase — several call sites pair this with the test check. */
+export function isBuildProcess(env: NodeJS.ProcessEnv = process.env): boolean {
+  return typeof process !== "undefined" && env.NEXT_PHASE === "phase-production-build";
+}
diff --git a/src/shared/validation/compressionConfigSchemas.ts b/src/shared/validation/compressionConfigSchemas.ts
index b640cf99ff3..745d2840438 100644
--- a/src/shared/validation/compressionConfigSchemas.ts
+++ b/src/shared/validation/compressionConfigSchemas.ts
@@ -314,6 +314,7 @@ export const compressionSettingsUpdateSchema = z
     ultra: ultraConfigSchema.optional(),
     contextBudget: contextBudgetConfigSchema.optional(),
     contextEditing: contextEditingConfigSchema.optional(),
+    liveZone: z.object({ enabled: z.boolean() }).strict().optional(),
     engines: z.record(z.string(), engineToggleSchema).optional(),
     enginesExplicit: z.boolean().optional(),
     activeComboId: z.string().nullable().optional(),
diff --git a/src/shared/validation/schemas/provider.ts b/src/shared/validation/schemas/provider.ts
index 95ee820d552..a315489958d 100644
--- a/src/shared/validation/schemas/provider.ts
+++ b/src/shared/validation/schemas/provider.ts
@@ -433,6 +433,7 @@ export const updateProviderConnectionSchema = z
       .optional(),
     proxyEnabled: z.boolean().optional(),
     perKeyProxyEnabled: z.boolean().optional(),
+    quotaVisible: z.boolean().optional(),
     // Partial patch of per-connection provider-specific settings (e.g. quota toggles)
     providerSpecificData: z
       .record(z.string(), z.unknown())
diff --git a/src/sse/handlers/chat.ts b/src/sse/handlers/chat.ts
index cefc5335616..3f639681592 100644
--- a/src/sse/handlers/chat.ts
+++ b/src/sse/handlers/chat.ts
@@ -22,9 +22,8 @@ import { resolveBareModelToConnectionDefault } from "@omniroute/open-sse/service
 import { errorResponse } from "@omniroute/open-sse/utils/error.ts";
 import { getImageModelEntry } from "@omniroute/open-sse/config/imageRegistry.ts";
 import { acceptHeaderForcesStream } from "@omniroute/open-sse/utils/aiSdkCompat.ts";
-import { isSelfInflictedUpstreamTimeout } from "@omniroute/open-sse/handlers/chatCore/cooldownClassification.ts";
 import { applyNoThinkingAlias } from "@omniroute/open-sse/utils/noThinkingAlias.ts";
-import { handleComboChat } from "@omniroute/open-sse/services/combo.ts";
+import { handleComboChat, shouldSkipConnDisable } from "@omniroute/open-sse/services/combo.ts";
 import { resolveRequestAutoControls } from "@omniroute/open-sse/services/autoCombo/requestControls.ts";
 import { resolveComboConfig } from "@omniroute/open-sse/services/comboConfig.ts";
 import { injectHandoffIntoBody } from "@omniroute/open-sse/services/contextHandoff.ts";
@@ -83,6 +82,7 @@ import {
   withCorrelationId,
 } from "./chatHelpers";
 import { connectionHasExtraKeys } from "@omniroute/open-sse/services/apiKeyRotator.ts";
+import { getComboFailureLogError, isRequestScopedUpstreamFailure } from "./comboFailureLogging";
 
 // Pipeline integration — wired modules
 import { classify429FromError, type FailureKind } from "@/shared/utils/classify429";
@@ -922,7 +922,7 @@ export async function handleChat(
           requestedModel: body?.model || resolvedModelStr,
           provider: "-",
           endpoint: clientRawRequest?.endpoint,
-          error: `[${response.status}] Combo "${combo.name}" failed — all targets exhausted`,
+          error: await getComboFailureLogError(response, combo.name),
           comboName: combo.name,
           apiKeyId: apiKeyInfo?.id ?? null,
           apiKeyName: apiKeyInfo?.name ?? null,
@@ -1726,13 +1726,7 @@ async function handleSingleModelChat(
         ((credentials.providerSpecificData?.extraApiKeys as string[] | undefined) ?? []).length >
           0 || connectionHasExtraKeys(credentials.connectionId);
       const is401 = result.status === 401;
-      // Our own timeout fired on a slow upstream; don't cool down a healthy account.
-      const skipConnectionDisable =
-        result.status === 499 ||
-        result.errorCode === "client_disconnected" ||
-        result.errorType === "client_disconnected" ||
-        (is401 && hasExtraKeys) ||
-        isSelfInflictedUpstreamTimeout(result.status, result.errorType, provider);
+      const skipConnectionDisable = shouldSkipConnDisable(result, is401, hasExtraKeys, provider);
 
       const { shouldFallback, cooldownMs } = skipConnectionDisable
         ? { shouldFallback: false, cooldownMs: 0 }
@@ -1785,6 +1779,7 @@ async function handleSingleModelChat(
       if (
         !forceLiveComboTest &&
         !isCombo &&
+        !isRequestScopedUpstreamFailure({ code: result.errorCode, type: result.errorType }) &&
         PROVIDER_BREAKER_FAILURE_STATUSES.has(Number(result.status))
       ) {
         breaker._onFailure();
diff --git a/src/sse/handlers/comboFailureLogging.ts b/src/sse/handlers/comboFailureLogging.ts
new file mode 100644
index 00000000000..1ccf4b1fdb6
--- /dev/null
+++ b/src/sse/handlers/comboFailureLogging.ts
@@ -0,0 +1,20 @@
+// Re-exported here (rather than a separate import line in chat.ts, a frozen
+// file-size chokepoint — see config/quality/file-size-baseline.json) so both
+// combo-failure-handling helpers chat.ts needs share a single import line.
+export { isRequestScopedUpstreamFailure } from "@omniroute/open-sse/services/combo/comboPredicates.ts";
+
+export async function getComboFailureLogError(
+  response: Response,
+  comboName: string
+): Promise<string> {
+  const fallback = `[${response.status}] Combo "${comboName}" failed`;
+  try {
+    const body = await response.clone().json();
+    const message = body?.error?.message;
+    return typeof message === "string" && message.trim()
+      ? `[${response.status}] ${message.trim()}`
+      : fallback;
+  } catch {
+    return fallback;
+  }
+}
diff --git a/src/types/databaseSettings.ts b/src/types/databaseSettings.ts
index f69b2141111..459f5803346 100644
--- a/src/types/databaseSettings.ts
+++ b/src/types/databaseSettings.ts
@@ -46,7 +46,10 @@ export interface DatabaseSettings {
     callLogs: number;
     usageHistory: number;
     memoryEntries: number;
+    domainCostHistory: number;
+    compressionCacheStats: number;
     xpAuditLog: number;
+    compressionRunTelemetry: number;
     autoCleanupEnabled: boolean;
   };
 
@@ -107,7 +110,10 @@ export const DEFAULT_DATABASE_SETTINGS: Omit<DatabaseSettings, "location" | "sta
     callLogs: 30,
     usageHistory: 30,
     memoryEntries: 30,
+    domainCostHistory: 30,
+    compressionCacheStats: 30,
     xpAuditLog: 30,
+    compressionRunTelemetry: 30,
     autoCleanupEnabled: true,
   },
   aggregation: {
diff --git a/stryker.conf.json b/stryker.conf.json
index 736aa7a1caa..bc5a1b1ad32 100644
--- a/stryker.conf.json
+++ b/stryker.conf.json
@@ -100,6 +100,7 @@
       "tests/unit/circuit-breaker-failure-kind.test.ts",
       "tests/unit/circuit-breaker-registry-cap.test.ts",
       "tests/unit/circuit-breaker-stream-controller-4602.test.ts",
+      "tests/unit/ddg-circuit-breaker-null-content-6999-7000.test.ts",
       "tests/unit/claude-code-parity.test.ts",
       "tests/unit/claude-effort-suffix-strip.test.ts",
       "tests/unit/claude-oauth-provider.test.ts",
diff --git a/tests/fixtures/router-eval/baseline.ndjson b/tests/fixtures/router-eval/baseline.ndjson
new file mode 100644
index 00000000000..ddc07175fc6
--- /dev/null
+++ b/tests/fixtures/router-eval/baseline.ndjson
@@ -0,0 +1,2 @@
+{"sampleId":"baseline-1","configId":"priority","selectedModel":"gpt-4.1","expectedModel":"gpt-4.1","latencyMs":120,"costUsd":0.004,"success":true,"timestamp":"2026-01-01T00:00:00.000Z"}
+{"sampleId":"baseline-2","configId":"priority","selectedModel":"gpt-4.1","expectedModel":"gpt-4.1","latencyMs":130,"costUsd":0.004,"success":true,"timestamp":"2026-01-01T00:01:00.000Z"}
diff --git a/tests/fixtures/router-eval/candidate.ndjson b/tests/fixtures/router-eval/candidate.ndjson
new file mode 100644
index 00000000000..a6e91603750
--- /dev/null
+++ b/tests/fixtures/router-eval/candidate.ndjson
@@ -0,0 +1,2 @@
+{"sampleId":"candidate-1","configId":"priority","selectedModel":"gpt-4.1","expectedModel":"gpt-4.1","latencyMs":110,"costUsd":0.004,"success":true,"timestamp":"2026-01-01T00:00:00.000Z"}
+{"sampleId":"candidate-2","configId":"priority","selectedModel":"gpt-4.1","expectedModel":"gpt-4.1","latencyMs":125,"costUsd":0.004,"success":true,"timestamp":"2026-01-01T00:01:00.000Z"}
diff --git a/tests/integration/all-statuses-route.test.ts b/tests/integration/all-statuses-route.test.ts
index 5bf72193a40..b3179862f1a 100644
--- a/tests/integration/all-statuses-route.test.ts
+++ b/tests/integration/all-statuses-route.test.ts
@@ -27,12 +27,11 @@ const apiKeysDb = await import("../../src/lib/db/apiKeys.ts");
 const { clearCache, setCached } = await import("../../src/lib/cliTools/batchStatusCache.ts");
 
 // Import the route under test
-const allStatusesRoute = await import(
-  "../../src/app/api/cli-tools/all-statuses/route.ts"
-);
+const allStatusesRoute = await import("../../src/app/api/cli-tools/all-statuses/route.ts");
 
 // Import CLI_TOOLS to know how many tools exist
 const { CLI_TOOLS } = await import("../../src/shared/constants/cliTools.ts");
+const { getCliPrimaryConfigPath } = await import("../../src/shared/services/cliRuntime.ts");
 
 const TOOL_COUNT = Object.keys(CLI_TOOLS).length;
 
@@ -105,14 +104,8 @@ test("happy path: returns status map covering all tools in CLI_TOOLS", async ()
     // Each returned entry should have detection and config fields
     for (const [toolId, entry] of Object.entries(body)) {
       const e = entry as Record<string, unknown>;
-      assert.ok(
-        "detection" in e,
-        `tool ${toolId} missing detection field`
-      );
-      assert.ok(
-        "config" in e,
-        `tool ${toolId} missing config field`
-      );
+      assert.ok("detection" in e, `tool ${toolId} missing detection field`);
+      assert.ok("config" in e, `tool ${toolId} missing config field`);
     }
   } else {
     // If 500 (e.g., runtime detection fails in CI), error body must be sanitized
@@ -190,14 +183,8 @@ test("timeout in 1 tool: others succeed + slot has error field (no full request
           typeof entry.error === "string",
           `tool ${toolId} error should be a string, got ${typeof entry.error}`
         );
-        assert.ok(
-          "detection" in entry,
-          `tool ${toolId} with error should still have detection`
-        );
-        assert.ok(
-          "config" in entry,
-          `tool ${toolId} with error should still have config`
-        );
+        assert.ok("detection" in entry, `tool ${toolId} with error should still have detection`);
+        assert.ok("config" in entry, `tool ${toolId} with error should still have config`);
       }
     }
   }
@@ -248,3 +235,22 @@ test("cache miss: different mtime forces re-execution (cache not used)", async (
   // The entry should exist — fresh execution was performed (no crash)
   assert.ok(toolId in body, `expected ${toolId} after cache miss re-execution`);
 });
+
+test("refresh=true bypasses a matching cached CLI result", async () => {
+  const toolId = "codex";
+  const cachedVersion = "stale-codex-version-from-cache";
+  const configPath = getCliPrimaryConfigPath(toolId);
+  const mtimeMs = configPath && fs.existsSync(configPath) ? fs.statSync(configPath).mtimeMs : 0;
+  setCached(toolId, mtimeMs, {
+    detection: { installed: true, runnable: true, version: cachedVersion },
+    config: { status: "configured" },
+  });
+
+  const response = await allStatusesRoute.GET(
+    new Request("http://localhost/api/cli-tools/all-statuses?refresh=true")
+  );
+
+  assert.equal(response.status, 200);
+  const body = (await response.json()) as Record<string, { detection?: { version?: string } }>;
+  assert.notEqual(body[toolId]?.detection?.version, cachedVersion);
+});
diff --git a/tests/theoldllm-stress.test.ts b/tests/theoldllm-stress.test.ts
index 2125edff5ae..e4b6ab59c6a 100644
--- a/tests/theoldllm-stress.test.ts
+++ b/tests/theoldllm-stress.test.ts
@@ -40,32 +40,30 @@ describe("TheOldLlmExecutor", () => {
     assert.strictEqual(headers["Content-Type"], "application/json");
     assert.ok(
       headers["User-Agent"].includes("Chrome/"),
-      `expected Chrome UA, got ${headers["User-Agent"]}`,
+      `expected Chrome UA, got ${headers["User-Agent"]}`
     );
     assert.ok(
       headers["User-Agent"].includes("Mozilla/5.0"),
-      `expected Mozilla UA, got ${headers["User-Agent"]}`,
+      `expected Mozilla UA, got ${headers["User-Agent"]}`
     );
   });
 
   it("maps model aliases to upstream slugs", () => {
     const cases: Record<string, string> = {
       "gpt-5.4": "GPT_5_4",
-      "GPT_5_3": "GPT_5_3",
-      "gpt_5_2": "GPT_5_2",
+      GPT_5_3: "GPT_5_3",
+      gpt_5_2: "GPT_5_2",
       "gpt-4o": "GPT_4O",
       "claude-4.6-opus": "CLAUDE_4_6_OPUS",
       "claude sonnet 4": "CLAUDE_4_6_SONNET",
-      "claude_haiku_3_5": "CLAUDE_4_5_HAIKU",
+      claude_haiku_3_5: "CLAUDE_4_5_HAIKU",
       "weird-model": "GPT_5_4",
     };
 
-    const transformRequest = (executor as any).transformRequest.bind(
-      executor,
-    ) as (
+    const transformRequest = (executor as any).transformRequest.bind(executor) as (
       model: string,
       body: Record<string, unknown>,
-      stream: boolean,
+      stream: boolean
     ) => Record<string, unknown>;
 
     for (const [model, expected] of Object.entries(cases)) {
@@ -73,7 +71,7 @@ describe("TheOldLlmExecutor", () => {
       assert.strictEqual(
         updated.model,
         expected,
-        `model ${model} mapped to ${expected}, got ${updated.model}`,
+        `model ${model} mapped to ${expected}, got ${updated.model}`
       );
     }
   });
@@ -89,7 +87,7 @@ describe("TheOldLlmExecutor", () => {
           error: () => {},
           debug: () => {},
         }),
-        true,
+        true
       );
 
       globalThis.fetch = async () => makeResponse(401) as any;
@@ -100,7 +98,7 @@ describe("TheOldLlmExecutor", () => {
           error: () => {},
           debug: () => {},
         }),
-        false,
+        false
       );
     } finally {
       globalThis.fetch = originalFetch;
@@ -112,15 +110,10 @@ describe("TheOldLlmExecutor", () => {
     warmTokenCache();
     try {
       let calls = 0;
-      const responses = [
-        () => makeResponse(401, MOCK_ERR),
-        () => makeResponse(200, MOCK_SSE),
-      ];
+      const responses = [() => makeResponse(401, MOCK_ERR), () => makeResponse(200, MOCK_SSE)];
 
       globalThis.fetch = async () =>
-        responses[
-          calls++ < responses.length ? calls - 1 : responses.length - 1
-        ]() as any;
+        responses[calls++ < responses.length ? calls - 1 : responses.length - 1]() as any;
 
       const result = await executor.execute({
         model: "gpt-5.4",
@@ -144,6 +137,45 @@ describe("TheOldLlmExecutor", () => {
     }
   });
 
+  it("does not retry a Vercel egress denial as a stale request token", async () => {
+    const originalFetch = globalThis.fetch;
+    try {
+      let calls = 0;
+      globalThis.fetch = async () => {
+        calls++;
+        return new Response(
+          JSON.stringify({ error: { code: "403", message: "Forbidden", id: "fra1::test" } }),
+          {
+            status: 403,
+            headers: {
+              "content-type": "application/json",
+              "x-vercel-mitigated": "deny",
+            },
+          }
+        );
+      };
+
+      const result = await executor.execute({
+        model: "gpt-5.4",
+        body: { messages: [{ role: "user", content: "ping" }] },
+        stream: true,
+        signal: null,
+        credentials: {},
+        log: { debug() {}, info() {}, warn() {}, error() {} },
+      });
+
+      assert.strictEqual(calls, 1);
+      assert.strictEqual(result.response.status, 403);
+      const json = (await result.response.json()) as {
+        error?: { code?: string; message?: string };
+      };
+      assert.strictEqual(json.error?.code, "THEOLDLLM_VERCEL_MITIGATED");
+      assert.match(json.error?.message || "", /residential.*proxy/i);
+    } finally {
+      globalThis.fetch = originalFetch;
+    }
+  });
+
   it("lets cancellation abort before upstream work", async () => {
     const controller = new AbortController();
     controller.abort(new Error("cancelled"));
@@ -202,7 +234,7 @@ describe("TheOldLlmExecutor", () => {
             warn: () => {},
             error: () => {},
           },
-        }),
+        })
       );
 
       await Promise.all(requests);
diff --git a/tests/unit/account-fallback-lockout-eviction.test.ts b/tests/unit/account-fallback-lockout-eviction.test.ts
new file mode 100644
index 00000000000..68537b58f38
--- /dev/null
+++ b/tests/unit/account-fallback-lockout-eviction.test.ts
@@ -0,0 +1,129 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omr-lockout-eviction-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+process.env.API_KEY_SECRET = "test-lockout-eviction-secret";
+
+const {
+  lockModel,
+  isModelLocked,
+  clearAllModelLockouts,
+  evictModelLockoutOverflow,
+  getModelLockoutSize,
+  MODEL_LOCKOUT_EVICTION_CAP,
+} = await import("../../open-sse/services/accountFallback.ts");
+
+const REASON = "rate_limited";
+// Far from expiring — entry.until is well in the future.
+const ACTIVE_COOLDOWN_MS = 60_000;
+// Negative cooldown backdates `until` into the past (Date.now() + cooldownMs),
+// deterministically producing an already-expired entry without a real sleep.
+const EXPIRED_COOLDOWN_MS = -60_000;
+
+test("eviction removes expired entries beyond the cap", () => {
+  clearAllModelLockouts();
+
+  const cap = MODEL_LOCKOUT_EVICTION_CAP ?? 1000;
+  const extra = 50;
+  const total = cap + extra;
+
+  for (let i = 0; i < total; i++) {
+    lockModel(`evict-test-p-${i}`, `conn-${i}`, `m-${i}`, REASON, EXPIRED_COOLDOWN_MS);
+  }
+
+  assert.equal(getModelLockoutSize(), total);
+
+  evictModelLockoutOverflow();
+
+  // Every entry here is already expired, so nothing is protected — eviction
+  // should shrink exactly back down to the cap.
+  assert.equal(
+    getModelLockoutSize(),
+    cap,
+    "all-expired overflow should be evicted down to exactly the cap"
+  );
+});
+
+test("eviction is idempotent when under cap", () => {
+  clearAllModelLockouts();
+
+  const cap = MODEL_LOCKOUT_EVICTION_CAP ?? 1000;
+  const under = cap - 10;
+  for (let i = 0; i < under; i++) {
+    lockModel(`idemp-p-${i}`, `conn-${i}`, `m-${i}`, REASON, ACTIVE_COOLDOWN_MS);
+  }
+
+  assert.equal(getModelLockoutSize(), under);
+
+  evictModelLockoutOverflow();
+
+  assert.equal(getModelLockoutSize(), under);
+});
+
+test("eviction never removes a still-active lockout, even when the map exceeds the cap purely with active entries", () => {
+  clearAllModelLockouts();
+
+  const cap = MODEL_LOCKOUT_EVICTION_CAP ?? 1000;
+
+  // Lock a "victim" model FIRST — the oldest insertion-order entry, i.e.
+  // exactly what the old (buggy) insertion-order eviction deleted first.
+  lockModel("victim-provider", "victim-conn", "victim-model", REASON, ACTIVE_COOLDOWN_MS);
+  assert.ok(
+    isModelLocked("victim-provider", "victim-conn", "victim-model"),
+    "precondition: victim is locked"
+  );
+
+  // Push the map over the cap using only MORE active (not expired) entries.
+  for (let i = 0; i < cap + 10; i++) {
+    lockModel(`overflow-p-${i}`, `conn-${i}`, `m-${i}`, REASON, ACTIVE_COOLDOWN_MS);
+  }
+  assert.ok(getModelLockoutSize() > cap, "precondition: map exceeds the cap purely with active locks");
+
+  evictModelLockoutOverflow();
+
+  // The core regression (#6923): an active lock must survive eviction
+  // regardless of insertion order, even while the map stays over the
+  // nominal cap — a real, currently-cooling-down model must never be
+  // silently reported as unlocked just because the Map got large.
+  assert.ok(
+    isModelLocked("victim-provider", "victim-conn", "victim-model"),
+    "a still-active lockout must survive eviction even when the map is over cap"
+  );
+});
+
+test("eviction removes expired entries while an active lockout in the same overflow survives", () => {
+  clearAllModelLockouts();
+
+  const cap = MODEL_LOCKOUT_EVICTION_CAP ?? 1000;
+  const extra = 50;
+
+  // Victim locked FIRST (oldest insertion order) with an active cooldown.
+  lockModel("victim2-provider", "victim2-conn", "victim2-model", REASON, ACTIVE_COOLDOWN_MS);
+
+  // Fill past the cap with EXPIRED entries — these should be evicted.
+  for (let i = 0; i < cap + extra; i++) {
+    lockModel(`stale-p-${i}`, `conn-${i}`, `m-${i}`, REASON, EXPIRED_COOLDOWN_MS);
+  }
+
+  const beforeSize = getModelLockoutSize();
+  assert.equal(beforeSize, cap + extra + 1);
+
+  evictModelLockoutOverflow();
+
+  assert.ok(
+    isModelLocked("victim2-provider", "victim2-conn", "victim2-model"),
+    "active victim lockout must survive eviction"
+  );
+  // The oldest entry (the victim) is protected because it's active, so
+  // eviction skips it and instead removes (extra + 1) of the expired
+  // fillers to close the gap — landing exactly back at the cap.
+  assert.equal(
+    getModelLockoutSize(),
+    cap,
+    "expired fillers should be evicted down to exactly the cap, victim included"
+  );
+});
diff --git a/tests/unit/agent-bridge-dns-params-7271.test.ts b/tests/unit/agent-bridge-dns-params-7271.test.ts
new file mode 100644
index 00000000000..013d5ade898
--- /dev/null
+++ b/tests/unit/agent-bridge-dns-params-7271.test.ts
@@ -0,0 +1,37 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { POST } from "../../src/app/api/tools/agent-bridge/agents/[id]/dns/route.ts";
+
+/**
+ * Regression for #7271: Agent Bridge "Start DNS" returned HTTP 404.
+ *
+ * Next.js 16 makes a dynamic route's `params` a Promise. The handler destructured
+ * `id` directly from the (un-awaited) Promise, so `id` was `undefined` and the
+ * agent lookup always missed — producing `404 Unknown agent: undefined` for every
+ * request, including valid agents.
+ *
+ * This test drives the handler with the Next.js 16 params shape (a Promise) and an
+ * unknown agent id. It never reaches the DNS side-effects (the 404 fires before
+ * `addDNSEntry`), so it is a pure, side-effect-free reproduction: the 404 message
+ * must echo the *resolved* id — proving `params` was awaited — not `undefined`.
+ */
+test("POST /agents/[id]/dns awaits params — 404 echoes the real id, not 'undefined'", async () => {
+  const agentId = "definitely-not-a-real-agent-7271";
+  const request = new Request("http://localhost/api/tools/agent-bridge/agents/x/dns", {
+    method: "POST",
+    headers: { "content-type": "application/json" },
+    body: JSON.stringify({ enabled: false }),
+  });
+
+  const res = await POST(request, { params: Promise.resolve({ id: agentId }) });
+  const body = await res.json();
+
+  assert.equal(res.status, 404);
+  // Before the fix, `id` was `undefined`, so the message read "Unknown agent: undefined".
+  assert.equal(body.error.message, `Unknown agent: ${agentId}`);
+  assert.ok(
+    !body.error.message.includes("undefined"),
+    "params must be awaited — message should not contain 'undefined'"
+  );
+});
diff --git a/tests/unit/agentrouter-models-discovery-7016.test.ts b/tests/unit/agentrouter-models-discovery-7016.test.ts
new file mode 100644
index 00000000000..122028b7cfd
--- /dev/null
+++ b/tests/unit/agentrouter-models-discovery-7016.test.ts
@@ -0,0 +1,74 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { PROVIDER_MODELS_CONFIG } from "../../src/app/api/providers/[id]/models/discovery/providerModelsConfig.ts";
+import { CLAUDE_CODE_COMPATIBLE_USER_AGENT } from "../../open-sse/services/claudeCodeCompatible.ts";
+
+// Regression guard for #7016 — AgentRouter's "Import from /models" must hit the
+// live /v1/models endpoint. Previously the provider had no PROVIDER_MODELS_CONFIG
+// entry, so model discovery fell through to the local catalog and import reported
+// "API unavailable — using local catalog". The live request must also carry the
+// Claude Code wire image (the same one the chat path uses) or the gateway WAF
+// rejects it; AgentRouter keeps its own x-api-key auth (#6056).
+
+test("agentrouter has a live /v1/models discovery config", () => {
+  const cfg = PROVIDER_MODELS_CONFIG.agentrouter;
+  assert.ok(cfg, "expected an agentrouter entry in PROVIDER_MODELS_CONFIG");
+  assert.equal(cfg.method, "GET");
+  assert.equal(cfg.url, "https://agentrouter.org/v1/models");
+  assert.equal(typeof cfg.buildHeaders, "function");
+  assert.equal(typeof cfg.parseResponse, "function");
+});
+
+test("agentrouter discovery headers carry the Claude Code wire image + x-api-key", () => {
+  const cfg = PROVIDER_MODELS_CONFIG.agentrouter;
+  const headers = cfg.buildHeaders!("sk-agentrouter");
+
+  // CC wire-image markers (mirrors the chat path — see agentrouter-cc-wire-image.test.ts).
+  assert.equal(headers["User-Agent"], CLAUDE_CODE_COMPATIBLE_USER_AGENT);
+  assert.equal(headers["x-app"], "cli");
+  assert.equal(headers["anthropic-dangerous-direct-browser-access"], "true");
+  assert.ok(headers["anthropic-beta"], "expected the CC anthropic-beta header");
+  assert.ok(headers["X-Stainless-Package-Version"], "expected CC X-Stainless anchors");
+
+  // CRUX: AgentRouter keeps its OWN x-api-key auth (NOT the CC Bearer).
+  assert.equal(headers["x-api-key"], "sk-agentrouter");
+  assert.equal(headers["Authorization"], undefined);
+});
+
+test("agentrouter discovery parseResponse maps an OpenAI-style model list", () => {
+  const cfg = PROVIDER_MODELS_CONFIG.agentrouter;
+  const models = cfg.parseResponse!({
+    object: "list",
+    data: [
+      { id: "claude-opus-4-6", object: "model", owned_by: "anthropic" },
+      { id: "glm-5.1", object: "model", owned_by: "zhipu" },
+    ],
+  });
+  assert.deepEqual(
+    models,
+    [
+      { id: "claude-opus-4-6", object: "model", owned_by: "anthropic" },
+      { id: "glm-5.1", object: "model", owned_by: "zhipu" },
+    ]
+  );
+
+  // Tolerant of the alternate `models` envelope shape too.
+  assert.deepEqual(cfg.parseResponse!({ models: [{ id: "deepseek-v3.2" }] }), [
+    { id: "deepseek-v3.2" },
+  ]);
+});
+
+test("agentrouter discovery headers leak no Authorization variant (case-insensitive)", () => {
+  const cfg = PROVIDER_MODELS_CONFIG.agentrouter;
+  const headers = cfg.buildHeaders!("sk-agentrouter");
+  const hasAuthVariant = Object.keys(headers).some((k) => k.toLowerCase() === "authorization");
+  assert.equal(hasAuthVariant, false, "no Authorization/authorization header must survive");
+});
+
+test("agentrouter parseResponse handles a bare array response (#7016)", () => {
+  const cfg = PROVIDER_MODELS_CONFIG.agentrouter;
+  assert.deepEqual(cfg.parseResponse!([{ id: "claude-sonnet-4-6" }]), [
+    { id: "claude-sonnet-4-6" },
+  ]);
+});
diff --git a/tests/unit/agnes-provider.test.ts b/tests/unit/agnes-provider.test.ts
new file mode 100644
index 00000000000..784703ebff0
--- /dev/null
+++ b/tests/unit/agnes-provider.test.ts
@@ -0,0 +1,70 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+const { APIKEY_PROVIDERS } = await import("../../src/shared/constants/providers.ts");
+const { REGISTRY: providerRegistry } = await import("../../open-sse/config/providerRegistry.ts");
+
+const AGNES_CHAT_URL = "https://apihub.agnes-ai.com/v1/chat/completions";
+
+test("agnes is registered as an API-key provider with complete metadata", () => {
+  const entry = APIKEY_PROVIDERS.agnes;
+  assert.ok(entry, "APIKEY_PROVIDERS.agnes must be defined");
+  assert.equal(entry.id, "agnes");
+  assert.equal(entry.alias, "agnes");
+  assert.equal(entry.name, "Agnes AI");
+  assert.equal(entry.icon, "auto_awesome");
+  assert.equal(entry.color, "#10B981");
+  assert.equal(entry.textIcon, "AG");
+  assert.equal(entry.website, "https://agnes-ai.com");
+  assert.equal(entry.hasFree, true);
+  assert.ok(entry.freeNote, "freeNote must be defined");
+  assert.ok(entry.authHint, "authHint must be defined");
+});
+
+test("agnes registry entry uses OpenAI format with bearer API-key auth", () => {
+  const entry = providerRegistry.agnes;
+  assert.ok(entry, "providerRegistry.agnes must be defined");
+  assert.equal(entry.id, "agnes");
+  assert.equal(entry.format, "openai");
+  assert.equal(entry.executor, "default");
+  assert.equal(entry.authType, "apikey");
+  assert.equal(entry.authHeader, "bearer");
+  assert.equal(entry.baseUrl, AGNES_CHAT_URL);
+});
+
+test("agnes ships two models with correct capabilities", () => {
+  const entry = providerRegistry.agnes;
+  assert.equal(entry.models.length, 2, "must have 2 models");
+
+  const flash2 = entry.models.find((m) => m.id === "agnes-2.0-flash");
+  assert.ok(flash2, "agnes-2.0-flash must be defined");
+  assert.equal(flash2.contextLength, 524288);
+  assert.equal(flash2.maxOutputTokens, 65536);
+  assert.equal(flash2.supportsReasoning, true);
+  assert.equal(flash2.supportsVision, true);
+  assert.equal(flash2.toolCalling, true);
+  assert.equal(flash2.interleavedField, "reasoning_content");
+
+  const flash15 = entry.models.find((m) => m.id === "agnes-1.5-flash");
+  assert.ok(flash15, "agnes-1.5-flash must be defined");
+  assert.equal(flash15.contextLength, 262144);
+  assert.equal(flash15.maxOutputTokens, 65536);
+  assert.equal(flash15.supportsVision, true);
+  assert.equal(flash15.supportsReasoning, undefined, "1.5-flash has no thinking mode");
+  assert.equal(flash15.toolCalling, undefined, "1.5-flash has no documented tool calling");
+});
+
+test("agnes has no collision with zenmux-free sapiens-ai prefixed models", (t) => {
+  const zenmux = providerRegistry["zenmux-free"];
+  if (!zenmux) {
+    t.skip("zenmux-free not registered in this environment");
+    return;
+  }
+  const agnesInZenmux = zenmux.models.filter((m) => m.id.includes("agnes"));
+  for (const m of agnesInZenmux) {
+    assert.ok(
+      m.id.startsWith("sapiens-ai/"),
+      `zenmux agnes model ${m.id} must use sapiens-ai/ prefix to avoid collision`
+    );
+  }
+});
diff --git a/tests/unit/agy-pro-fallback-chain-3786.test.ts b/tests/unit/agy-pro-fallback-chain-3786.test.ts
index 1f76ddc4d1a..1762d6436a0 100644
--- a/tests/unit/agy-pro-fallback-chain-3786.test.ts
+++ b/tests/unit/agy-pro-fallback-chain-3786.test.ts
@@ -190,6 +190,136 @@ test("(#3786) happy path: first id 200 makes exactly ONE upstream call (zero ext
   }
 });
 
+// ---------------------------------------------------------------------------
+// Behavioral: executor catches exceptions and continues the fallback chain
+// ---------------------------------------------------------------------------
+
+test("(#3786) exception on first candidate (timeout) falls through to second candidate returning 200", async () => {
+  const executor = new AntigravityExecutor();
+  const originalFetch = globalThis.fetch;
+  seedAntigravityVersionCache("2026.04.17-test");
+  const modelsTried: string[] = [];
+
+  globalThis.fetch = (async (_url: string, init?: RequestInit) => {
+    const m = envelopeModel(init);
+    modelsTried.push(m);
+    if (m === "gemini-3.1-pro-high") throw new Error("upstream timeout after 30000ms");
+    return makeSuccessSSE();
+  }) as typeof fetch;
+
+  try {
+    const result = await executor.execute({
+      model: "antigravity/gemini-3.1-pro-high",
+      body: { request: { contents: [] } },
+      stream: false,
+      credentials: { accessToken: "token", projectId: "project-1" },
+      log: { debug() {}, warn() {}, info() {} },
+    });
+    const payload = (await result.response.json()) as ChatCompletionPayload;
+
+    assert.equal(result.response.status, 200, "second candidate should succeed after first threw");
+    assert.equal(payload.choices[0].message.content, "OK");
+    // First candidate retried internally (URL-level retries on throw), then second succeeded.
+    assert.equal(modelsTried[0], "gemini-3.1-pro-high", "first call targets first candidate");
+    assert.ok(modelsTried.includes("gemini-pro-agent"), "must eventually try second candidate");
+    assert.ok(
+      modelsTried.lastIndexOf("gemini-3.1-pro-high") < modelsTried.indexOf("gemini-pro-agent"),
+      "first candidate exhausted before second tried"
+    );
+  } finally {
+    globalThis.fetch = originalFetch;
+  }
+});
+
+test("(#3786) all candidates throw exceptions -- error includes 'chain exhausted'", async () => {
+  const executor = new AntigravityExecutor();
+  const originalFetch = globalThis.fetch;
+  seedAntigravityVersionCache("2026.04.17-test");
+  const modelsTried: string[] = [];
+
+  globalThis.fetch = (async (_url: string, init?: RequestInit) => {
+    const m = envelopeModel(init);
+    modelsTried.push(m);
+    throw new Error(`timeout on ${m}`);
+  }) as typeof fetch;
+
+  try {
+    await assert.rejects(
+      () =>
+        executor.execute({
+          model: "antigravity/gemini-3.1-pro-high",
+          body: { request: { contents: [] } },
+          stream: false,
+          credentials: { accessToken: "token", projectId: "project-1" },
+          log: { debug() {}, warn() {}, info() {} },
+        }),
+      (err: Error) => {
+        assert.ok(
+          err.message.includes("chain exhausted"),
+          `error must mention chain exhausted, got: ${err.message}`
+        );
+        assert.ok(
+          err.message.includes("timeout on"),
+          `error must include last error message, got: ${err.message}`
+        );
+        return true;
+      }
+    );
+    // All 3 candidates tried (each retries internally on throw).
+    assert.ok(modelsTried.includes("gemini-3.1-pro-high"), "tried first candidate");
+    assert.ok(modelsTried.includes("gemini-pro-agent"), "tried second candidate");
+    assert.ok(modelsTried.includes("gemini-3-pro-high"), "tried third candidate");
+    // Ordering: all first-candidate attempts before second, etc.
+    const lastFirst = modelsTried.lastIndexOf("gemini-3.1-pro-high");
+    const firstSecond = modelsTried.indexOf("gemini-pro-agent");
+    const lastSecond = modelsTried.lastIndexOf("gemini-pro-agent");
+    const firstThird = modelsTried.indexOf("gemini-3-pro-high");
+    assert.ok(lastFirst < firstSecond, "first candidate exhausted before second tried");
+    assert.ok(lastSecond < firstThird, "second candidate exhausted before third tried");
+  } finally {
+    globalThis.fetch = originalFetch;
+  }
+});
+
+test("(#3786) mixed path: 400 on first, exception on second, 200 on third", async () => {
+  const executor = new AntigravityExecutor();
+  const originalFetch = globalThis.fetch;
+  seedAntigravityVersionCache("2026.04.17-test");
+  const modelsTried: string[] = [];
+
+  globalThis.fetch = (async (_url: string, init?: RequestInit) => {
+    const m = envelopeModel(init);
+    modelsTried.push(m);
+    if (m === "gemini-3.1-pro-high") return make400(m);
+    if (m === "gemini-pro-agent") throw new Error("connection reset");
+    return makeSuccessSSE();
+  }) as typeof fetch;
+
+  try {
+    const result = await executor.execute({
+      model: "antigravity/gemini-3.1-pro-high",
+      body: { request: { contents: [] } },
+      stream: false,
+      credentials: { accessToken: "token", projectId: "project-1" },
+      log: { debug() {}, warn() {}, info() {} },
+    });
+    const payload = (await result.response.json()) as ChatCompletionPayload;
+
+    assert.equal(result.response.status, 200, "third candidate should succeed");
+    assert.equal(payload.choices[0].message.content, "OK");
+    // First candidate (400, no retry) -> second candidate (throws, retried internally) -> third (200).
+    assert.equal(modelsTried[0], "gemini-3.1-pro-high", "first call targets first candidate");
+    assert.ok(modelsTried.includes("gemini-pro-agent"), "second candidate attempted");
+    assert.ok(modelsTried.includes("gemini-3-pro-high"), "third candidate reached");
+    assert.ok(
+      modelsTried.lastIndexOf("gemini-pro-agent") < modelsTried.indexOf("gemini-3-pro-high"),
+      "second candidate exhausted before third tried"
+    );
+  } finally {
+    globalThis.fetch = originalFetch;
+  }
+});
+
 test("(#3786) a non-pro model that 400s does NOT trigger the fallback chain", async () => {
   const executor = new AntigravityExecutor();
   const originalFetch = globalThis.fetch;
@@ -217,3 +347,76 @@ test("(#3786) a non-pro model that 400s does NOT trigger the fallback chain", as
     globalThis.fetch = originalFetch;
   }
 });
+
+test("(#3786) mixed: first 400 + last throws returns firstResult (original 400) instead of throwing", async () => {
+  const executor = new AntigravityExecutor();
+  const originalFetch = globalThis.fetch;
+  seedAntigravityVersionCache("2026.04.17-test");
+  const modelsTried: string[] = [];
+
+  globalThis.fetch = (async (_url: string, init?: RequestInit) => {
+    const m = envelopeModel(init);
+    modelsTried.push(m);
+    if (m === "gemini-3.1-pro-high") return make400(m);
+    // Second and third candidates throw
+    throw new Error(`connection reset on ${m}`);
+  }) as typeof fetch;
+
+  try {
+    const result = await executor.execute({
+      model: "antigravity/gemini-3.1-pro-high",
+      body: { request: { contents: [] } },
+      stream: false,
+      credentials: { accessToken: "token", projectId: "project-1" },
+      log: { debug() {}, warn() {}, info() {} },
+    });
+
+    // Returns the original 400 from firstResult, not a thrown error.
+    assert.equal(result.response.status, 400, "should return first candidate's 400");
+    const payload = (await result.response.json()) as ErrorPayload;
+    assert.ok(payload.error, "must carry an error object");
+    assert.equal(typeof payload.error.message, "string");
+    assert.ok(!payload.error.message.includes("at /"), "no raw stack trace (hard rule #12)");
+    // All candidates tried (executeOnce retries internally, so duplicates are expected).
+    assert.ok(modelsTried.includes("gemini-3.1-pro-high"), "first candidate tried");
+    assert.ok(modelsTried.includes("gemini-pro-agent"), "second candidate tried");
+    assert.ok(modelsTried.includes("gemini-3-pro-high"), "third candidate tried");
+  } finally {
+    globalThis.fetch = originalFetch;
+  }
+});
+
+test("(#3786) AbortError from standard Error (not DOMException) propagates immediately", async () => {
+  const executor = new AntigravityExecutor();
+  const originalFetch = globalThis.fetch;
+  seedAntigravityVersionCache("2026.04.17-test");
+  // Signal is NOT aborted -- this exercises the new Error.name === "AbortError" path.
+  const controller = new AbortController();
+
+  globalThis.fetch = (async (_url: string, _init?: RequestInit) => {
+    // Simulate a polyfill/test env that throws standard Error with name AbortError
+    const err = new Error("The operation was aborted");
+    err.name = "AbortError";
+    throw err;
+  }) as typeof fetch;
+
+  try {
+    await assert.rejects(
+      () =>
+        executor.execute({
+          model: "antigravity/gemini-3.1-pro-high",
+          body: { request: { contents: [] } },
+          stream: false,
+          credentials: { accessToken: "token", projectId: "project-1" },
+          signal: controller.signal,
+          log: { debug() {}, warn() {}, info() {} },
+        }),
+      (err: Error) => {
+        assert.equal(err.name, "AbortError", "must propagate as AbortError");
+        return true;
+      }
+    );
+  } finally {
+    globalThis.fetch = originalFetch;
+  }
+});
diff --git a/tests/unit/antigravity-model-aliases.test.ts b/tests/unit/antigravity-model-aliases.test.ts
index 718250c8ecf..9c80666ff57 100644
--- a/tests/unit/antigravity-model-aliases.test.ts
+++ b/tests/unit/antigravity-model-aliases.test.ts
@@ -99,21 +99,27 @@ test("isUserCallableAntigravityModelId only allows public chat-capable model IDs
 
 test("ANTIGRAVITY_PUBLIC_MODELS exposes captured Antigravity 2.0.1 names and capabilities", () => {
   // #3184: Claude is exposed in the antigravity catalog (same backend as `agy`, verified).
+  // #7129: Opus 4.6, Sonnet 4.6, and Sonnet 5 graduated to a 1M-token context window at GA
+  // (Anthropic docs, platform.claude.com/docs/en/build-with-claude/context-windows: "Claude
+  // Opus 4.8, Claude Opus 4.7, Claude Opus 4.6, Claude Sonnet 5, and Claude Sonnet 4.6 have a
+  // 1M-token context window ... on the Claude API, Amazon Bedrock, Google Cloud, and Microsoft
+  // Foundry" — Google Cloud coverage extends to the Antigravity-hosted ids exercised here).
   assert.deepEqual(getPublicModel("claude-opus-4-6-thinking"), {
     id: "claude-opus-4-6-thinking",
     name: "Claude Opus 4.6 (Thinking)",
-    contextLength: 200000,
+    contextLength: 1048576,
     maxOutputTokens: 65536,
     supportsReasoning: true,
     supportsVision: true,
     toolCalling: true,
   });
   assert.equal(getPublicModel("claude-sonnet-4-6").name, "Claude Sonnet 4.6 (Thinking)");
+  assert.equal(getPublicModel("claude-sonnet-4-6").contextLength, 1048576);
   // claude-sonnet-5 was added to the Antigravity catalog alongside the existing Claude entries.
   assert.deepEqual(getPublicModel("claude-sonnet-5"), {
     id: "claude-sonnet-5",
     name: "Claude Sonnet 5 (Thinking)",
-    contextLength: 200000,
+    contextLength: 1048576,
     maxOutputTokens: 65536,
     supportsReasoning: true,
     supportsVision: true,
diff --git a/tests/unit/antigravity-server-side-tools-6914.test.ts b/tests/unit/antigravity-server-side-tools-6914.test.ts
new file mode 100644
index 00000000000..73dfd18d7ba
--- /dev/null
+++ b/tests/unit/antigravity-server-side-tools-6914.test.ts
@@ -0,0 +1,43 @@
+// #6914: server-side tool invocations must be requested from Antigravity whenever the
+// caller sends tools, and must NOT be forced when the request carries none. Lives in
+// its own file (not executor-antigravity.test.ts) because that suite is frozen at the
+// test-file-size cap.
+import { test } from "node:test";
+import assert from "node:assert/strict";
+
+import { AntigravityExecutor } from "../../open-sse/executors/antigravity.ts";
+
+test("AntigravityExecutor.transformRequest includes includeServerSideToolInvocations when tools are present", async () => {
+  const executor = new AntigravityExecutor();
+  const body = {
+    request: {
+      contents: [{ role: "user", parts: [{ text: "Hello" }] }],
+      tools: [{ functionDeclarations: [{ name: "search" }] }],
+    },
+  };
+
+  const result = await executor.transformRequest("antigravity/gemini-3.1-pro", body, true, {
+    projectId: "project-1",
+  });
+
+  if (result instanceof Response) throw new Error("Unexpected Response from transformRequest");
+  assert.deepEqual(result.request.toolConfig, {
+    functionCallingConfig: { mode: "VALIDATED", includeServerSideToolInvocations: true },
+  });
+});
+
+test("AntigravityExecutor.transformRequest does not include includeServerSideToolInvocations when no tools", async () => {
+  const executor = new AntigravityExecutor();
+  const body = {
+    request: {
+      contents: [{ role: "user", parts: [{ text: "Hello" }] }],
+    },
+  };
+
+  const result = await executor.transformRequest("antigravity/gemini-3.1-pro", body, true, {
+    projectId: "project-1",
+  });
+
+  if (result instanceof Response) throw new Error("Unexpected Response from transformRequest");
+  assert.equal(result.request.toolConfig, undefined);
+});
diff --git a/tests/unit/api/compression/rtk-toml-import-route.test.ts b/tests/unit/api/compression/rtk-toml-import-route.test.ts
new file mode 100644
index 00000000000..959e43453a6
--- /dev/null
+++ b/tests/unit/api/compression/rtk-toml-import-route.test.ts
@@ -0,0 +1,125 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+import { makeManagementSessionRequest } from "../../../helpers/managementSession.ts";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-route-"));
+const ORIGINAL_DATA_DIR = process.env.DATA_DIR;
+const ORIGINAL_INITIAL_PASSWORD = process.env.INITIAL_PASSWORD;
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const core = await import("../../../../src/lib/db/core.ts");
+const settingsDb = await import("../../../../src/lib/db/settings.ts");
+const importRoute = await import("../../../../src/app/api/context/rtk/import/route.ts");
+
+const SAMPLE = `schema_version = 1
+[filters.route-test]
+description = "Route test"
+match_command = "^route-test"
+strip_lines_matching = ["^noise"]
+
+[[tests.route-test]]
+name = "removes noise"
+input = "noise\\nkept"
+expected = "kept"
+`;
+
+async function reset() {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+  fs.mkdirSync(TEST_DATA_DIR, { recursive: true });
+  delete process.env.INITIAL_PASSWORD;
+}
+
+async function post(body: unknown): Promise<Response> {
+  return importRoute.POST(
+    await makeManagementSessionRequest("http://localhost/api/context/rtk/import", {
+      method: "POST",
+      body,
+    })
+  );
+}
+
+test.beforeEach(reset);
+
+test.after(() => {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+  if (ORIGINAL_DATA_DIR === undefined) delete process.env.DATA_DIR;
+  else process.env.DATA_DIR = ORIGINAL_DATA_DIR;
+  if (ORIGINAL_INITIAL_PASSWORD === undefined) delete process.env.INITIAL_PASSWORD;
+  else process.env.INITIAL_PASSWORD = ORIGINAL_INITIAL_PASSWORD;
+});
+
+test("POST requires management authentication when login is configured", async () => {
+  process.env.INITIAL_PASSWORD = "rtk-toml-route-password";
+  await settingsDb.updateSettings({ requireLogin: true });
+
+  const response = await importRoute.POST(
+    new Request("http://localhost/api/context/rtk/import", {
+      method: "POST",
+      headers: { "content-type": "application/json" },
+      body: JSON.stringify({ action: "validate", content: SAMPLE }),
+    })
+  );
+
+  assert.equal(response.status, 401);
+});
+
+test("validate returns tests and does not write filters.toml", async () => {
+  const response = await post({ action: "validate", content: SAMPLE });
+  const body = await response.json();
+
+  assert.equal(response.status, 200);
+  assert.equal(body.passed, true);
+  assert.equal(body.filters[0].id, "route-test");
+  assert.equal(body.outcomes[0].passed, true);
+  assert.equal(fs.existsSync(path.join(TEST_DATA_DIR, "rtk", "filters.toml")), false);
+});
+
+test("install writes the global file, refuses implicit overwrite, and creates a backup", async () => {
+  const first = await post({ action: "install", content: SAMPLE });
+  assert.equal(first.status, 200);
+  assert.equal(fs.readFileSync(path.join(TEST_DATA_DIR, "rtk", "filters.toml"), "utf8"), SAMPLE);
+
+  const refused = await post({ action: "install", content: SAMPLE });
+  assert.equal(refused.status, 400);
+
+  const replacement = SAMPLE.replace("Route test", "Replacement");
+  const replaced = await post({ action: "install", content: replacement, overwrite: true });
+  const replacedBody = await replaced.json();
+  assert.equal(replaced.status, 200);
+  assert.equal(replacedBody.backupCreated, true);
+  assert.equal(
+    fs.readFileSync(path.join(TEST_DATA_DIR, "rtk", "filters.toml.bak"), "utf8"),
+    SAMPLE
+  );
+});
+
+test("invalid TOML, unsafe regex, failing tests, and oversized input return safe 400 errors", async () => {
+  const inputs = [
+    "not = [valid",
+    `schema_version = 1\n[filters.bad]\nmatch_command = "(a+)+$"`,
+    `schema_version = 1
+[filters.bad]
+match_command = "^bad"
+strip_lines_matching = ["^noise"]
+[[tests.bad]]
+name = "fails"
+input = "noise\\nkept"
+expected = "wrong"`,
+    "x".repeat(1024 * 1024 + 1),
+  ];
+
+  for (const content of inputs) {
+    const response = await post({ action: "install", content });
+    const body = await response.json();
+    assert.equal(response.status, 400);
+    assert.equal(typeof body.error?.message, "string");
+    assert.ok(!body.error.message.includes(" at "));
+    assert.ok(!body.error.message.includes(process.cwd()));
+  }
+});
diff --git a/tests/unit/api/v1/provider-plugin-manifest-route.test.ts b/tests/unit/api/v1/provider-plugin-manifest-route.test.ts
index 9af7baa46d5..e71f6607822 100644
--- a/tests/unit/api/v1/provider-plugin-manifest-route.test.ts
+++ b/tests/unit/api/v1/provider-plugin-manifest-route.test.ts
@@ -6,11 +6,13 @@ import {
   OPTIONS,
 } from "../../../../src/app/api/v1/provider-plugin-manifest/route.ts";
 
-test("provider plugin manifest route returns JSON-safe manifest", async () => {
-  const response = await GET();
+test("provider plugin manifest returns a stable ETag with its cache policy", async () => {
+  const response = await GET(new Request("http://localhost/api/v1/provider-plugin-manifest"));
   const body = await response.json();
 
   assert.equal(response.status, 200);
+  assert.equal(response.headers.get("Cache-Control"), "public, max-age=60");
+  assert.match(response.headers.get("ETag") ?? "", /^"[A-Za-z0-9_-]+"$/);
   assert.equal(response.headers.get("Content-Type"), "application/json");
   assert.equal(body.schemaVersion, 1);
   assert.equal(body.generatedFrom, "open-sse/config/providers");
@@ -28,3 +30,31 @@ test("provider plugin manifest route handles CORS preflight", async () => {
   assert.equal(response.headers.get("Access-Control-Allow-Methods"), "GET, OPTIONS");
   assert.equal(response.headers.get("Access-Control-Allow-Headers"), "*");
 });
+
+test("provider plugin manifest supports conditional sidecar refreshes", async () => {
+  const initial = await GET(new Request("http://localhost/api/v1/provider-plugin-manifest"));
+  const etag = initial.headers.get("ETag");
+
+  const response = await GET(
+    new Request("http://localhost/api/v1/provider-plugin-manifest", {
+      headers: { "If-None-Match": etag ?? "" },
+    })
+  );
+
+  assert.equal(response.status, 304);
+  assert.equal(response.headers.get("ETag"), etag);
+  assert.equal(await response.text(), "");
+});
+
+test("provider plugin manifest accepts weak conditional validators", async () => {
+  const initial = await GET(new Request("http://localhost/api/v1/provider-plugin-manifest"));
+  const etag = initial.headers.get("ETag");
+
+  const response = await GET(
+    new Request("http://localhost/api/v1/provider-plugin-manifest", {
+      headers: { "If-None-Match": `W/${etag}` },
+    })
+  );
+
+  assert.equal(response.status, 304);
+});
diff --git a/tests/unit/apikey-connection-health-check.test.ts b/tests/unit/apikey-connection-health-check.test.ts
index 7f5b0b0d8e7..582f0ad50f1 100644
--- a/tests/unit/apikey-connection-health-check.test.ts
+++ b/tests/unit/apikey-connection-health-check.test.ts
@@ -171,3 +171,77 @@ test("connection with both apiKey and refreshToken: refresh path is tried", asyn
     "dual-auth connection with stale refresh token should be expired (refresh path takes precedence)"
   );
 });
+
+test("sweep processes all connections with stagger + jitter delay", async () => {
+  await resetStorage();
+
+  // Create multiple connections; set isActive=false so checkConnection
+  // returns immediately at the !conn.isActive guard without OAuth calls.
+  const c1 = await providersDb.createProviderConnection({
+    provider: "openai",
+    authType: "oauth",
+    name: "Stagger Test 1",
+    email: "t1@example.com",
+    refreshToken: "test-rt",
+    isActive: false,
+  });
+  const c2 = await providersDb.createProviderConnection({
+    provider: "openai",
+    authType: "oauth",
+    name: "Stagger Test 2",
+    email: "t2@example.com",
+    refreshToken: "test-rt",
+    isActive: false,
+  });
+  const c3 = await providersDb.createProviderConnection({
+    provider: "openai",
+    authType: "oauth",
+    name: "Stagger Test 3",
+    email: "t3@example.com",
+    refreshToken: "test-rt",
+    isActive: false,
+  });
+
+  // Clear any health-check skip config
+  const origSetting = process.env.HEALTHCHECK_SKIP_PROVIDERS;
+  const origJitterMin = process.env.HEALTHCHECK_JITTER_MIN_MS;
+  const origJitterMax = process.env.HEALTHCHECK_JITTER_MAX_MS;
+  delete process.env.HEALTHCHECK_SKIP_PROVIDERS;
+  process.env.HEALTHCHECK_STAGGER_MS = "1";
+  process.env.HEALTHCHECK_JITTER_MIN_MS = "100";
+  process.env.HEALTHCHECK_JITTER_MAX_MS = "100"; // fixed jitter = deterministic
+
+  try {
+    // Import sweep — exported for testing from tokenHealthCheck
+    const { sweep } = await import("../../src/lib/tokenHealthCheck.ts");
+
+    const start = Date.now();
+    await sweep();
+    const elapsed = Date.now() - start;
+
+    // 3 connections -> 2 gaps. Each gap waits
+    // HEALTHCHECK_STAGGER_MS (1ms) + HEALTHCHECK_JITTER_MIN_MS (100ms) = 101ms.
+    // Without jitter: 2 * 1ms = ~2ms. With jitter: 2 * 101ms = ~202ms.
+    // The assert proves the jitter is actually applied.
+    assert.ok(
+      elapsed >= 190,
+      `sweep took ${elapsed}ms — expected >= 190ms with jitter applied (jitter-free baseline would be ~2ms)`
+    );
+
+    // Verify all connections still exist (sweep didn't error out mid-loop)
+    const reloaded1 = await providersDb.getProviderConnectionById(c1.id);
+    const reloaded2 = await providersDb.getProviderConnectionById(c2.id);
+    const reloaded3 = await providersDb.getProviderConnectionById(c3.id);
+    assert.ok(reloaded1, "connection 1 should still exist");
+    assert.ok(reloaded2, "connection 2 should still exist");
+    assert.ok(reloaded3, "connection 3 should still exist");
+  } finally {
+    if (origSetting !== undefined) process.env.HEALTHCHECK_SKIP_PROVIDERS = origSetting;
+    else delete process.env.HEALTHCHECK_SKIP_PROVIDERS;
+    if (origJitterMin !== undefined) process.env.HEALTHCHECK_JITTER_MIN_MS = origJitterMin;
+    else delete process.env.HEALTHCHECK_JITTER_MIN_MS;
+    if (origJitterMax !== undefined) process.env.HEALTHCHECK_JITTER_MAX_MS = origJitterMax;
+    else delete process.env.HEALTHCHECK_JITTER_MAX_MS;
+    delete process.env.HEALTHCHECK_STAGGER_MS;
+  }
+});
diff --git a/tests/unit/auggie-executor.test.ts b/tests/unit/auggie-executor.test.ts
index c5418ac546a..f8fcd25ea5e 100644
--- a/tests/unit/auggie-executor.test.ts
+++ b/tests/unit/auggie-executor.test.ts
@@ -14,9 +14,8 @@ import fs from "node:fs";
 import os from "node:os";
 import path from "node:path";
 
-const { AuggieExecutor, buildAuggiePrompt, resolveAuggieBin, resolveAuggieModel } = await import(
-  "@omniroute/open-sse/executors/auggie"
-);
+const { AuggieExecutor, buildAuggiePrompt, resolveAuggieBin, resolveAuggieModel } =
+  await import("@omniroute/open-sse/executors/auggie");
 
 const TMP_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-auggie-test-"));
 
@@ -27,7 +26,6 @@ function writeFakeBin(name: string, script: string): string {
   return p;
 }
 
-
 async function readSseEvents(response: Response): Promise<Record<string, unknown>[]> {
   const text = await response.text();
   const events: Record<string, unknown>[] = [];
@@ -61,7 +59,13 @@ test("buildAuggiePrompt flattens system/user/assistant turns with role tags", ()
 
 test("buildAuggiePrompt flattens array-shaped content blocks and skips empty turns", () => {
   const prompt = buildAuggiePrompt([
-    { role: "user", content: [{ type: "text", text: "part1 " }, { type: "text", text: "part2" }] },
+    {
+      role: "user",
+      content: [
+        { type: "text", text: "part1 " },
+        { type: "text", text: "part2" },
+      ],
+    },
     { role: "user", content: "" },
     { role: "user", content: [] },
   ]);
@@ -93,7 +97,7 @@ test("execute() surfaces a sanitized 'CLI not found' error on ENOENT (streaming)
   try {
     const executor = new AuggieExecutor();
     const { response } = await executor.execute({
-      model: "claude-sonnet-4.6",
+      model: "sonnet4.6",
       body: { messages: [{ role: "user", content: "hi" }] },
       stream: true,
       credentials: {} as never,
@@ -120,7 +124,7 @@ test("execute() surfaces a sanitized 'CLI not found' error on ENOENT (non-stream
   try {
     const executor = new AuggieExecutor();
     const { response } = await executor.execute({
-      model: "claude-sonnet-4.6",
+      model: "sonnet4.6",
       body: { messages: [{ role: "user", content: "hi" }] },
       stream: false,
       credentials: {} as never,
@@ -138,13 +142,16 @@ test("execute() surfaces a sanitized 'CLI not found' error on ENOENT (non-stream
 // ─── execute(): non-zero exit → sanitized error ────────────────────────────
 
 test("execute() surfaces a sanitized error when the CLI exits non-zero (streaming)", async () => {
-  const bin = writeFakeBin("fake-auggie-fail.sh", 'echo "boom at /home/attacker/secret.ts:42" 1>&2\nexit 3');
+  const bin = writeFakeBin(
+    "fake-auggie-fail.sh",
+    'echo "boom at /home/attacker/secret.ts:42" 1>&2\nexit 3'
+  );
   const prevBin = process.env.AUGGIE_BIN;
   process.env.AUGGIE_BIN = bin;
   try {
     const executor = new AuggieExecutor();
     const { response } = await executor.execute({
-      model: "claude-sonnet-4.6",
+      model: "sonnet4.6",
       body: { messages: [{ role: "user", content: "hi" }] },
       stream: true,
       credentials: {} as never,
@@ -168,7 +175,7 @@ test("execute() surfaces a sanitized error when the CLI exits non-zero (non-stre
   try {
     const executor = new AuggieExecutor();
     const { response } = await executor.execute({
-      model: "claude-sonnet-4.6",
+      model: "sonnet4.6",
       body: { messages: [{ role: "user", content: "hi" }] },
       stream: false,
       credentials: {} as never,
@@ -191,7 +198,7 @@ test("execute() with stream=true returns SSE deltas + [DONE]", async () => {
   try {
     const executor = new AuggieExecutor();
     const { response } = await executor.execute({
-      model: "claude-sonnet-4.6",
+      model: "sonnet4.6",
       body: { messages: [{ role: "user", content: "say hi" }] },
       stream: true,
       credentials: {} as never,
@@ -221,7 +228,7 @@ test("execute() with stream=false returns a single chat.completion JSON body", a
   try {
     const executor = new AuggieExecutor();
     const { response } = await executor.execute({
-      model: "claude-sonnet-4.6",
+      model: "sonnet4.6",
       body: { messages: [{ role: "user", content: "say hi" }] },
       stream: false,
       credentials: {} as never,
@@ -249,8 +256,41 @@ test("resolveAuggieModel defaults to a real allowlisted model when unset", () =>
 });
 
 test("resolveAuggieModel accepts a known registry model id verbatim", () => {
-  const result = resolveAuggieModel("claude-haiku-4.5");
-  assert.deepEqual(result, { ok: true, model: "claude-haiku-4.5" });
+  const result = resolveAuggieModel("haiku4.5");
+  assert.deepEqual(result, { ok: true, model: "haiku4.5" });
+});
+
+// ─── resolveAuggieModel: pre-v0.32.0 alias back-compat (#7032) ────────────
+// The v0.32.0 registry rename dropped ~10 old model IDs outright. Saved combos
+// created before the rename still reference those old IDs, so resolveAuggieModel
+// must transparently map each one to its v0.32.0 replacement instead of erroring
+// with "Unknown Auggie model".
+
+test("resolveAuggieModel resolves a pre-v0.32.0 saved model id via the alias map (#7032)", () => {
+  const result = resolveAuggieModel("claude-sonnet-4.6");
+  assert.deepEqual(result, { ok: true, model: "sonnet4.6" });
+});
+
+test("resolveAuggieModel resolves every pre-v0.32.0 alias to an allowlisted v0.32.0 id (#7032)", () => {
+  const OLD_TO_NEW: Record<string, string> = {
+    "claude-sonnet-4.6": "sonnet4.6",
+    "claude-sonnet-4.6-thinking": "sonnet4.6",
+    "claude-opus-4.6": "opus4.6",
+    "claude-haiku-4.5": "haiku4.5",
+    "gemini-3.1-pro": "gemini-3.1-pro-preview",
+    "gemini-3.0-flash": "gemini-3.1-pro-preview",
+    "gpt-5.5-high": "gpt5.5",
+    "gpt-5.5-medium": "gpt5.5",
+    "gpt-5.4-high": "gpt5.4",
+    "gpt-5.4-medium": "gpt5.4",
+  };
+  for (const [oldId, newId] of Object.entries(OLD_TO_NEW)) {
+    assert.deepEqual(
+      resolveAuggieModel(oldId),
+      { ok: true, model: newId },
+      `${oldId} should resolve to ${newId}`
+    );
+  }
 });
 
 // ─── execute(): model allowlist (argument-injection defense) ──────────────
@@ -314,7 +354,7 @@ test("execute() spawns with a valid allowlisted model, no shell, and a '--' argv
   try {
     const executor = new AuggieExecutor();
     const { response } = await executor.execute({
-      model: "claude-opus-4.6",
+      model: "opus4.6",
       body: { messages: [{ role: "user", content: "hi" }] },
       stream: false,
       credentials: {} as never,
@@ -323,7 +363,7 @@ test("execute() spawns with a valid allowlisted model, no shell, and a '--' argv
     const body = await response.json();
     assert.equal(body.choices[0].message.content, "ok");
     const argv = fs.readFileSync(argvFile, "utf8").trim().split("\n");
-    assert.deepEqual(argv, ["--print", "--quiet", "--model", "claude-opus-4.6", "--"]);
+    assert.deepEqual(argv, ["--print", "--quiet", "--model", "opus4.6", "--"]);
   } finally {
     if (prevBin === undefined) delete process.env.AUGGIE_BIN;
     else process.env.AUGGIE_BIN = prevBin;
@@ -342,7 +382,7 @@ test("execute() aborts a long-running CLI process instead of hanging (streaming)
     const executor = new AuggieExecutor();
     const controller = new AbortController();
     const { response } = await executor.execute({
-      model: "claude-sonnet-4.6",
+      model: "sonnet4.6",
       body: { messages: [{ role: "user", content: "hi" }] },
       stream: true,
       credentials: {} as never,
@@ -378,3 +418,99 @@ test("buildUrl/buildHeaders/transformRequest match the CLI-passthrough shape", (
   assert.deepEqual(executor.buildHeaders(), {});
   assert.equal(executor.transformRequest(), null);
 });
+
+// ─── initAuggieModels / resolveAuggieModel auto-fetch ────────────────────
+
+test("initAuggieModels discovers models from `auggie model list` output", async () => {
+  const { initAuggieModels, resolveAuggieModel, __resetAuggieModels } =
+    await import("@omniroute/open-sse/executors/auggie");
+  __resetAuggieModels(); // wipe stale cache from previous tests
+  const bin = writeFakeBin(
+    "fake-auggie-list.sh",
+    `printf '[new-model-alpha]\\nignored\\n[new-model-beta]'`
+  );
+  const prevBin = process.env.AUGGIE_BIN;
+  process.env.AUGGIE_BIN = bin;
+  try {
+    await initAuggieModels();
+    const r = resolveAuggieModel("new-model-alpha");
+    assert.deepEqual(r, { ok: true, model: "new-model-alpha" });
+  } finally {
+    if (prevBin === undefined) delete process.env.AUGGIE_BIN;
+    else process.env.AUGGIE_BIN = prevBin;
+    __resetAuggieModels(); // restore for subsequent tests
+  }
+});
+
+test("execute() spawns a model discovered via auto-fetch", async () => {
+  const { __resetAuggieModels } = await import("@omniroute/open-sse/executors/auggie");
+  __resetAuggieModels();
+  const bin = writeFakeBin(
+    "fake-auggie-autofetch.sh",
+    [
+      `case "$*" in`,
+      `  *model\\ list*)`,
+      `    printf '[auto-discovered-model]'`,
+      `    exit 0`,
+      `    ;;`,
+      `  *)`,
+      `    printf 'auto-ok'`,
+      `    ;;`,
+      `esac`,
+    ].join("\n")
+  );
+  const prevBin = process.env.AUGGIE_BIN;
+  process.env.AUGGIE_BIN = bin;
+  try {
+    const executor = new AuggieExecutor();
+    const { response } = await executor.execute({
+      model: "auto-discovered-model",
+      body: { messages: [{ role: "user", content: "hi" }] },
+      stream: false,
+      credentials: {} as never,
+    });
+    assert.equal(response.status, 200);
+    const body = await response.json();
+    assert.equal(body.choices[0].message.content, "auto-ok");
+  } finally {
+    if (prevBin === undefined) delete process.env.AUGGIE_BIN;
+    else process.env.AUGGIE_BIN = prevBin;
+    __resetAuggieModels();
+  }
+});
+
+test("initAuggieModels failure does not block execute() for a known model", async () => {
+  const { __resetAuggieModels } = await import("@omniroute/open-sse/executors/auggie");
+  __resetAuggieModels();
+  const bin = writeFakeBin(
+    "fake-auggie-crash.sh",
+    [
+      `case "$*" in`,
+      `  *model\\ list*)`,
+      `    exit 1  # crash on model list so init fails`,
+      `    ;;`,
+      `  *)`,
+      `    printf 'known-model-ok'`,
+      `    ;;`,
+      `esac`,
+    ].join("\n")
+  );
+  const prevBin = process.env.AUGGIE_BIN;
+  process.env.AUGGIE_BIN = bin;
+  try {
+    const executor = new AuggieExecutor();
+    const { response } = await executor.execute({
+      model: "haiku4.5", // static-registry model, should work regardless
+      body: { messages: [{ role: "user", content: "hi" }] },
+      stream: false,
+      credentials: {} as never,
+    });
+    assert.equal(response.status, 200);
+    const body = await response.json();
+    assert.equal(body.choices[0].message.content, "known-model-ok");
+  } finally {
+    if (prevBin === undefined) delete process.env.AUGGIE_BIN;
+    else process.env.AUGGIE_BIN = prevBin;
+    __resetAuggieModels();
+  }
+});
diff --git a/tests/unit/auto-combo-context-advertising.test.ts b/tests/unit/auto-combo-context-advertising.test.ts
index e316cdae523..390b1265e7c 100644
--- a/tests/unit/auto-combo-context-advertising.test.ts
+++ b/tests/unit/auto-combo-context-advertising.test.ts
@@ -64,7 +64,8 @@ test("computeAdvertisedLimits returns MAX of candidates' known context windows",
     "virtualFactory should export computeAdvertisedLimits()"
   );
 
-  // gemini has registry defaultContextLength=1048576; claude has 200000.
+  // gemini has registry defaultContextLength=1048576; claude-sonnet-4-6 has 1000000 (#7129:
+  // 1M GA per Anthropic docs) -- gemini's binary-1M window still wins as the MAX.
   const result = computeAdvertisedLimits([
     { provider: "claude", model: "claude-sonnet-4-6" },
     { provider: "gemini", model: "gemini-2.5-pro" },
@@ -171,7 +172,10 @@ test("resolveComboContextLimit regression: claude target must not be compressed
     model: "claude-sonnet-4-6",
     comboTargetLimits: [8000],
   });
-  assert.equal(result.limit, 200000);
+  // #7129: claude-sonnet-4-6's own registry limit is 1M GA (was 200000) -- the point of this
+  // regression test is that the target's OWN limit wins over the 8k sibling, not the specific
+  // magic number, so the expectation tracks the registry's current (correct) value.
+  assert.equal(result.limit, 1000000);
   assert.equal(result.source, "target");
 });
 
diff --git a/tests/unit/autoCombo/provider-family-combos.test.ts b/tests/unit/autoCombo/provider-family-combos.test.ts
index 54a68612279..27a30d8e4fb 100644
--- a/tests/unit/autoCombo/provider-family-combos.test.ts
+++ b/tests/unit/autoCombo/provider-family-combos.test.ts
@@ -123,7 +123,12 @@ describe("auto/<family> materialization (#6453)", () => {
     assert.equal(combo.id, "auto/glm");
     assert.equal(combo.strategy, "auto");
     const providerIds = combo.models.map((m) => m.providerId).sort();
-    assert.deepEqual(providerIds, ["glm", "zai"]);
+    // Includes the always-on `auggie` no-auth candidate: its registry (v0.32.0 CLI
+    // model ids) advertises a literal "glm-5.2" model, and — same as the
+    // "degrades gracefully" test below documents for opencode/minimax — a
+    // no-auth backend that genuinely serves a family model IS a legitimate
+    // member of the family pool, not just credentialed provider_connections rows.
+    assert.deepEqual(providerIds, ["auggie", "glm", "zai"]);
     assert.ok(combo.models.every((m) => m.model.endsWith("glm-5.2")));
   });
 
diff --git a/tests/unit/base-executor-sanitize-effort.test.ts b/tests/unit/base-executor-sanitize-effort.test.ts
index cfa62972013..943e09b95a4 100644
--- a/tests/unit/base-executor-sanitize-effort.test.ts
+++ b/tests/unit/base-executor-sanitize-effort.test.ts
@@ -543,3 +543,38 @@ test("sanitizeReasoningEffortForProvider: opencode-go with non-DeepSeek model st
   assert.notEqual(result, body);
   assert.equal((result as any).reasoning_effort, "xhigh");
 });
+
+test("sanitizeReasoningEffortForProvider: #7044 output_config.effort (Claude native) xhigh is downgraded, not bypassed", () => {
+  const log = makeLog();
+  const body = {
+    model: "claude-opus-4-6",
+    output_config: { effort: "xhigh" },
+    messages: [{ role: "user", content: "hi" }],
+  };
+  const result = sanitizeReasoningEffortForProvider(body, "claude", "claude-opus-4-6", log);
+  assert.notEqual(result, body, "must return a new object when mutating");
+  assert.equal(
+    (result as any).output_config.effort,
+    "high",
+    "xhigh downgraded to high on the output_config carrier"
+  );
+  assert.ok(
+    !("reasoning_effort" in (result as any)),
+    "no spurious reasoning_effort injected when only output_config was present"
+  );
+  assert.ok(
+    log.messages.some(([tag, m]) => tag === "REASONING_SANITIZE" && /xhigh → high/.test(m)),
+    "logs the downgrade"
+  );
+});
+
+test("sanitizeReasoningEffortForProvider: #7044 output_config.effort high passes through unchanged", () => {
+  const body = {
+    model: "claude-opus-4-6",
+    output_config: { effort: "high" },
+    messages: [{ role: "user", content: "hi" }],
+  };
+  const result = sanitizeReasoningEffortForProvider(body, "claude", "claude-opus-4-6", null);
+  assert.equal(result, body, "high is supported — body returned unchanged");
+  assert.equal((result as any).output_config.effort, "high");
+});
diff --git a/tests/unit/batch-status-cache.test.ts b/tests/unit/batch-status-cache.test.ts
index 458da72cb0b..46c1887c17f 100644
--- a/tests/unit/batch-status-cache.test.ts
+++ b/tests/unit/batch-status-cache.test.ts
@@ -59,6 +59,14 @@ test("getCached returns entry when mtime is 0 and stored mtime is also 0", () =>
   assert.deepEqual(result, status);
 });
 
+test("getCached expires unchanged negative results after 30 seconds", () => {
+  const status = makeStatus(false);
+  setCached("codex", 0, status, 1_000);
+
+  assert.deepEqual(getCached("codex", 0, 30_999), status);
+  assert.equal(getCached("codex", 0, 31_000), null);
+});
+
 // ── invalidate ────────────────────────────────────────────────────────────────
 
 test("invalidate removes entry from cache", () => {
diff --git a/tests/unit/chatcore-compression-analytics-write.test.ts b/tests/unit/chatcore-compression-analytics-write.test.ts
index 15768b22e17..ab3d056b8fb 100644
--- a/tests/unit/chatcore-compression-analytics-write.test.ts
+++ b/tests/unit/chatcore-compression-analytics-write.test.ts
@@ -12,9 +12,8 @@ const testDataDir = fs.mkdtempSync(path.join(os.tmpdir(), "omni-comp-analytics-t
 process.env.DATA_DIR = testDataDir;
 
 const coreDb = await import("../../src/lib/db/core.ts");
-const { writeCompressionAnalytics } = await import(
-  "../../open-sse/handlers/chatCore/compressionAnalyticsWrite.ts"
-);
+const { writeCompressionAnalytics } =
+  await import("../../open-sse/handlers/chatCore/compressionAnalyticsWrite.ts");
 
 type Stats = Parameters<typeof writeCompressionAnalytics>[0]["stats"];
 
@@ -37,7 +36,7 @@ function analyticsRow(requestId: string): Record<string, unknown> | undefined {
     return coreDb
       .getDbInstance()
       .prepare(
-        "SELECT mode, engine, original_tokens AS orig, compressed_tokens AS comp, tokens_saved AS saved, validation_fallback AS vf FROM compression_analytics WHERE request_id = ?"
+        "SELECT mode, engine, original_tokens AS orig, compressed_tokens AS comp, tokens_saved AS saved, estimated_usd_saved AS usd, validation_fallback AS vf FROM compression_analytics WHERE request_id = ?"
       )
       .get(requestId) as Record<string, unknown> | undefined;
   } catch {
@@ -122,6 +121,19 @@ test("inserts a per-engine breakdown when present", async () => {
   assert.equal(breakdownCount("ca-req-3"), 2);
 });
 
+test("inserts the analytics row when calculateCost throws", async () => {
+  await writeCompressionAnalytics(baseOpts("ca-req-cost-error"), {
+    calculateCost: async () => {
+      throw new Error("forced cost-estimate failure");
+    },
+  });
+
+  const row = analyticsRow("ca-req-cost-error");
+  assert.ok(row, "expected a compression_analytics row despite the cost-estimate failure");
+  assert.equal(row!.saved, 120);
+  assert.equal(row!.usd, null);
+});
+
 test("never rejects even on a bad write (fail-open)", async () => {
   coreDb.resetDbInstance();
   await assert.doesNotReject(writeCompressionAnalytics(baseOpts("ca-req-4")));
diff --git a/tests/unit/chatcore-sanitization.test.ts b/tests/unit/chatcore-sanitization.test.ts
index b92a098abf3..50cf0fe12b1 100644
--- a/tests/unit/chatcore-sanitization.test.ts
+++ b/tests/unit/chatcore-sanitization.test.ts
@@ -178,7 +178,7 @@ test("chatCore sanitization normalizes max_output_tokens into max_tokens", async
     },
   });
 
-  assert.equal(copied.call.body.max_tokens, 0);
+  assert.equal(copied.call.body.max_tokens, undefined);
   assert.equal("max_output_tokens" in copied.call.body, false);
   assert.equal(preserved.call.body.max_tokens, 7);
   assert.equal("max_output_tokens" in preserved.call.body, false);
diff --git a/tests/unit/chatcore-streaming-pipeline.test.ts b/tests/unit/chatcore-streaming-pipeline.test.ts
index c112ff6d217..bd3b3057010 100644
--- a/tests/unit/chatcore-streaming-pipeline.test.ts
+++ b/tests/unit/chatcore-streaming-pipeline.test.ts
@@ -7,9 +7,8 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
 
-const { assembleStreamingPipeline } = await import(
-  "../../open-sse/handlers/chatCore/streamingPipeline.ts"
-);
+const { assembleStreamingPipeline } =
+  await import("../../open-sse/handlers/chatCore/streamingPipeline.ts");
 
 // A fake stream: each pipeThrough appends the transform's tag and returns a new fake stream.
 function fakeStream(tag: string, log: string[]) {
@@ -100,3 +99,37 @@ test("full chain order: pii → progress → heartbeat → echo", () => {
   assembleStreamingPipeline(baseArgs({ echoModel: "alias-x" }), deps);
   assert.deepEqual(log, ["pii-flag", "progress", "heartbeat", "echo"]);
 });
+
+test("pipeline assembly creates performance mark and measure entries", () => {
+  performance.clearMarks();
+  performance.clearMeasures();
+  const { deps } = makeDeps();
+  assembleStreamingPipeline(baseArgs(), deps);
+  assert.equal(performance.getEntriesByName("omni-pipeline-start").length, 1, "start mark");
+  assert.equal(performance.getEntriesByName("omni-pipeline-end").length, 1, "end mark");
+  assert.equal(performance.getEntriesByName("omni-pipeline").length, 1, "measure");
+});
+
+test("re-entering pipeline clears previous timeline entries", () => {
+  performance.clearMarks();
+  performance.clearMeasures();
+  const { deps } = makeDeps();
+  // First call creates entries
+  assembleStreamingPipeline(baseArgs(), deps);
+  // Second call — clearMarks/clearMeasures runs before new marks, so count stays 1
+  assembleStreamingPipeline(baseArgs(), deps);
+  assert.equal(performance.getEntriesByName("omni-pipeline-start").length, 1, "start cleared");
+  assert.equal(performance.getEntriesByName("omni-pipeline-end").length, 1, "end cleared");
+  assert.equal(performance.getEntriesByName("omni-pipeline").length, 1, "measure cleared");
+});
+
+test("performance measure has positive duration", () => {
+  performance.clearMarks();
+  performance.clearMeasures();
+  const { deps } = makeDeps();
+  assembleStreamingPipeline(baseArgs(), deps);
+  const [entry] = performance.getEntriesByName("omni-pipeline");
+  assert.ok(entry, "measure exists");
+  assert.equal(entry.entryType, "measure");
+  assert.ok(entry.duration >= 0, `duration ${entry.duration} >= 0`);
+});
diff --git a/tests/unit/claude-context-1m-supported-models.test.ts b/tests/unit/claude-context-1m-supported-models.test.ts
new file mode 100644
index 00000000000..dbac4c55e29
--- /dev/null
+++ b/tests/unit/claude-context-1m-supported-models.test.ts
@@ -0,0 +1,153 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import path from "node:path";
+
+const REPO_ROOT = path.resolve(import.meta.dirname, "../..");
+
+// ─── Parse CONTEXT_1M_SUPPORTED_MODELS from source (no module import needed) ───
+
+function parseSupportedModels(): string[] {
+  const src = fs.readFileSync(
+    path.join(REPO_ROOT, "open-sse/services/claudeCodeCompatible.ts"),
+    "utf8"
+  );
+  // Strip type annotations before matching to handle `const X: string[] = [...]`
+  const match = src
+    .replace(/:\s*\w+(\[\])?/, "")
+    .match(/CONTEXT_1M_SUPPORTED_MODELS\s*=\s*\[([\s\S]*?)\]/);
+  if (!match) throw new Error("CONTEXT_1M_SUPPORTED_MODELS not found in source");
+  return match[1]
+    .split(",")
+    .map((s) => s.replace(/["'\s]/g, "").toLowerCase())
+    .filter(Boolean);
+}
+
+// ─── Parse Claude models from provider registry (source-level) ───
+// Uses brace-counting to isolate each model block within the models array,
+// avoiding false matches on the provider-level id field.
+
+function parseClaudeRegistryModels(): Array<{
+  provider: string;
+  modelId: string;
+  contextLength: number;
+}> {
+  const registryDir = path.join(REPO_ROOT, "open-sse/config/providers/registry/claude");
+  const results: Array<{
+    provider: string;
+    modelId: string;
+    contextLength: number;
+  }> = [];
+
+  for (const file of fs.readdirSync(registryDir).filter((f) => f.endsWith(".ts"))) {
+    const src = fs.readFileSync(path.join(registryDir, file), "utf8");
+
+    // Extract provider id
+    const providerIdMatch = src.match(/id:\s*["']([^"']+)["']/);
+    const providerId = providerIdMatch?.[1] ?? file.replace(".ts", "");
+
+    // Find the models array
+    const modelsStart = src.indexOf("models:");
+    if (modelsStart === -1) continue;
+    const arrayStart = src.indexOf("[", modelsStart);
+    if (arrayStart === -1) continue;
+
+    // Extract individual model objects using brace counting.
+    // Skips any non-model nested braces by requiring both id and contextLength
+    // in the same brace-delimited block.
+    let braceDepth = 0;
+    let objStart = -1;
+    const modelBlocks: string[] = [];
+    for (let i = arrayStart; i < src.length; i++) {
+      if (src[i] === "{") {
+        if (braceDepth === 0) objStart = i;
+        braceDepth++;
+      } else if (src[i] === "}") {
+        braceDepth--;
+        if (braceDepth === 0 && objStart !== -1) {
+          modelBlocks.push(src.slice(objStart, i + 1));
+          objStart = -1;
+        }
+      } else if (src[i] === "]" && braceDepth === 0) {
+        break;
+      }
+    }
+
+    for (const block of modelBlocks) {
+      const idMatch = block.match(/id:\s*["']([^"']+)["']/);
+      const ctxMatch = block.match(/contextLength:\s*(\d+)/);
+      if (idMatch && ctxMatch) {
+        results.push({
+          provider: providerId,
+          modelId: idMatch[1].toLowerCase(),
+          contextLength: parseInt(ctxMatch[1], 10),
+        });
+      }
+    }
+  }
+  return results;
+}
+
+const CONTEXT_1M_THRESHOLD = 200_000;
+
+function isModel1mSupported(modelId: string, supportedModels: string[]): boolean {
+  const normalized = modelId.replace(/:.*$/, "").toLowerCase();
+  return supportedModels.some((s) => normalized === s || normalized.startsWith(`${s}-`));
+}
+
+// ─── Sanity: parser must find at least one model ───
+
+test("parser finds at least one Claude model in the registry", () => {
+  const models = parseClaudeRegistryModels();
+  assert.ok(
+    models.length > 0,
+    "parseClaudeRegistryModels returned 0 models — check that open-sse/config/providers/registry/claude/ exists and contains model definitions"
+  );
+});
+
+// ─── Forward: every high-context Claude model must be in the allowlist ───
+
+test("every Claude model with contextLength > 200K is in CONTEXT_1M_SUPPORTED_MODELS", () => {
+  const supportedModels = parseSupportedModels();
+  const claudeModels = parseClaudeRegistryModels();
+  const violations: string[] = [];
+
+  for (const m of claudeModels) {
+    if (m.contextLength > CONTEXT_1M_THRESHOLD && !isModel1mSupported(m.modelId, supportedModels)) {
+      violations.push(`${m.provider}/${m.modelId} (contextLength=${m.contextLength})`);
+    }
+  }
+
+  assert.deepEqual(
+    violations,
+    [],
+    `Claude models with contextLength > 200K missing from CONTEXT_1M_SUPPORTED_MODELS.\n` +
+      `Add the model prefix to the allowlist in claudeCodeCompatible.ts.\n` +
+      `Violations:\n  ${violations.join("\n  ")}`
+  );
+});
+
+// ─── Reverse: every allowlist entry must have a matching high-context model ───
+
+test("every CONTEXT_1M_SUPPORTED_MODELS entry has a matching Claude model with contextLength > 200K", () => {
+  const supportedModels = parseSupportedModels();
+  const claudeModels = parseClaudeRegistryModels();
+  const orphans: string[] = [];
+
+  for (const supported of supportedModels) {
+    const found = claudeModels.some(
+      (m) =>
+        m.contextLength > CONTEXT_1M_THRESHOLD &&
+        (m.modelId === supported || m.modelId.startsWith(`${supported}-`))
+    );
+    if (!found) orphans.push(supported);
+  }
+
+  assert.deepEqual(
+    orphans,
+    [],
+    `CONTEXT_1M_SUPPORTED_MODELS entries with no matching high-context Claude model.\n` +
+      `Remove stale entries or check model id spelling.\n` +
+      `Orphans:\n  ${orphans.join("\n  ")}`
+  );
+});
diff --git a/tests/unit/claude-to-gemini-budget-tokens-zero-6813.test.ts b/tests/unit/claude-to-gemini-budget-tokens-zero-6813.test.ts
new file mode 100644
index 00000000000..422e2b5be63
--- /dev/null
+++ b/tests/unit/claude-to-gemini-budget-tokens-zero-6813.test.ts
@@ -0,0 +1,42 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+const { claudeToGeminiRequest } = await import(
+  "../../open-sse/translator/request/claude-to-gemini.ts"
+);
+
+// Regression for #6813 — the Claude→Gemini thinking transform used a truthy
+// check (`body.thinking.budget_tokens`) that silently dropped `budget_tokens: 0`
+// (dynamic thinking). The value must be preserved as `thinkingBudget: 0`.
+
+test("Claude -> Gemini preserves thinking.budget_tokens: 0 (dynamic thinking) (#6813)", () => {
+  const result = claudeToGeminiRequest(
+    "gemini-2.5-pro",
+    {
+      messages: [{ role: "user", content: [{ type: "text", text: "hi" }] }],
+      thinking: { type: "enabled", budget_tokens: 0 },
+    },
+    false
+  );
+
+  assert.deepEqual(result.generationConfig.thinkingConfig, {
+    thinkingBudget: 0,
+    includeThoughts: true,
+  });
+});
+
+test("Claude -> Gemini prefers a positive thinking.budget_tokens over budget 0 semantics", () => {
+  const result = claudeToGeminiRequest(
+    "gemini-2.5-pro",
+    {
+      messages: [{ role: "user", content: [{ type: "text", text: "hi" }] }],
+      thinking: { type: "enabled", budget_tokens: 8192 },
+    },
+    false
+  );
+
+  assert.deepEqual(result.generationConfig.thinkingConfig, {
+    thinkingBudget: 8192,
+    includeThoughts: true,
+  });
+});
diff --git a/tests/unit/cli-runtime-detection.test.ts b/tests/unit/cli-runtime-detection.test.ts
index 5c46a14161a..bd996616f89 100644
--- a/tests/unit/cli-runtime-detection.test.ts
+++ b/tests/unit/cli-runtime-detection.test.ts
@@ -170,6 +170,7 @@ describe("Healthcheck — checkRunnable", () => {
       assert.ok(result.installed, `Expected installed=true, got reason=${result.reason}`);
       if (result.runnable) {
         assert.ok(result.reason === null, `Expected no reason, got ${result.reason}`);
+        assert.equal(result.version, "1.0.0");
       }
     } finally {
       if (prev !== undefined) process.env.CLI_CLINE_BIN = prev;
@@ -177,6 +178,27 @@ describe("Healthcheck — checkRunnable", () => {
     }
   });
 
+  it("should detect Claude through an explicit read-only executable path", async () => {
+    const previousOverride = process.env.CLI_CLAUDE_BIN;
+    const script =
+      process.platform === "win32"
+        ? createFile(tmpDir, "claude.cmd", "@echo off\necho 2.1.211 (Claude Code)\n")
+        : createFile(tmpDir, "claude", "#!/bin/sh\necho '2.1.211 (Claude Code)'\n");
+    if (process.platform !== "win32") fs.chmodSync(script, 0o555);
+    process.env.CLI_CLAUDE_BIN = script;
+
+    try {
+      const result = await getCliRuntimeStatus("claude");
+      assert.equal(result.installed, true);
+      assert.equal(result.runnable, true);
+      assert.equal(result.commandPath, script);
+      assert.equal(result.version, "2.1.211 (Claude Code)");
+    } finally {
+      if (previousOverride === undefined) delete process.env.CLI_CLAUDE_BIN;
+      else process.env.CLI_CLAUDE_BIN = previousOverride;
+    }
+  });
+
   it("should detect qodercli via env override and mark it runnable", async () => {
     const prev = process.env.CLI_QODER_BIN;
     const script =
@@ -208,13 +230,35 @@ describe("Unknown tool", () => {
   });
 });
 
-// ─── continue tool (requiresBinary: false) ────────────────────
+// ─── Continue CLI (`cn`) ──────────────────────────────────────
 
-describe("continue tool — no binary required", () => {
-  it("should report installed=true without checking binary", async () => {
-    const result = await getCliRuntimeStatus("continue");
-    assert.equal(result.installed, true);
-    assert.equal(result.reason, "not_required");
+describe("Continue CLI detection", () => {
+  it("should not report Continue as installed when the cn binary is absent", async () => {
+    const previousPath = process.env.PATH;
+    const previousOverride = process.env.CLI_CONTINUE_BIN;
+    process.env.PATH = "";
+    delete process.env.CLI_CONTINUE_BIN;
+
+    try {
+      const result = await getCliRuntimeStatus("continue");
+      assert.equal(result.installed, false);
+      assert.equal(result.runnable, false);
+      assert.equal(result.reason, "not_found");
+      assert.equal(result.requiresBinary, true);
+    } finally {
+      if (previousPath === undefined) delete process.env.PATH;
+      else process.env.PATH = previousPath;
+      if (previousOverride === undefined) delete process.env.CLI_CONTINUE_BIN;
+      else process.env.CLI_CONTINUE_BIN = previousOverride;
+    }
+  });
+
+  it("should enumerate cn in Continue's known installation paths", () => {
+    const knownPaths = getKnownToolPaths("continue");
+    assert.ok(
+      knownPaths.some((knownPath) => /^cn(?:\.cmd)?$/i.test(path.basename(knownPath))),
+      "Continue detection should search for the cn executable"
+    );
   });
 });
 
diff --git a/tests/unit/clinepass-thinking-budget.test.ts b/tests/unit/clinepass-thinking-budget.test.ts
index db49bac5830..ade39fa1b7b 100644
--- a/tests/unit/clinepass-thinking-budget.test.ts
+++ b/tests/unit/clinepass-thinking-budget.test.ts
@@ -3,8 +3,9 @@ import assert from "node:assert/strict";
 
 import { DefaultExecutor } from "../../open-sse/executors/default.ts";
 
-// DefaultExecutor.ensureThinkingBudget — clinepass-gated max_tokens floor for
+// DefaultExecutor.ensureThinkingBudget — max_tokens floor for
 // reasoning models (prevents empty content when the budget is undersized).
+// Previously gated to clinepass only; now applies to all providers (#6912).
 
 test("bumps undersized max_tokens to 4096 for a clinepass reasoning model", () => {
   const executor = new DefaultExecutor("clinepass");
@@ -64,14 +65,17 @@ test("no-op for a non-reasoning clinepass model", () => {
   assert.equal(body.max_tokens, 100);
 });
 
-test("no-op for a non-clinepass provider (gate)", () => {
-  const executor = new DefaultExecutor("openrouter");
+test("bumps undersized max_tokens for a non-clinepass reasoning provider (gate removed, #6912)", () => {
+  // Issue #6912: ensureThinkingBudget was gated to clinepass only.
+  // Now it applies to all providers. Use nvidia (non-clinepass) which has
+  // deepseek-ai/deepseek-v4-pro with supportsReasoning in the registry.
+  const executor = new DefaultExecutor("nvidia");
   const body = {
-    model: "cline-pass/deepseek-v4-pro",
+    model: "deepseek-ai/deepseek-v4-pro",
     reasoning_effort: "high",
     max_tokens: 100,
   } as Record<string, unknown>;
 
-  executor.ensureThinkingBudget(body, "cline-pass/deepseek-v4-pro");
-  assert.equal(body.max_tokens, 100);
+  executor.ensureThinkingBudget(body, "deepseek-ai/deepseek-v4-pro");
+  assert.equal(body.max_tokens, 4096);
 });
diff --git a/tests/unit/codex-reset-credits.test.ts b/tests/unit/codex-reset-credits.test.ts
index a9fb0e282ad..e9466858d2a 100644
--- a/tests/unit/codex-reset-credits.test.ts
+++ b/tests/unit/codex-reset-credits.test.ts
@@ -136,6 +136,153 @@ test("consumeCodexResetCredit accepts alreadyRedeemed as success", async () => {
   assert.equal(result.outcome, "alreadyRedeemed");
 });
 
+test("consumeCodexResetCredit automatically redeems the soonest-expiring available credit", async () => {
+  const connection = (await createCodexConnection()) as { id: string };
+  let consumedCreditId: string | null = null;
+
+  globalThis.fetch = async (url, init = {}) => {
+    if (String(url).endsWith("/rate-limit-reset-credits")) {
+      return new Response(
+        JSON.stringify({
+          credits: [
+            {
+              id: "credit-later",
+              status: "available",
+              expires_at: "2099-08-20T10:00:00.000Z",
+            },
+            {
+              id: "credit-no-expiry",
+              status: "available",
+              expires_at: null,
+            },
+            {
+              id: "credit-first",
+              status: "available",
+              expires_at: "2099-07-20T10:00:00.000Z",
+            },
+          ],
+        }),
+        { status: 200 }
+      );
+    }
+    if (String(url).includes("/rate-limit-reset-credits/consume")) {
+      consumedCreditId = JSON.parse(String(init.body)).credit_id;
+      return new Response(JSON.stringify({ code: "reset" }), { status: 200 });
+    }
+    if (String(url).includes("/backend-api/wham/usage")) {
+      return new Response(JSON.stringify({ rate_limit: {} }), { status: 200 });
+    }
+    return new Response("unexpected", { status: 500 });
+  };
+
+  await resetCredits.consumeCodexResetCredit(connection.id, "redeem-fefo");
+  assert.equal(consumedCreditId, "credit-first");
+});
+
+test("consumeCodexResetCredit redeems an explicitly selected available credit", async () => {
+  const connection = (await createCodexConnection()) as { id: string };
+  let consumedCreditId: string | null = null;
+
+  globalThis.fetch = async (url, init = {}) => {
+    if (String(url).endsWith("/rate-limit-reset-credits")) {
+      return new Response(
+        JSON.stringify({
+          credits: [
+            { id: "credit-first", status: "available", expires_at: "2099-07-20T10:00:00Z" },
+            { id: "credit-chosen", status: "available", expires_at: "2099-08-20T10:00:00Z" },
+          ],
+        }),
+        { status: 200 }
+      );
+    }
+    if (String(url).includes("/rate-limit-reset-credits/consume")) {
+      consumedCreditId = JSON.parse(String(init.body)).credit_id;
+      return new Response(JSON.stringify({ code: "reset" }), { status: 200 });
+    }
+    if (String(url).includes("/backend-api/wham/usage")) {
+      return new Response(JSON.stringify({ rate_limit: {} }), { status: 200 });
+    }
+    return new Response("unexpected", { status: 500 });
+  };
+
+  await resetCredits.consumeCodexResetCredit(connection.id, "redeem-selected", "credit-chosen");
+  assert.equal(consumedCreditId, "credit-chosen");
+});
+
+test("listCodexResetCredits returns available credits ordered by expiry", async () => {
+  const connection = (await createCodexConnection()) as { id: string };
+
+  globalThis.fetch = async (url) => {
+    if (String(url).endsWith("/rate-limit-reset-credits")) {
+      return new Response(
+        JSON.stringify({
+          available_count: 3,
+          credits: [
+            {
+              id: "credit-no-expiry",
+              status: "available",
+              expires_at: null,
+              title: "No expiry",
+            },
+            {
+              id: "credit-redeemed",
+              status: "redeemed",
+              expires_at: "2099-06-01T00:00:00Z",
+            },
+            {
+              id: "credit-later",
+              status: "available",
+              expires_at: "2099-08-01T00:00:00Z",
+              title: "Later",
+            },
+            {
+              id: "credit-first",
+              status: "available",
+              expires_at: "2099-07-01T00:00:00Z",
+              title: "First",
+              description: "Expires first",
+            },
+          ],
+        }),
+        { status: 200 }
+      );
+    }
+    return new Response("unexpected", { status: 500 });
+  };
+
+  const result = await resetCredits.listCodexResetCredits(connection.id);
+  assert.equal(result.availableCount, 3);
+  assert.deepEqual(
+    result.credits.map((credit) => credit.selectionToken),
+    ["credit-first", "credit-later", "credit-no-expiry"]
+  );
+  assert.equal(result.credits[0].title, "First");
+  assert.equal(result.credits[0].description, "Expires first");
+  assert.equal(result.credits[0].expiresAt, "2099-07-01T00:00:00Z");
+});
+
+test("consumeCodexResetCredit rejects an unavailable manual selection", async () => {
+  const connection = (await createCodexConnection()) as { id: string };
+
+  globalThis.fetch = async (url) => {
+    if (String(url).endsWith("/rate-limit-reset-credits")) {
+      return new Response(JSON.stringify({ credits: [{ id: "credit-other" }] }), {
+        status: 200,
+      });
+    }
+    return new Response("unexpected", { status: 500 });
+  };
+
+  await assert.rejects(
+    () =>
+      resetCredits.consumeCodexResetCredit(connection.id, "redeem-unavailable", "credit-missing"),
+    (error: unknown) =>
+      error instanceof resetCredits.CodexResetCreditError &&
+      error.status === 409 &&
+      error.code === "selected_credit_unavailable"
+  );
+});
+
 for (const code of ["noCredit", "nothingToReset"]) {
   test(`consumeCodexResetCredit maps ${code} to 409`, async () => {
     const connection = (await createCodexConnection()) as { id: string };
diff --git a/tests/unit/combo-auto-config-split.test.ts b/tests/unit/combo-auto-config-split.test.ts
index a833cae0c43..dd755b3a106 100644
--- a/tests/unit/combo-auto-config-split.test.ts
+++ b/tests/unit/combo-auto-config-split.test.ts
@@ -3,6 +3,7 @@ import assert from "node:assert/strict";
 
 import { parseAutoConfig } from "@omniroute/open-sse/services/combo/autoConfig.ts";
 import { DEFAULT_WEIGHTS } from "@omniroute/open-sse/services/autoCombo/scoring.ts";
+import { MODE_PACKS } from "@omniroute/open-sse/services/autoCombo/modePacks.ts";
 
 // Split guard for Block J Task 2: parseAutoConfig was extracted verbatim from
 // handleComboChat's inline auto-strategy config block. These assertions pin the
@@ -62,6 +63,22 @@ test("explicit candidatePool, weights, exploration and budget are honored", () =
   assert.equal(cfg.modePack, "coding");
 });
 
+test("valid modePack overrides configured weights for fallback scoring", () => {
+  const cfg = parseAutoConfig(
+    {
+      name: "c",
+      autoConfig: {
+        weights: { ...DEFAULT_WEIGHTS, latencyInv: 0 },
+        modePack: "ship-fast",
+      },
+    } as never,
+    []
+  );
+
+  assert.equal(cfg.modePack, "ship-fast");
+  assert.equal(cfg.weights, MODE_PACKS["ship-fast"]);
+});
+
 test("config.auto is preferred over top-level config", () => {
   const cfg = parseAutoConfig(
     { name: "c", config: { auto: { routerStrategy: "cost" }, routerStrategy: "rules" } } as never,
diff --git a/tests/unit/combo-breaker-429.test.ts b/tests/unit/combo-breaker-429.test.ts
index 98fbc966ed6..988ebfb9c1c 100644
--- a/tests/unit/combo-breaker-429.test.ts
+++ b/tests/unit/combo-breaker-429.test.ts
@@ -67,3 +67,14 @@ test("skipProviderBreaker:true suppresses recording regardless of status", () =>
     );
   }
 });
+
+test("request-scoped synthetic 502 does NOT record a whole-provider breaker failure", () => {
+  assert.equal(
+    shouldRecordProviderBreakerFailure({
+      ...OTHER_ARGS,
+      status: 502,
+      requestScopedFailure: true,
+    }),
+    false
+  );
+});
diff --git a/tests/unit/combo-context-window-filter.test.ts b/tests/unit/combo-context-window-filter.test.ts
index 27f139fd558..342e360ce45 100644
--- a/tests/unit/combo-context-window-filter.test.ts
+++ b/tests/unit/combo-context-window-filter.test.ts
@@ -16,7 +16,8 @@ process.env.DATA_DIR = TEST_DATA_DIR;
 const core = await import("../../src/lib/db/core.ts");
 const { saveModelsDevCapabilities, clearModelsDevCapabilities } =
   await import("../../src/lib/modelsDevSync.ts");
-const { filterTargetsByRequestCompatibility } = await import("../../open-sse/services/combo.ts");
+const { filterTargetsByRequestCompatibility, getKnownContextOverflow, handleComboChat } =
+  await import("../../open-sse/services/combo.ts");
 
 test.after(() => {
   core.resetDbInstance();
@@ -54,6 +55,14 @@ function capabilityEntry(limitContext: number | null) {
   };
 }
 
+function capabilityEntryWithLimits(limitInput: number | null, limitContext: number | null, limitOutput = 4096) {
+  return {
+    ...capabilityEntry(limitContext),
+    limit_input: limitInput,
+    limit_output: limitOutput,
+  };
+}
+
 function target(modelStr: string) {
   return {
     kind: "model" as const,
@@ -74,6 +83,16 @@ function largeContextBody() {
   };
 }
 
+// Build a body whose input estimates to roughly `tokens` tokens. estimateTokens
+// is `ceil(charCount / 4)` over the JSON-serialized payload, so a run of
+// `tokens * 4` characters lands the estimate near `tokens` (the wrapper is
+// negligible at this scale).
+function bigContextBody(tokens: number) {
+  return {
+    messages: [{ role: "user", content: "x".repeat(tokens * 4) }],
+  };
+}
+
 const noopLog = { info() {}, warn() {}, error() {}, debug() {} };
 
 test("known compatible context target wins over unknown-context targets", () => {
@@ -162,3 +181,186 @@ test("all known-too-small context targets still fall back to strategy order", ()
     ["unit-known-context/tiny", "unit-known-context/small"]
   );
 });
+
+test("known context overflow reports the largest target limit", () => {
+  saveModelsDevCapabilities({
+    "unit-known-context": {
+      tiny: capabilityEntry(8_000),
+      small: capabilityEntry(16_000),
+    },
+  });
+
+  const overflow = getKnownContextOverflow(
+    [target("unit-known-context/tiny"), target("unit-known-context/small")],
+    largeContextBody()
+  );
+
+  assert.ok(overflow);
+  assert.ok(overflow.requiredContextTokens > overflow.maxKnownContextTokens);
+  assert.equal(overflow.maxKnownContextTokens, 16_000);
+  assert.equal(overflow.targetCount, 2);
+});
+
+test("#7177 an empty messages array is not counted as real content at an exact-boundary limit", () => {
+  // Regression: some combo entrypoints default a caller-omitted `messages` to `[]`. The
+  // estimator used to JSON.stringify whatever keys were merely *present* on the body,
+  // so an empty array still contributed a few phantom "structural" tokens (JSON braces/
+  // brackets), which was enough to trip a false-positive overflow when max_tokens exactly
+  // equals the target's context window (a common config where limit_input === limit_output
+  // === limit_context) even though there is no real input to account for.
+  saveModelsDevCapabilities({
+    "unit-known-context": {
+      exact: capabilityEntry(4_096),
+    },
+  });
+
+  const overflow = getKnownContextOverflow([target("unit-known-context/exact")], {
+    messages: [],
+    max_tokens: 4_096,
+  });
+
+  assert.equal(overflow, null);
+});
+
+test("unknown context metadata keeps overflow detection fail-open", () => {
+  saveModelsDevCapabilities({
+    "unit-known-context": {
+      tiny: capabilityEntry(8_000),
+    },
+  });
+
+  const overflow = getKnownContextOverflow(
+    [target("unit-known-context/tiny"), target("unit-unknown-context/mystery")],
+    largeContextBody()
+  );
+
+  assert.equal(overflow, null);
+});
+
+test("combo rejects a known oversized request before upstream dispatch", async () => {
+  saveModelsDevCapabilities({
+    "unit-known-context": {
+      tiny: capabilityEntry(8_000),
+      small: capabilityEntry(16_000),
+    },
+  });
+  let dispatches = 0;
+
+  const response = await handleComboChat({
+    body: largeContextBody(),
+    combo: {
+      name: "known-context-overflow",
+      strategy: "priority",
+      models: ["unit-known-context/tiny", "unit-known-context/small"],
+    },
+    handleSingleModel: async () => {
+      dispatches += 1;
+      return new Response("unexpected", { status: 200 });
+    },
+    log: noopLog,
+  });
+
+  assert.equal(response.status, 400);
+  assert.equal(dispatches, 0);
+  const body = await response.json();
+  assert.equal(body.error.code, "context_length_exceeded");
+  assert.equal(body.diagnostics.terminalReason, "context_length_exceeded");
+  assert.equal(body.diagnostics.attempted, 0);
+});
+
+test("input-only maxInputTokens is not double-counted against the output reserve (#7039)", () => {
+  // Faithful reproduction of #7039 (Codex gpt-5.5-xhigh):
+  //   maxInputTokens = 272_000, contextWindow = 400_000, maxOutputTokens = 128_000
+  // With max_tokens = 32_000 the OLD code required
+  //   maxInputTokens >= estimatedInputTokens + 32_000
+  // i.e. it allowed only ~240K of input against a real 272K input cap — the
+  // output reserve was double-counted against an already input-only cap. Here
+  // the input (~256K tokens) sits between the buggy allowance (240K) and the
+  // real cap (272K): the fix keeps the target, the bug drops it.
+  saveModelsDevCapabilities({
+    "unit-7039": {
+      "codex-like": capabilityEntryWithLimits(272_000, 400_000, 128_000),
+      huge: capabilityEntryWithLimits(1_000_000, 1_000_000, 500_000),
+    },
+  });
+
+  const out = filterTargetsByRequestCompatibility(
+    [target("unit-7039/codex-like"), target("unit-7039/huge")],
+    { ...bigContextBody(256_000), max_tokens: 32_000 },
+    noopLog
+  );
+
+  assert.deepEqual(
+    out.map((entry) => entry.modelStr),
+    ["unit-7039/codex-like", "unit-7039/huge"]
+  );
+});
+
+test("small input-only maxInputTokens keeps a target whose input fits even though output reserve would overflow the cap (#7039)", () => {
+  // A second, lightweight reproduction: with maxInputTokens = 100 the input-only
+  // cap comfortably holds the ~11-token input, but the old code compared it
+  // against input + output (~411) and rejected the target. The fix keeps it.
+  saveModelsDevCapabilities({
+    "unit-7039-small": {
+      "input-capped": capabilityEntryWithLimits(100, 1_000_000, 500),
+      huge: capabilityEntryWithLimits(1_000_000, 1_000_000, 500_000),
+    },
+  });
+
+  const out = filterTargetsByRequestCompatibility(
+    [target("unit-7039-small/input-capped"), target("unit-7039-small/huge")],
+    { messages: [{ role: "user", content: "hello" }], max_tokens: 400 },
+    noopLog
+  );
+
+  assert.deepEqual(
+    out.map((entry) => entry.modelStr),
+    ["unit-7039-small/input-capped", "unit-7039-small/huge"]
+  );
+});
+
+test("input-only maxInputTokens still rejects when the input itself exceeds the cap", () => {
+  // The fix must not let a genuinely-too-small input cap pass. `too-small` has
+  // maxInputTokens = 1, which cannot even hold the ~11-token input, so it must
+  // still be dropped while the compatible target survives.
+  saveModelsDevCapabilities({
+    "unit-7039-too-small": {
+      "too-small": capabilityEntryWithLimits(1, 1_000_000, 500),
+      huge: capabilityEntryWithLimits(1_000_000, 1_000_000, 500_000),
+    },
+  });
+
+  const out = filterTargetsByRequestCompatibility(
+    [target("unit-7039-too-small/too-small"), target("unit-7039-too-small/huge")],
+    { messages: [{ role: "user", content: "hello" }], max_tokens: 400 },
+    noopLog
+  );
+
+  assert.deepEqual(
+    out.map((entry) => entry.modelStr),
+    ["unit-7039-too-small/huge"]
+  );
+});
+
+test("maxInputTokens defaulting to contextWindow still rejects when input + output exceeds the total window (#7039 follow-up)", () => {
+  // Shared-window model where maxInputTokens equals the total window size.
+  // The input alone fits the input cap, but input + output overflows the
+  // window, so the target must be rejected instead of passing on the input cap.
+  saveModelsDevCapabilities({
+    "unit-7039-window": {
+      "shared-window": capabilityEntryWithLimits(400_000, 400_000, 200_000),
+      huge: capabilityEntryWithLimits(1_000_000, 1_000_000, 500_000),
+    },
+  });
+
+  const out = filterTargetsByRequestCompatibility(
+    [target("unit-7039-window/shared-window"), target("unit-7039-window/huge")],
+    { messages: [{ role: "user", content: "x".repeat(350_000 * 4) }], max_tokens: 100_000 },
+    noopLog
+  );
+
+  assert.deepEqual(
+    out.map((entry) => entry.modelStr),
+    ["unit-7039-window/huge"]
+  );
+});
diff --git a/tests/unit/combo-failure-log-message.test.ts b/tests/unit/combo-failure-log-message.test.ts
new file mode 100644
index 00000000000..61775e76c81
--- /dev/null
+++ b/tests/unit/combo-failure-log-message.test.ts
@@ -0,0 +1,29 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+const { getComboFailureLogError } = await import("../../src/sse/handlers/comboFailureLogging.ts");
+
+test("combo failure log preserves the concrete response error", async () => {
+  const response = new Response(
+    JSON.stringify({
+      error: {
+        message: "Request context exceeds every known target limit",
+        code: "context_length_exceeded",
+      },
+    }),
+    { status: 400, headers: { "Content-Type": "application/json" } }
+  );
+
+  assert.equal(
+    await getComboFailureLogError(response, "large-context-combo"),
+    "[400] Request context exceeds every known target limit"
+  );
+});
+
+test("combo failure log uses a neutral fallback for an unreadable response", async () => {
+  const response = new Response("not-json", { status: 503 });
+  assert.equal(
+    await getComboFailureLogError(response, "temporary-combo"),
+    '[503] Combo "temporary-combo" failed'
+  );
+});
diff --git a/tests/unit/combo-least-used-account.test.ts b/tests/unit/combo-least-used-account.test.ts
new file mode 100644
index 00000000000..67b929a8164
--- /dev/null
+++ b/tests/unit/combo-least-used-account.test.ts
@@ -0,0 +1,86 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { sortTargetsByUsage } from "../../open-sse/services/combo/targetSorters.ts";
+import {
+  recordComboRequest,
+  resetComboMetrics,
+  getComboMetrics,
+} from "../../open-sse/services/comboMetrics.ts";
+import type { ResolvedComboTarget } from "../../open-sse/services/combo/types.ts";
+
+function target(modelStr: string, executionKey: string, provider = "codex"): ResolvedComboTarget {
+  return {
+    kind: "model",
+    stepId: executionKey,
+    executionKey,
+    modelStr,
+    provider,
+    providerId: null,
+    connectionId: null,
+    weight: 1,
+    label: null,
+  };
+}
+
+test("least-used distributes across distinct accounts of the same model (#7015)", () => {
+  const combo = "test-7015-least-used";
+  resetComboMetrics(combo);
+
+  // Same model, two accounts. Account A serves 5 requests, account B serves 0.
+  for (let i = 0; i < 5; i++) {
+    recordComboRequest(combo, "codex/gpt-5.5", {
+      success: true,
+      latencyMs: 10,
+      strategy: "least-used",
+      target: { executionKey: "acct-a", provider: "codex" },
+    });
+  }
+
+  const metrics = getComboMetrics(combo);
+  assert.equal(metrics?.byTarget["acct-a"]?.requests, 5, "acct-a usage recorded");
+  // The shared modelStr usage is aggregated; the per-account bug ignores it for B.
+  assert.equal(metrics?.byModel["codex/gpt-5.5"]?.requests, 5);
+
+  const targets = [
+    target("codex/gpt-5.5", "acct-a"),
+    target("codex/gpt-5.5", "acct-b"),
+  ];
+  const ordered = sortTargetsByUsage(targets, combo);
+
+  // The unused account (B, 0 requests) must be selected first — not always A.
+  assert.equal(ordered[0].executionKey, "acct-b", "unused account should be first");
+  assert.equal(ordered[1].executionKey, "acct-a");
+
+  resetComboMetrics(combo);
+});
+
+test("least-used orders three same-model accounts by ascending per-account usage (#7015)", () => {
+  const combo = "test-7015-least-used-3";
+  resetComboMetrics(combo);
+
+  const counts: Record<string, number> = { "acct-a": 3, "acct-b": 1, "acct-c": 7 };
+  for (const [key, n] of Object.entries(counts)) {
+    for (let i = 0; i < n; i++) {
+      recordComboRequest(combo, "codex/gpt-5.5", {
+        success: true,
+        latencyMs: 10,
+        strategy: "least-used",
+        target: { executionKey: key, provider: "codex" },
+      });
+    }
+  }
+
+  const targets = [
+    target("codex/gpt-5.5", "acct-a"),
+    target("codex/gpt-5.5", "acct-b"),
+    target("codex/gpt-5.5", "acct-c"),
+  ];
+  const ordered = sortTargetsByUsage(targets, combo);
+  assert.deepEqual(
+    ordered.map((t) => t.executionKey),
+    ["acct-b", "acct-a", "acct-c"]
+  );
+
+  resetComboMetrics(combo);
+});
diff --git a/tests/unit/combo-resolve-auto-strategy-split.test.ts b/tests/unit/combo-resolve-auto-strategy-split.test.ts
index 7ffceeae168..84c0fe114e9 100644
--- a/tests/unit/combo-resolve-auto-strategy-split.test.ts
+++ b/tests/unit/combo-resolve-auto-strategy-split.test.ts
@@ -86,3 +86,139 @@ test("all candidates quota-cutoff-blocked -> early 429 Response", async () => {
     assert.equal(result.earlyResponse.status, 429);
   }
 });
+
+// #7008 follow-up: parseAutoConfig() (see combo-auto-config-split.test.ts) already
+// makes `weights` honor a combo's own STORED modePack. But resolveAutoStrategyOrder()
+// also supports a per-request `X-OmniRoute-Mode` override (relayOptions.mode) that can
+// pick a *different* mode pack than the one stored on the combo for that single
+// request — and `weights` must track that EFFECTIVE (post-override) modePack, not the
+// stored one, so scoreAutoTargets' fallback ranking doesn't drift from the primary
+// selection. These three synthetic candidates are built so every scoring factor is
+// IDENTICAL between them except cost/latency/stability, and "dominant" clearly wins
+// selection under any weight profile (so the engine's own randomized tier-rotation
+// never affects which one becomes `orderedTargets[0]`) — isolating the assertion to
+// the *fallback ranking order* of the remaining two, which is exactly what
+// scoreAutoTargets (and only scoreAutoTargets) controls.
+const weightSensitiveCandidates = () =>
+  [
+    {
+      kind: "model",
+      stepId: "dominant",
+      executionKey: "groq>dominant-model",
+      modelStr: "dominant-model",
+      provider: "groq",
+      model: "dominant-model",
+      quotaRemaining: 100,
+      quotaTotal: 100,
+      circuitBreakerState: "CLOSED",
+      costPer1MTokens: 0.01,
+      p95LatencyMs: 10,
+      latencyStdDev: 1,
+      errorRate: 0,
+    },
+    {
+      // Wins under "quality-first" (high taskFit/stability weight): low latencyStdDev
+      // (=> high stability), but the highest cost and highest latency in the pool.
+      kind: "model",
+      stepId: "quality-leaning",
+      executionKey: "openai>model-alpha",
+      modelStr: "model-alpha",
+      provider: "openai",
+      model: "model-alpha",
+      quotaRemaining: 100,
+      quotaTotal: 100,
+      circuitBreakerState: "CLOSED",
+      costPer1MTokens: 20,
+      p95LatencyMs: 5000,
+      latencyStdDev: 1,
+      errorRate: 0,
+    },
+    {
+      // Wins under "ship-fast" (high latencyInv/health weight): lowest latency and
+      // lowest cost in the pool, but the highest latencyStdDev (=> low stability).
+      kind: "model",
+      stepId: "speed-leaning",
+      executionKey: "anthropic>model-beta",
+      modelStr: "model-beta",
+      provider: "anthropic",
+      model: "model-beta",
+      quotaRemaining: 100,
+      quotaTotal: 100,
+      circuitBreakerState: "CLOSED",
+      costPer1MTokens: 1,
+      p95LatencyMs: 100,
+      latencyStdDev: 900,
+      errorRate: 0,
+    },
+  ] as never;
+
+const weightSensitiveDeps = (autoConfig: Record<string, unknown>, mode?: string) =>
+  ({
+    orderedTargets: [
+      target("groq", "dominant-model"),
+      target("openai", "model-alpha"),
+      target("anthropic", "model-beta"),
+    ],
+    body: { messages: [{ role: "user", content: "hi" }] },
+    combo: {
+      id: "auto-mode-override",
+      name: "auto-mode-override",
+      autoConfig: {
+        // Fixed candidatePool skips the DB-backed expandAutoComboCandidatePool
+        // path entirely (see the `candidatePool.length > 0` early-return there) —
+        // this test only cares about weight-driven ranking, not pool expansion.
+        candidatePool: ["groq", "openai", "anthropic"],
+        explorationRate: 0,
+        ...autoConfig,
+      },
+    },
+    settings: null,
+    config: {},
+    relayOptions: mode ? { mode } : null,
+    resilienceSettings: { quotaPreflight: { enabled: false } },
+    log: noopLog,
+    buildAutoCandidates: (async () => weightSensitiveCandidates()) as never,
+  }) as never;
+
+test("per-request X-OmniRoute-Mode override changes the EFFECTIVE weights used for fallback ranking, not just selection", async () => {
+  // Combo's own stored modePack is "quality-first" (would rank model-alpha before
+  // model-beta if the override were ignored), but the request overrides to
+  // "ship-fast" for this one call.
+  const overridden = await resolveAutoStrategyOrder(
+    weightSensitiveDeps({ modePack: "quality-first" }, "ship-fast")
+  );
+  assert.ok("orderedTargets" in overridden, "expected a normal ordering result, not earlyResponse");
+  if (!("orderedTargets" in overridden)) return;
+
+  // A combo natively configured with "ship-fast" (no override at all) is the ground
+  // truth for what "effective modePack = ship-fast" should rank like.
+  const native = await resolveAutoStrategyOrder(weightSensitiveDeps({ modePack: "ship-fast" }));
+  assert.ok("orderedTargets" in native, "expected a normal ordering result, not earlyResponse");
+  if (!("orderedTargets" in native)) return;
+
+  // "dominant" overwhelms every weight profile tried here, so it is always the
+  // engine's selection (position 0) regardless of any exploration/tier-rotation
+  // randomness — the two asserted positions below are populated exclusively by
+  // scoreAutoTargets' deterministic weight-driven sort.
+  assert.equal(overridden.orderedTargets[0].provider, "groq");
+  assert.equal(native.orderedTargets[0].provider, "groq");
+
+  // Ship-fast weights favor low-latency/high-health over stability, so the
+  // speed-leaning candidate outranks the quality-leaning one when the effective
+  // modePack is ship-fast — whether that's because it's natively configured that
+  // way, or because a per-request override made it so.
+  assert.equal(
+    overridden.orderedTargets[1].provider,
+    "anthropic",
+    "request-level ship-fast override should rank the speed-leaning candidate above the quality-leaning one"
+  );
+  assert.equal(overridden.orderedTargets[2].provider, "openai");
+
+  // The override case must match the native ship-fast case EXACTLY — proving the
+  // override drives an identical effective weight vector for fallback ranking, not
+  // just for the initial selection.
+  assert.deepEqual(
+    overridden.orderedTargets.map((t) => t.provider),
+    native.orderedTargets.map((t) => t.provider)
+  );
+});
diff --git a/tests/unit/combo-scoring-inspector.test.ts b/tests/unit/combo-scoring-inspector.test.ts
index 4f6d63c5801..473dc1e4ff6 100644
--- a/tests/unit/combo-scoring-inspector.test.ts
+++ b/tests/unit/combo-scoring-inspector.test.ts
@@ -21,11 +21,14 @@ const quotaSnapshotsDb = await import("../../src/lib/db/quotaSnapshots.ts");
 const callLogs = await import("../../src/lib/usage/callLogs.ts");
 const comboMetrics = await import("../../open-sse/services/comboMetrics.ts");
 const inspector = await import("../../src/lib/usage/comboScoringInspector.ts");
+const comboHealthDashboard = await import("../../src/lib/usage/comboHealthDashboard.ts");
 const route = await import("../../src/app/api/usage/combo-scoring-inspector/route.ts");
 const { normalizeComboStep } = await import("../../src/lib/combos/steps.ts");
 const { lockModel, clearAllModelLockouts } =
   await import("../../open-sse/services/accountFallback.ts");
 const { resetAllCircuitBreakers } = await import("../../src/shared/utils/circuitBreaker.ts");
+const { DEFAULT_WEIGHTS } = await import("../../open-sse/services/autoCombo/scoring.ts");
+const { MODE_PACKS } = await import("../../open-sse/services/autoCombo/modePacks.ts");
 
 async function resetStorage() {
   comboMetrics.resetAllComboMetrics();
@@ -41,10 +44,11 @@ async function enableManagementAuth() {
   await settingsDb.updateSettings({ requireLogin: true, password: "" });
 }
 
-async function seedAutoCombo() {
+async function seedAutoCombo(comboOverrides: Record<string, unknown> = {}) {
   const comboInput = {
     name: "combo-scoring-auto",
     strategy: "auto",
+    ...comboOverrides,
     models: [
       {
         kind: "model",
@@ -165,6 +169,8 @@ test("scoring inspector ranks targets and explains score contributions", async (
   assert.equal(response.method, "read_only_recompute");
   assert.equal(response.combos.length, 1);
   assert.equal(response.combos[0].strategy, "auto");
+  assert.equal(response.combos[0].weightSource, "default");
+  assert.equal(response.combos[0].modePack, null);
   assert.equal(response.combos[0].targets.length, 2);
   assert.equal(response.combos[0].selectedExecutionKey, firstStep.id);
   assert.equal(response.combos[0].targets[0].executionKey, firstStep.id);
@@ -181,6 +187,60 @@ test("scoring inspector ranks targets and explains score contributions", async (
   );
 });
 
+test("scoring inspector reports mode packs over explicit auto weights", async () => {
+  const combo = await combosDb.createCombo({
+    name: "combo-scoring-mode-pack",
+    strategy: "auto",
+    models: ["openai/gpt-4o-mini"],
+    config: {
+      auto: {
+        modePack: "ship-fast",
+        weights: { ...DEFAULT_WEIGHTS },
+      },
+    },
+  });
+
+  const response = await inspector.buildComboScoringInspectorResponse({
+    range: "24h",
+    horizon: "7d",
+    comboId: String(combo.id),
+    combos: [combo],
+    skipAutopilot: true,
+  });
+
+  assert.equal(response.combos.length, 1);
+  assert.equal(response.combos[0].weightSource, "mode_pack");
+  assert.equal(response.combos[0].modePack, "ship-fast");
+  assert.deepEqual(response.combos[0].weights, MODE_PACKS["ship-fast"]);
+});
+
+test("scoring inspector reports valid explicit auto weights", async () => {
+  const explicitWeights = {
+    ...DEFAULT_WEIGHTS,
+    latencyInv: 0.2,
+    costInv: 0.07,
+  };
+  const combo = await combosDb.createCombo({
+    name: "combo-scoring-explicit-weights",
+    strategy: "auto",
+    models: ["openai/gpt-4o-mini"],
+    autoConfig: { weights: explicitWeights },
+  });
+
+  const response = await inspector.buildComboScoringInspectorResponse({
+    range: "24h",
+    horizon: "7d",
+    comboId: String(combo.id),
+    combos: [combo],
+    skipAutopilot: true,
+  });
+
+  assert.equal(response.combos.length, 1);
+  assert.equal(response.combos[0].weightSource, "explicit");
+  assert.equal(response.combos[0].modePack, null);
+  assert.deepEqual(response.combos[0].weights, explicitWeights);
+});
+
 test("scoring inspector marks non-auto combos as explanatory recompute", async () => {
   const combo = await combosDb.createCombo({
     name: "combo-scoring-priority",
@@ -267,11 +327,15 @@ test("scoring inspector skipAutopilot avoids rebuilding autopilot report", async
     },
     skipAutopilot: true,
   };
-  Object.defineProperty(options, "combos", {
-    get() {
-      throw new Error("autopilot should not read combos when skipped");
-    },
-  });
+  // #7087: this used to assert that `options.combos` is never even READ when
+  // health/forecast/skipAutopilot are all supplied (via a getter that throws on
+  // access) — but that was pinning the exact bug: resolveConfiguredCombos() must
+  // check `options.combos` FIRST, unconditionally, so a caller-supplied `combos`
+  // array (e.g. from buildComboHealthDashboardResponse()) is never silently
+  // discarded just because health/forecast/autopilot were already resolved too.
+  // This test's own `options` genuinely has no `combos` (undefined), which still
+  // exercises the "no combos supplied + already have health signals -> skip the
+  // getCombos() DB round-trip" branch — no getter/trap needed for that.
 
   const response = await inspector.buildComboScoringInspectorResponse(options);
 
@@ -417,3 +481,41 @@ test("scoring inspector route requires auth, validates query, and returns 404",
   assert.equal(body.combos.length, 1);
   assert.equal(body.combos[0].targets.length, 2);
 });
+
+// #7087 follow-up: resolveConfiguredCombos() used to unconditionally return `[]`
+// whenever healthResponse + forecastResponse + (skipAutopilot || autopilotReport)
+// were ALL supplied, regardless of whether options.combos was also populated. That
+// is exactly the call shape comboHealthDashboard.ts::buildComboHealthDashboardResponse
+// always uses (it fetches combos once, then threads combos/healthResponse/
+// forecastResponse/autopilotReport into buildComboScoringInspectorResponse
+// together) -- so through the real dashboard integration, `combosById`/`combosByName`
+// (which resolveInspectorWeights() reads to report a combo's real configured
+// modePack/weights) always came back empty, and every combo's weightSource silently
+// fell back to "default" no matter what was actually configured. This test drives
+// the real buildComboHealthDashboardResponse() end-to-end (not
+// buildComboScoringInspectorResponse() directly, which the PR's own tests already
+// covered) with a combo configured for modePack "ship-fast".
+test("scoring inspector reports the configured weight source through the dashboard integration", async () => {
+  const { combo } = await seedAutoCombo({
+    config: { auto: { modePack: "ship-fast" } },
+  });
+
+  const dashboard = await comboHealthDashboard.buildComboHealthDashboardResponse({
+    range: "24h",
+    horizon: "7d",
+    comboId: String(combo.id),
+    taskType: "coding",
+    combos: [combo],
+  });
+
+  assert.equal(dashboard.errors.scoring, undefined);
+  assert.ok(dashboard.scoring, "expected a scoring inspector response, not null");
+  assert.equal(dashboard.scoring?.combos.length, 1);
+  assert.equal(
+    dashboard.scoring?.combos[0].weightSource,
+    "mode_pack",
+    "dashboard's own combos should drive the reported weight source, not silently fall back to default"
+  );
+  assert.equal(dashboard.scoring?.combos[0].modePack, "ship-fast");
+  assert.deepEqual(dashboard.scoring?.combos[0].weights, MODE_PACKS["ship-fast"]);
+});
diff --git a/tests/unit/combo-stickiness-responses-input-7270.test.ts b/tests/unit/combo-stickiness-responses-input-7270.test.ts
new file mode 100644
index 00000000000..125b551b5b8
--- /dev/null
+++ b/tests/unit/combo-stickiness-responses-input-7270.test.ts
@@ -0,0 +1,235 @@
+/**
+ * Regression for #7270 — combo session stickiness is a silent no-op on the OpenAI
+ * Responses API (`/v1/responses`).
+ *
+ * The stickiness key was derived exclusively from `body.messages`, but Responses-API
+ * requests carry their turns in `.input` and never populate `.messages`. So
+ * deriveMessageHash(body.messages) resolved to null, stickiness failed open, and a
+ * single conversation was re-ordered every turn as if stickiness were disabled —
+ * regardless of strategy or the toggle.
+ *
+ * This drives the REAL handleComboChat with a round-robin combo and an `.input`-shaped
+ * body and asserts:
+ *  - a single sessionless Responses-API conversation re-pins to its turn-1 connection
+ *    (FAILS before the fix: the combo rotates A → B → C …);
+ *  - distinct conversations still spread across connections (spreading preserved);
+ *  - the new normalizeStickinessMessages() helper covers both wire shapes.
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-stick-resp-7270-"));
+const ORIGINAL_DATA_DIR = process.env.DATA_DIR;
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const { handleComboChat } = await import("../../open-sse/services/combo.ts");
+const stick = await import("../../open-sse/services/combo/sessionStickiness.ts");
+const dbCore = await import("../../src/lib/db/core.ts");
+
+function makeLog() {
+  return { info() {}, warn() {}, debug() {}, error() {} };
+}
+
+function rrCombo(name: string) {
+  return {
+    name,
+    strategy: "round-robin",
+    config: { maxRetries: 0 },
+    models: [
+      { kind: "model", provider: "codex", providerId: "codex", model: "m-a", connectionId: "conn-A", id: `${name}-0` },
+      { kind: "model", provider: "codex", providerId: "codex", model: "m-b", connectionId: "conn-B", id: `${name}-1` },
+      { kind: "model", provider: "glm-cn", providerId: "glm-cn", model: "m-c", connectionId: "conn-C", id: `${name}-2` },
+    ],
+  };
+}
+
+// Responses-API shape: turns live in `.input` (array of message items with
+// input_text content parts), and `.messages` is absent.
+function responsesBody(combo: Record<string, unknown>, firstMessage: string) {
+  return {
+    model: combo.name,
+    input: [{ role: "user", content: [{ type: "input_text", text: firstMessage }] }],
+    stream: false,
+  };
+}
+
+async function dispatchConnection(
+  combo: Record<string, unknown>,
+  body: Record<string, unknown>
+): Promise<string> {
+  let conn = "?";
+  await handleComboChat({
+    body,
+    combo,
+    allCombos: [combo],
+    isModelAvailable: async () => true,
+    relayOptions: undefined,
+    signal: undefined,
+    settings: {},
+    log: makeLog(),
+    handleSingleModel: async (
+      _b: unknown,
+      modelStr: string,
+      target?: { connectionId?: string | null }
+    ) => {
+      conn = target?.connectionId ?? "?";
+      return Response.json({ choices: [{ message: { role: "assistant", content: modelStr } }] });
+    },
+  });
+  return conn;
+}
+
+test.beforeEach(() => {
+  stick.clearAllStickyBindings();
+  // Fail-open saturation (unknown → full headroom) so we exercise the stickiness
+  // MECHANISM, not the headroom gate.
+  stick.__setStickinessHeadroomFetcherForTests(async () => undefined);
+});
+
+test.after(() => {
+  stick.__setStickinessHeadroomFetcherForTests(null);
+  dbCore.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+  if (ORIGINAL_DATA_DIR === undefined) delete process.env.DATA_DIR;
+  else process.env.DATA_DIR = ORIGINAL_DATA_DIR;
+});
+
+test("normalizeStickinessMessages: covers .messages, .input array and .input string", () => {
+  const fromMessages = stick.normalizeStickinessMessages({
+    messages: [{ role: "user", content: "hi" }],
+  });
+  assert.deepEqual(fromMessages, [{ role: "user", content: "hi" }]);
+
+  const fromInputArray = stick.normalizeStickinessMessages({
+    input: [{ role: "user", content: [{ type: "input_text", text: "hi" }] }],
+  });
+  assert.ok(Array.isArray(fromInputArray) && fromInputArray.length === 1);
+
+  const fromInputString = stick.normalizeStickinessMessages({ input: "plain string turn" });
+  assert.deepEqual(fromInputString, [{ role: "user", content: "plain string turn" }]);
+
+  // .messages wins over .input when both present (Chat Completions precedence).
+  const both = stick.normalizeStickinessMessages({
+    messages: [{ role: "user", content: "from messages" }],
+    input: "from input",
+  });
+  assert.deepEqual(both, [{ role: "user", content: "from messages" }]);
+
+  assert.equal(stick.normalizeStickinessMessages({}), null);
+  assert.equal(stick.normalizeStickinessMessages(null), null);
+});
+
+test("normalizeStickinessMessages: .input array of PLAIN STRINGS maps to user messages (#7270 bare-string-array gap)", () => {
+  // The Responses API also allows `.input` to be an array of bare strings (each string
+  // is shorthand for an `input_text` user message) — the exact shape
+  // `responsesInputNormalization.ts`'s `normalizeCodexResponsesInputItem` already
+  // special-cases (`typeof itemValue === "string"`). Before this fix, the
+  // Array.isArray(input) branch cast the array straight through unmapped, so
+  // deriveMessageHash's `messages.find(m => m?.role === "user")` never matched a bare
+  // string and the key stayed null (fail-open), same bug as #7270 for this narrower shape.
+  const fromPlainStringArray = stick.normalizeStickinessMessages({
+    input: ["First turn of the conversation, plain string item"],
+  });
+  assert.deepEqual(fromPlainStringArray, [
+    { role: "user", content: "First turn of the conversation, plain string item" },
+  ]);
+
+  const key = stick.deriveMessageHash(fromPlainStringArray);
+  assert.ok(key !== null, "a bare-string .input array item must still yield a stickiness key");
+  assert.match(key!, /^[a-f0-9]{16}$/);
+
+  // Mixed arrays (bare string turn followed by a structured item) must only map the
+  // bare-string entries, leaving object items untouched.
+  const mixed = stick.normalizeStickinessMessages({
+    input: ["plain turn", { role: "user", content: [{ type: "input_text", text: "hi" }] }],
+  });
+  assert.deepEqual(mixed?.[0], { role: "user", content: "plain turn" });
+  assert.deepEqual(mixed?.[1], { role: "user", content: [{ type: "input_text", text: "hi" }] });
+});
+
+test("an .input-shaped Responses body yields a stable, non-null stickiness key", () => {
+  const body = responsesBody(rrCombo("k"), "First turn of the conversation");
+  // Old behavior: deriveMessageHash(body.messages) is null (bug).
+  assert.equal(
+    stick.deriveMessageHash(body.messages as never),
+    null,
+    "body.messages is absent on the Responses API — the old key source is null"
+  );
+  // New behavior: the normalized view produces a stable key.
+  const key1 = stick.deriveMessageHash(stick.normalizeStickinessMessages(body));
+  const key2 = stick.deriveMessageHash(stick.normalizeStickinessMessages(body));
+  assert.ok(key1 !== null, "normalized Responses body must yield a key");
+  assert.match(key1!, /^[a-f0-9]{16}$/);
+  assert.equal(key1, key2, "stable across calls");
+});
+
+test("round-robin: a sessionless Responses-API conversation re-pins across turns (#7270)", async () => {
+  const combo = rrCombo("rr-resp-stick");
+  const conns: string[] = [];
+  for (let turn = 0; turn < 5; turn++) {
+    conns.push(
+      await dispatchConnection(combo, responsesBody(combo, "Refactor the streaming handler."))
+    );
+  }
+  // Turns 2..5 must all reuse turn 1's connection. Before the fix the RR combo rotates
+  // (conn-A → conn-B → conn-C → …) because the .input key resolved to null → set size 3.
+  assert.equal(
+    new Set(conns.slice(1)).size,
+    1,
+    `turns 2..5 must stick to one connection, got: ${conns.join(", ")}`
+  );
+  assert.equal(conns[1], conns[0], "turn 2 must reuse turn 1's connection");
+});
+
+// Responses-API shape variant: `.input` is an array of PLAIN STRING items (each string
+// is shorthand for a user message), not an array of message objects. Before this fix,
+// normalizeStickinessMessages cast the array straight through unmapped, so
+// deriveMessageHash never found a `role === "user"` entry and stickiness stayed
+// fail-open for this narrower wire shape too.
+function responsesBodyPlainStringArray(combo: Record<string, unknown>, firstMessage: string) {
+  return {
+    model: combo.name,
+    input: [firstMessage],
+    stream: false,
+  };
+}
+
+test("round-robin: a sessionless Responses-API conversation with .input as a plain-string array re-pins across turns (#7270 bare-string-array gap)", async () => {
+  const combo = rrCombo("rr-resp-stick-strarr");
+  const conns: string[] = [];
+  for (let turn = 0; turn < 5; turn++) {
+    conns.push(
+      await dispatchConnection(
+        combo,
+        responsesBodyPlainStringArray(combo, "Refactor the streaming handler (string-array input).")
+      )
+    );
+  }
+  // Turns 2..5 must all reuse turn 1's connection. Before the fix, the bare-string
+  // .input array items yield a null key → the RR combo rotates (conn-A → conn-B → conn-C).
+  assert.equal(
+    new Set(conns.slice(1)).size,
+    1,
+    `turns 2..5 must stick to one connection, got: ${conns.join(", ")}`
+  );
+  assert.equal(conns[1], conns[0], "turn 2 must reuse turn 1's connection");
+});
+
+test("round-robin: distinct Responses-API conversations still spread (#7270 spreading guard)", async () => {
+  const combo = rrCombo("rr-resp-spread");
+  const hist: Record<string, number> = {};
+  for (let i = 0; i < 6; i++) {
+    const conn = await dispatchConnection(
+      combo,
+      responsesBody(combo, `conversation number ${i} — distinct first turn`)
+    );
+    hist[conn] = (hist[conn] || 0) + 1;
+  }
+  assert.ok(
+    Object.keys(hist).length > 1,
+    `distinct conversations must spread across connections, got: ${JSON.stringify(hist)}`
+  );
+});
diff --git a/tests/unit/combo/combo-target-exhaustion.test.ts b/tests/unit/combo/combo-target-exhaustion.test.ts
index c6b03691878..4ac95bcbd35 100644
--- a/tests/unit/combo/combo-target-exhaustion.test.ts
+++ b/tests/unit/combo/combo-target-exhaustion.test.ts
@@ -93,6 +93,25 @@ test("connection-level 5xx with a connectionId poisons exhaustedConnections (#17
   assert.equal(s.exhaustedProviders.size, 0);
 });
 
+test("request-scoped failed-response 502 does not exhaust the connection", () => {
+  const s = sets();
+  const exhausted = applyComboTargetExhaustion(target(), {
+    ...baseOpts,
+    result: { status: 502, headers: null },
+    fallbackResult: {},
+    errorText: "upstream reported a failed response without usable output",
+    structuredError: {
+      code: "upstream_response_failed",
+      type: "upstream_response_error",
+    },
+    sets: s,
+  });
+
+  assert.equal(exhausted, false);
+  assert.equal(s.exhaustedProviders.size, 0);
+  assert.equal(s.exhaustedConnections.size, 0);
+});
+
 test("connection-level 5xx without a connectionId poisons exhaustedProviders (#1731)", () => {
   const s = sets();
   applyComboTargetExhaustion(target({ connectionId: null }), {
diff --git a/tests/unit/compression/body-adapter.test.ts b/tests/unit/compression/body-adapter.test.ts
index be584bf2113..4bc3103d3d9 100644
--- a/tests/unit/compression/body-adapter.test.ts
+++ b/tests/unit/compression/body-adapter.test.ts
@@ -78,6 +78,92 @@ describe("compression body adapter", () => {
     assert.equal(input[0].call_id, "call_1");
   });
 
+  it("applies RTK compression to Codex custom_tool_call_output items", () => {
+    const repeatedOutput = Array.from({ length: 20 }, () => "same noisy line").join("\n");
+    const body = {
+      input: [
+        {
+          type: "custom_tool_call_output",
+          call_id: "call_patch_1",
+          output: repeatedOutput,
+        },
+      ],
+    };
+
+    const result = applyRtkCompression(body);
+
+    assert.equal(result.compressed, true);
+    assert.ok(!("messages" in result.body), "Responses body must not leak synthetic messages");
+    const input = result.body.input as typeof body.input;
+    assert.equal(input[0].type, "custom_tool_call_output");
+    assert.equal(input[0].call_id, "call_patch_1");
+    assert.match(input[0].output, /\[rtk:dropped/);
+  });
+
+  it("preserves wrapped custom tool output metadata while compressing its output", () => {
+    const repeatedOutput = Array.from({ length: 20 }, () => "same noisy line").join("\n");
+    const body = {
+      input: [
+        {
+          type: "custom_tool_call_output",
+          call_id: "call_exec_1",
+          output: JSON.stringify({ output: repeatedOutput, metadata: { exitCode: 0 } }),
+        },
+      ],
+    };
+
+    const result = applyRtkCompression(body);
+    const input = result.body.input as typeof body.input;
+    const restoredOutput = JSON.parse(input[0].output) as {
+      output: string;
+      metadata: { exitCode: number };
+    };
+
+    assert.equal(result.compressed, true);
+    assert.match(restoredOutput.output, /\[rtk:dropped/);
+    assert.deepEqual(restoredOutput.metadata, { exitCode: 0 });
+  });
+
+  it("restores custom tool output to content when that was the source field", () => {
+    const repeatedOutput = Array.from({ length: 20 }, () => "same noisy line").join("\n");
+    const body = {
+      input: [
+        {
+          type: "custom_tool_call_output",
+          call_id: "call_exec_2",
+          content: repeatedOutput,
+        },
+      ],
+    };
+
+    const result = applyRtkCompression(body);
+    const input = result.body.input as Array<Record<string, unknown>>;
+
+    assert.equal(result.compressed, true);
+    assert.match(input[0].content as string, /\[rtk:dropped/);
+    assert.ok(!("output" in input[0]), "restore must not add a conflicting output field");
+  });
+
+  it("restores function call output to content when that was the source field", () => {
+    const repeatedOutput = Array.from({ length: 20 }, () => "same noisy line").join("\n");
+    const body = {
+      input: [
+        {
+          type: "function_call_output",
+          call_id: "call_2",
+          content: repeatedOutput,
+        },
+      ],
+    };
+
+    const result = applyRtkCompression(body);
+    const input = result.body.input as Array<Record<string, unknown>>;
+
+    assert.equal(result.compressed, true);
+    assert.match(input[0].content as string, /\[rtk:dropped/);
+    assert.ok(!("output" in input[0]), "restore must not add a conflicting output field");
+  });
+
   it("restores compressed array output on Responses function_call_output items", () => {
     const repeatedOutput = Array.from({ length: 20 }, () => "same noisy line").join("\n");
     const body = {
diff --git a/tests/unit/compression/ccr-cross-tenant.test.ts b/tests/unit/compression/ccr-cross-tenant.test.ts
index c7361772565..044002a5c69 100644
--- a/tests/unit/compression/ccr-cross-tenant.test.ts
+++ b/tests/unit/compression/ccr-cross-tenant.test.ts
@@ -184,9 +184,9 @@ describe("ccr security: [MEDIUM] scoped retrieval feedback (shouldSkipCompressio
   });
 });
 
-// ─── bounded memory (FIFO eviction) ──────────────────────────────────────────
+// ─── bounded memory (LRU eviction) ───────────────────────────────────────────
 
-describe("ccr security: [MEDIUM] bounded memory (FIFO eviction)", () => {
+describe("ccr security: [MEDIUM] bounded memory (LRU eviction)", () => {
   beforeEach(() => resetCcrStore());
 
   it("MAX_CCR_ENTRIES is exported and positive", () => {
@@ -194,7 +194,7 @@ describe("ccr security: [MEDIUM] bounded memory (FIFO eviction)", () => {
     assert.ok(MAX_CCR_ENTRIES > 0, "MAX_CCR_ENTRIES must be positive");
   });
 
-  it("FIFO eviction: oldest entry is evicted when cap is reached", () => {
+  it("LRU eviction: least-recently-used entry is evicted when cap is reached", () => {
     const principal = "evictionTest";
 
     // Fill to exactly cap with unique blocks
@@ -205,7 +205,10 @@ describe("ccr security: [MEDIUM] bounded memory (FIFO eviction)", () => {
       storeBlock(`eviction block ${i} ${"y".repeat(30)}`, principal);
     }
 
-    // Confirm first block is present before overflow
+    const secondText = `eviction block 1 ${"y".repeat(30)}`;
+    const secondHash = contentHash(secondText);
+
+    // Touch the first block so the second one becomes least recently used.
     assert.equal(
       retrieveBlock(firstHash, principal),
       firstText,
@@ -215,12 +218,13 @@ describe("ccr security: [MEDIUM] bounded memory (FIFO eviction)", () => {
     // Insert one more unique block — should evict the first (oldest) entry
     storeBlock(`eviction overflow block ${"z".repeat(50)}`, principal);
 
-    // First block should now be gone
+    // The untouched second block should now be gone; the recently used first remains.
     assert.equal(
-      retrieveBlock(firstHash, principal),
+      retrieveBlock(secondHash, principal),
       null,
-      "[MEDIUM] FIFO eviction: oldest block must be evicted when cap is reached"
+      "[MEDIUM] LRU eviction: least-recently-used block must be evicted when cap is reached"
     );
+    assert.equal(retrieveBlock(firstHash, principal), firstText);
   });
 
   it("store does not grow unboundedly beyond MAX_CCR_ENTRIES distinct principals", () => {
diff --git a/tests/unit/compression/ccr-mcp-integration.test.ts b/tests/unit/compression/ccr-mcp-integration.test.ts
new file mode 100644
index 00000000000..4a7eeb1c421
--- /dev/null
+++ b/tests/unit/compression/ccr-mcp-integration.test.ts
@@ -0,0 +1,220 @@
+import { describe, it, beforeEach } from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+
+process.env.DATA_DIR = mkdtempSync(join(tmpdir(), "omniroute-ccr-mcp-"));
+
+const ccr = await import("../../../open-sse/services/compression/engines/ccr/index.ts");
+const tools = await import("../../../open-sse/mcp-server/tools/compressionTools.ts");
+const schemas = await import("../../../open-sse/mcp-server/schemas/tools.ts");
+const { MCP_TOOL_SCOPES } = await import("../../../src/shared/constants/mcpScopes.ts");
+
+const readExtra = {
+  authInfo: { clientId: "tenant-a", scopes: ["read:compression"] },
+};
+const writeExtra = {
+  authInfo: { clientId: "tenant-a", scopes: ["write:compression"] },
+};
+
+describe("CCR in-memory store contract", () => {
+  beforeEach(() => ccr.resetCcrStore());
+
+  it("stores metadata, references, pagination, deletion, and caller-scoped stats", () => {
+    const first = ccr.tryStoreBlock("first\nblock", "tenant-a", {
+      contentType: "text/markdown",
+      source: "mcp",
+      ttlSeconds: 120,
+      now: 1_000,
+    });
+    const second = ccr.tryStoreBlock("second block", "tenant-a", {
+      source: "mcp",
+      now: 2_000,
+    });
+    assert.equal(first.stored, true);
+    assert.equal(second.stored, true);
+    if (!first.stored || !second.stored) return;
+
+    assert.deepEqual(ccr.buildCcrReference(first.hash, first.metadata.chars), {
+      hash: first.hash,
+      uri: `ccr://${first.hash}`,
+      marker: `[CCR retrieve hash=${first.hash} chars=11]`,
+    });
+    assert.equal(first.metadata.contentType, "text/markdown");
+    assert.equal(first.metadata.lines, 2);
+    assert.equal("content" in first.metadata, false);
+
+    const page = ccr.listCcrBlocks("tenant-a", { limit: 1, now: 2_000 });
+    assert.equal(page.total, 2);
+    assert.equal(page.entries[0].hash, second.hash);
+    assert.equal(page.hasMore, true);
+    assert.equal(ccr.getCcrStoreStats("tenant-a", 2_000).entries, 2);
+
+    assert.equal(ccr.deleteCcrBlock(first.hash, "tenant-b", 2_000), false);
+    assert.equal(ccr.deleteCcrBlock(first.hash, "tenant-a", 2_000), true);
+    assert.equal(ccr.inspectCcrBlock(first.hash, "tenant-a", 2_000), null);
+  });
+
+  it("expires entries deterministically and counts expiration eviction", () => {
+    const stored = ccr.tryStoreBlock("expires", "tenant-a", { ttlSeconds: 60, now: 1_000 });
+    assert.equal(stored.stored, true);
+    if (!stored.stored) return;
+    assert.ok(ccr.inspectCcrBlock(stored.hash, "tenant-a", 60_999));
+    assert.equal(ccr.inspectCcrBlock(stored.hash, "tenant-a", 61_000), null);
+    assert.equal(ccr.getCcrStoreStats("tenant-a", 61_000).lifecycle.expiredEvictions, 1);
+  });
+
+  it("expires only the requested entry on hot-path reads", () => {
+    const expired = ccr.tryStoreBlock("expired", "tenant-a", { ttlSeconds: 60, now: 1_000 });
+    const active = ccr.tryStoreBlock("active", "tenant-b", { ttlSeconds: 120, now: 1_000 });
+    assert.equal(expired.stored, true);
+    assert.equal(active.stored, true);
+    if (!expired.stored || !active.stored) return;
+
+    assert.equal(ccr.retrieveBlock(active.hash, "tenant-b", 61_000), "active");
+    assert.equal(ccr.getCcrStoreStats("tenant-a", 60_999).lifecycle.expiredEvictions, 0);
+    assert.equal(ccr.inspectCcrBlock(expired.hash, "tenant-a", 61_000), null);
+    assert.equal(ccr.getCcrStoreStats("tenant-a", 61_000).lifecycle.expiredEvictions, 1);
+  });
+
+  it("rejects blocks above the UTF-8 byte limit", () => {
+    const result = ccr.tryStoreBlock("ü".repeat(ccr.MAX_CCR_BLOCK_BYTES), "tenant-a");
+    assert.deepEqual(result.stored, false);
+    if (!result.stored) assert.equal(result.reason, "block_too_large");
+    assert.equal(ccr.getCcrStoreStats("tenant-a").lifecycle.rejectedStores, 1);
+  });
+
+  it("evicts least-recently-used owner entries to enforce the principal byte budget", () => {
+    const hashes: string[] = [];
+    for (let i = 0; i < 9; i++) {
+      const prefix = `${i}:`;
+      const content = prefix + "x".repeat(ccr.MAX_CCR_BLOCK_BYTES - prefix.length);
+      const result = ccr.tryStoreBlock(content, "tenant-a", { now: i + 1 });
+      assert.equal(result.stored, true);
+      if (result.stored) hashes.push(result.hash);
+    }
+    const stats = ccr.getCcrStoreStats("tenant-a", 10);
+    assert.ok(stats.bytes <= ccr.MAX_CCR_PRINCIPAL_BYTES);
+    assert.equal(stats.entries, 8);
+    assert.equal(ccr.inspectCcrBlock(hashes[0], "tenant-a", 10), null);
+    assert.ok(stats.lifecycle.capacityEvictions >= 1);
+  });
+
+  it("releases principal byte budget when an entry is deleted", () => {
+    const content = "x".repeat(ccr.MAX_CCR_BLOCK_BYTES);
+    const stored = ccr.tryStoreBlock(content, "tenant-a", { now: 1 });
+    assert.equal(stored.stored, true);
+    if (!stored.stored) return;
+    assert.equal(ccr.getCcrStoreStats("tenant-a", 1).bytes, ccr.MAX_CCR_BLOCK_BYTES);
+    assert.equal(ccr.deleteCcrBlock(stored.hash, "tenant-a", 1), true);
+    assert.equal(ccr.getCcrStoreStats("tenant-a", 1).bytes, 0);
+  });
+
+  it("keeps all read, list, inspect, delete, and stats operations principal-isolated", () => {
+    const result = ccr.tryStoreBlock("tenant-a secret", "tenant-a", { source: "mcp" });
+    assert.equal(result.stored, true);
+    if (!result.stored) return;
+    assert.equal(ccr.retrieveBlock(result.hash, "tenant-b"), null);
+    assert.equal(ccr.inspectCcrBlock(result.hash, "tenant-b"), null);
+    assert.equal(ccr.listCcrBlocks("tenant-b").total, 0);
+    assert.equal(ccr.deleteCcrBlock(result.hash, "tenant-b"), false);
+    assert.equal(ccr.getCcrStoreStats("tenant-b").entries, 0);
+    assert.deepEqual(ccr.getCcrStoreStats("tenant-b").lifecycle, {
+      expiredEvictions: 0,
+      capacityEvictions: 0,
+      rejectedStores: 0,
+    });
+    assert.equal(ccr.retrieveBlock(result.hash, "tenant-a"), "tenant-a secret");
+  });
+});
+
+describe("CCR MCP contracts and handlers", () => {
+  beforeEach(() => ccr.resetCcrStore());
+
+  it("registers all six tools canonically with least-privilege scopes", () => {
+    const expected = {
+      omniroute_ccr_store: ["write:compression"],
+      omniroute_ccr_retrieve: ["read:compression"],
+      omniroute_ccr_inspect: ["read:compression"],
+      omniroute_ccr_list: ["read:compression"],
+      omniroute_ccr_delete: ["write:compression"],
+      omniroute_ccr_stats: ["read:compression"],
+    } as const;
+    for (const [name, scopes] of Object.entries(expected)) {
+      assert.ok(schemas.MCP_TOOL_MAP[name], `${name} must exist in MCP_TOOL_MAP`);
+      assert.deepEqual(schemas.MCP_TOOL_MAP[name].scopes, scopes);
+      assert.deepEqual(MCP_TOOL_SCOPES[name], scopes);
+    }
+  });
+
+  it("validates the MCP store limit in UTF-8 bytes instead of JavaScript characters", () => {
+    const multibyte = "ü".repeat(ccr.MAX_CCR_BLOCK_BYTES / 2 + 1);
+    assert.equal(schemas.ccrStoreInput.safeParse({ content: multibyte }).success, false);
+    const result = ccr.tryStoreBlock(multibyte, "tenant-a", { source: "mcp" });
+    assert.equal(result.stored, false);
+    if (!result.stored) assert.equal(result.reason, "block_too_large");
+  });
+
+  it("stores through MCP and exposes metadata without content in inspect/list", async () => {
+    const stored = await tools.handleCcrStoreTool(
+      { content: "sensitive body", contentType: "text/plain", ttlSeconds: 120 },
+      writeExtra
+    );
+    assert.equal(stored.stored, true);
+    if (!stored.stored) return;
+
+    const inspected = await tools.handleCcrInspectTool({ hash: stored.reference.hash }, readExtra);
+    assert.equal(inspected.found, true);
+    assert.equal(JSON.stringify(inspected).includes("sensitive body"), false);
+    const listed = await tools.handleCcrListTool({ limit: 10 }, readExtra);
+    assert.equal(listed.total, 1);
+    assert.equal(JSON.stringify(listed).includes("sensitive body"), false);
+    assert.equal((await tools.handleCcrStatsTool({}, readExtra)).entries, 1);
+    assert.deepEqual(await tools.handleCcrDeleteTool({ hash: stored.reference.hash }, writeExtra), {
+      deleted: true,
+    });
+  });
+
+  it("does not include content in the CCR store audit input", () => {
+    const secret = "never-persist-this-content";
+    const auditInput = tools.buildCcrStoreAuditInput({
+      content: secret,
+      contentType: "text/plain",
+    });
+    assert.equal(JSON.stringify(auditInput).includes(secret), false);
+    assert.deepEqual(auditInput, {
+      bytes: Buffer.byteLength(secret, "utf8"),
+      contentType: "text/plain",
+      ttlSeconds: undefined,
+    });
+  });
+
+  it("refuses oversized full MCP responses and recommends ranged modes", async () => {
+    const content = "large\n".repeat(50_000);
+    const stored = ccr.tryStoreBlock(content, "tenant-a", { source: "mcp" });
+    assert.equal(stored.stored, true);
+    if (!stored.stored) return;
+    const full = await tools.handleCcrRetrieveTool({ hash: stored.hash }, readExtra);
+    assert.equal(full.found, true);
+    assert.equal("content" in full, false);
+    assert.equal("tooLargeForFull" in full && full.tooLargeForFull, true);
+
+    const head = await tools.handleCcrRetrieveTool(
+      { hash: stored.hash, mode: "head", n: 2 },
+      readExtra
+    );
+    assert.equal(head.found, true);
+    assert.equal("content" in head ? head.content : undefined, "large\nlarge");
+  });
+
+  it("does not reveal another principal's block through MCP handlers", async () => {
+    const stored = ccr.tryStoreBlock("only tenant a", "tenant-a", { source: "mcp" });
+    assert.equal(stored.stored, true);
+    if (!stored.stored) return;
+    const other = { authInfo: { clientId: "tenant-b", scopes: ["read:compression"] } };
+    assert.equal((await tools.handleCcrRetrieveTool({ hash: stored.hash }, other)).found, false);
+    assert.equal((await tools.handleCcrInspectTool({ hash: stored.hash }, other)).found, false);
+    assert.equal((await tools.handleCcrListTool({}, other)).total, 0);
+  });
+});
diff --git a/tests/unit/compression/compression-preview-api.test.ts b/tests/unit/compression/compression-preview-api.test.ts
index 9a34952d157..c751bafa7c7 100644
--- a/tests/unit/compression/compression-preview-api.test.ts
+++ b/tests/unit/compression/compression-preview-api.test.ts
@@ -8,6 +8,10 @@ import {
 
 describe("compression preview API contract", () => {
   it("accepts RTK and stacked preview payloads", () => {
+    assert.equal(
+      PreviewCompressionConfigSchema.safeParse({ liveZone: { enabled: true } }).success,
+      true
+    );
     assert.equal(
       PreviewRequestSchema.safeParse({
         messages: [{ role: "tool", content: "same\nsame\nsame" }],
diff --git a/tests/unit/compression/db.test.ts b/tests/unit/compression/db.test.ts
index 2df38957138..20ba863dbd8 100644
--- a/tests/unit/compression/db.test.ts
+++ b/tests/unit/compression/db.test.ts
@@ -52,6 +52,7 @@ describe("getCompressionSettings", () => {
     assert.equal(settings.cacheMinutes, 5);
     assert.equal(settings.preserveSystemPrompt, true);
     assert.equal(settings.preserveSystemPromptMode, "always");
+    assert.deepEqual(settings.liveZone, { enabled: false });
     assert.deepEqual(settings.comboOverrides, {});
     assert.equal(settings.ultra?.enabled, false);
     assert.equal(settings.ultra?.compressionRate, 0.5);
@@ -105,6 +106,16 @@ describe("updateCompressionSettings", () => {
     await updateCompressionSettings({ autoTriggerTokens: 0 } as any);
   });
 
+  it("round-trips cache-aligned live-zone compression", async () => {
+    await updateCompressionSettings({ liveZone: { enabled: true } });
+    let settings = await getCompressionSettings();
+    assert.deepEqual(settings.liveZone, { enabled: true });
+
+    await updateCompressionSettings({ liveZone: { enabled: false } });
+    settings = await getCompressionSettings();
+    assert.deepEqual(settings.liveZone, { enabled: false });
+  });
+
   it("updates multiple settings at once", async () => {
     await updateCompressionSettings({
       enabled: true,
diff --git a/tests/unit/compression/live-zone.test.ts b/tests/unit/compression/live-zone.test.ts
new file mode 100644
index 00000000000..fbddf3e7613
--- /dev/null
+++ b/tests/unit/compression/live-zone.test.ts
@@ -0,0 +1,254 @@
+import { beforeEach, describe, it } from "node:test";
+import assert from "node:assert/strict";
+
+import {
+  applyLiveZoneCompression,
+  getLiveZoneCacheStats,
+  resetLiveZoneCache,
+} from "../../../open-sse/services/compression/liveZone.ts";
+import type { CompressionResult } from "../../../open-sse/services/compression/types.ts";
+
+function result(body: Record<string, unknown>, compressed = true): CompressionResult {
+  return {
+    body,
+    compressed,
+    stats: {
+      originalTokens: 100,
+      compressedTokens: compressed ? 50 : 100,
+      savingsPercent: compressed ? 50 : 0,
+      techniquesUsed: compressed ? ["test"] : [],
+      mode: "rtk",
+      timestamp: Date.now(),
+    },
+  };
+}
+
+function toolCompressor(calls: Array<Record<string, unknown>>) {
+  return async (body: Record<string, unknown>): Promise<CompressionResult> => {
+    calls.push(structuredClone(body));
+    const field = Array.isArray(body.messages) ? "messages" : "input";
+    const items = body[field] as Array<Record<string, unknown>>;
+    return result({
+      ...body,
+      [field]: items.map((item) =>
+        item.role === "tool" ? { ...item, content: `compressed:${String(item.content)}` } : item
+      ),
+    });
+  };
+}
+
+const options = {
+  principalId: "key-1",
+  sessionId: "session-1",
+  variant: { mode: "rtk", config: {} },
+};
+
+beforeEach(() => resetLiveZoneCache());
+
+describe("cache-aligned live-zone compression", () => {
+  it("compresses normally once, then sends only appended items through the engine", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = toolCompressor(calls);
+    const first = {
+      messages: [
+        { role: "system", content: "stable" },
+        { role: "user", content: "run tests" },
+        { role: "tool", content: "100 noisy lines" },
+      ],
+    };
+    const firstResult = await applyLiveZoneCompression(first, options, compress);
+    const frozenBytes = JSON.stringify((firstResult.body.messages as unknown[]).slice(0, 3));
+
+    const second = {
+      messages: [
+        ...first.messages,
+        { role: "assistant", content: "Need the failing test" },
+        { role: "tool", content: "50 more noisy lines" },
+      ],
+    };
+    const secondResult = await applyLiveZoneCompression(second, options, compress);
+
+    assert.equal(calls.length, 2);
+    assert.deepEqual(calls[1].messages, [second.messages[4]]);
+    assert.equal(
+      JSON.stringify((secondResult.body.messages as unknown[]).slice(0, 3)),
+      frozenBytes,
+      "the provider-facing prefix must remain byte-stable"
+    );
+    assert.equal(
+      (secondResult.body.messages as Array<{ content: string }>)[3].content,
+      "Need the failing test"
+    );
+    assert.equal(
+      (secondResult.body.messages as Array<{ content: string }>)[4].content,
+      "compressed:50 more noisy lines"
+    );
+    assert.deepEqual(secondResult.stats?.liveZone, {
+      cacheHit: true,
+      frozenItems: 3,
+      liveItems: 2,
+    });
+  });
+
+  it("keeps transformed top-level prefix fields stable and skips unchanged requests", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = async (body: Record<string, unknown>): Promise<CompressionResult> => {
+      calls.push(structuredClone(body));
+      return result({
+        ...body,
+        instructions: `${String(body.instructions)}:${calls.length}`,
+      });
+    };
+    const first = {
+      instructions: "stable instructions",
+      tools: [{ type: "function", name: "run" }],
+      messages: [{ role: "user", content: "hello" }],
+    };
+    const firstResult = await applyLiveZoneCompression(first, options, compress);
+    const unchangedResult = await applyLiveZoneCompression(first, options, compress);
+    const appendedResult = await applyLiveZoneCompression(
+      { ...first, messages: [...first.messages, { role: "tool", content: "new output" }] },
+      options,
+      compress
+    );
+
+    assert.equal(calls.length, 2, "an unchanged request must not run the compressor again");
+    assert.equal(unchangedResult.body.instructions, firstResult.body.instructions);
+    assert.equal(appendedResult.body.instructions, firstResult.body.instructions);
+    assert.deepEqual(appendedResult.body.tools, firstResult.body.tools);
+    assert.deepEqual(calls[1].messages, [{ role: "tool", content: "new output" }]);
+  });
+
+  it("fails open to full compression when any raw prefix item changes", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = toolCompressor(calls);
+    await applyLiveZoneCompression(
+      { messages: [{ role: "user", content: "original" }] },
+      options,
+      compress
+    );
+    await applyLiveZoneCompression(
+      {
+        messages: [
+          { role: "user", content: "edited" },
+          { role: "tool", content: "new" },
+        ],
+      },
+      options,
+      compress
+    );
+
+    assert.equal((calls[1].messages as unknown[]).length, 2);
+  });
+
+  it("fails open when stable system or tool metadata changes", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = toolCompressor(calls);
+    const messages = [{ role: "user", content: "same" }];
+    await applyLiveZoneCompression(
+      { system_instruction: "first", tools: [{ name: "one" }], messages },
+      options,
+      compress
+    );
+    await applyLiveZoneCompression(
+      {
+        system_instruction: "changed",
+        tools: [{ name: "two" }],
+        messages: [...messages, { role: "tool", content: "new" }],
+      },
+      options,
+      compress
+    );
+
+    assert.equal((calls[1].messages as unknown[]).length, 2);
+  });
+
+  it("isolates cached prefixes by principal and compression variant", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = toolCompressor(calls);
+    const body = { messages: [{ role: "user", content: "same" }] };
+    await applyLiveZoneCompression(body, options, compress);
+    await applyLiveZoneCompression(
+      { messages: [...body.messages, { role: "tool", content: "other principal" }] },
+      { ...options, principalId: "key-2" },
+      compress
+    );
+    await applyLiveZoneCompression(
+      { messages: [...body.messages, { role: "tool", content: "other mode" }] },
+      { ...options, variant: { mode: "caveman", config: {} } },
+      compress
+    );
+
+    assert.equal((calls[1].messages as unknown[]).length, 2);
+    assert.equal((calls[2].messages as unknown[]).length, 2);
+    assert.equal(getLiveZoneCacheStats().entries, 3);
+  });
+
+  it("supports Responses input arrays and bypasses caching without an authenticated principal", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = toolCompressor(calls);
+    const first = { input: [{ role: "user", content: "hello" }] };
+    await applyLiveZoneCompression(first, options, compress);
+    await applyLiveZoneCompression(
+      { input: [...first.input, { role: "tool", content: "response output" }] },
+      options,
+      compress
+    );
+    assert.deepEqual(calls[1].input, [{ role: "tool", content: "response output" }]);
+
+    await applyLiveZoneCompression(first, { ...options, principalId: undefined }, compress);
+    await applyLiveZoneCompression(first, { ...options, principalId: undefined }, compress);
+    assert.equal(getLiveZoneCacheStats().entries, 1);
+  });
+
+  it("compresses only new Responses API tool outputs", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = toolCompressor(calls);
+    const first = { input: [{ type: "message", role: "user", content: "hello" }] };
+    await applyLiveZoneCompression(first, options, compress);
+    const functionCall = { type: "function_call", call_id: "call-1", name: "run" };
+    const output = { type: "function_call_output", call_id: "call-1", output: "noisy" };
+    const result = await applyLiveZoneCompression(
+      { input: [...first.input, functionCall, output] },
+      options,
+      compress
+    );
+
+    assert.deepEqual(calls[1].input, [output]);
+    assert.deepEqual((result.body.input as unknown[])[1], functionCall);
+  });
+
+  it("compresses tool_result roles and types in the live zone", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = toolCompressor(calls);
+    const first = { input: [{ role: "user", content: "hello" }] };
+    await applyLiveZoneCompression(first, options, compress);
+    const roleOutput = { role: "tool_result", content: "role output" };
+    const typeOutput = { type: "tool_result", content: "type output" };
+    await applyLiveZoneCompression(
+      { input: [...first.input, roleOutput, typeOutput] },
+      options,
+      compress
+    );
+
+    assert.deepEqual(calls[1].input, [roleOutput, typeOutput]);
+  });
+
+  it("bypasses live-zone reuse when a global hard budget is configured", async () => {
+    const calls: Array<Record<string, unknown>> = [];
+    const compress = toolCompressor(calls);
+    const hardBudgetOptions = {
+      ...options,
+      variant: { mode: "stacked", config: { targetTokens: 100 } },
+    };
+    const first = { messages: [{ role: "user", content: "hello" }] };
+    await applyLiveZoneCompression(first, hardBudgetOptions, compress);
+    await applyLiveZoneCompression(
+      { messages: [...first.messages, { role: "tool", content: "new" }] },
+      hardBudgetOptions,
+      compress
+    );
+    assert.equal((calls[1].messages as unknown[]).length, 2);
+    assert.equal(getLiveZoneCacheStats().entries, 0);
+  });
+});
diff --git a/tests/unit/compression/rtk-line-filter.test.ts b/tests/unit/compression/rtk-line-filter.test.ts
index 3704c7e174b..a3ac5a067e7 100644
--- a/tests/unit/compression/rtk-line-filter.test.ts
+++ b/tests/unit/compression/rtk-line-filter.test.ts
@@ -36,4 +36,20 @@ describe("RTK line filters", () => {
     assert.ok(result.text.includes("On branch"));
     assert.ok(!result.text.includes("nothing added"));
   });
+
+  it("does not truncate within a combined RTK TOML head and tail limit", () => {
+    const filter = {
+      ...loadRtkFilters().find((item) => item.id === "generic-output")!,
+      id: "toml-head-tail",
+      sourceFormat: "rtk-toml-v1" as const,
+      rtkTomlHeadLines: 2,
+      rtkTomlTailLines: 2,
+    };
+
+    const result = applyLineFilter("first\nsecond\nthird", filter);
+
+    assert.equal(result.text, "first\nsecond\nthird");
+    assert.ok(!result.appliedRules.includes("toml-head-tail:rtk-head"));
+    assert.ok(!result.appliedRules.includes("toml-head-tail:rtk-head-tail"));
+  });
 });
diff --git a/tests/unit/compression/rtk-toml-compatibility.test.ts b/tests/unit/compression/rtk-toml-compatibility.test.ts
new file mode 100644
index 00000000000..cc5b9cacef0
--- /dev/null
+++ b/tests/unit/compression/rtk-toml-compatibility.test.ts
@@ -0,0 +1,313 @@
+import { afterEach, describe, it } from "node:test";
+import assert from "node:assert/strict";
+import crypto from "node:crypto";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+import { applyLineFilter } from "../../../open-sse/services/compression/engines/rtk/lineFilter.ts";
+import {
+  getRtkFilterCatalog,
+  getRtkFilterLoadDiagnostics,
+  loadRtkFilters,
+  matchRtkFilter,
+} from "../../../open-sse/services/compression/engines/rtk/filterLoader.ts";
+import {
+  installGlobalRtkTomlV1,
+  parseRtkTomlV1,
+  RtkTomlCompatibilityError,
+} from "../../../open-sse/services/compression/engines/rtk/tomlCompatibility.ts";
+
+const originalCwd = process.cwd();
+const originalDataDir = process.env.DATA_DIR;
+
+const SAMPLE = `schema_version = 1
+
+[filters.my-tool]
+description = "Compact my-tool output"
+match_command = "^my-tool\\\\s+build"
+strip_ansi = true
+strip_lines_matching = ["^noise", "^\\\\s*$"]
+replace = [{ pattern = "duration: ([0-9]+)ms", replacement = "t=$1ms" }]
+match_output = [{ pattern = "all good", message = "my-tool: ok", unless = "error" }]
+truncate_lines_at = 12
+head_lines = 2
+tail_lines = 1
+max_lines = 4
+on_empty = "my-tool: empty"
+
+[[tests.my-tool]]
+name = "filters and truncates"
+input = "noise banner\\nfirst line\\nsecond line\\nthird line\\nfourth line\\n"
+expected = "first line\\nsecond line\\n... (1 lines omitted)\\nfourth line"
+
+[[tests.my-tool]]
+name = "short circuits success"
+input = "all good"
+expected = "my-tool: ok"
+`;
+
+afterEach(() => {
+  process.chdir(originalCwd);
+  if (originalDataDir === undefined) delete process.env.DATA_DIR;
+  else process.env.DATA_DIR = originalDataDir;
+  loadRtkFilters({ refresh: true, customFiltersEnabled: false });
+});
+
+describe("RTK TOML schema v1 compatibility", () => {
+  it("reports a safe location for invalid TOML syntax", () => {
+    assert.throws(
+      () => parseRtkTomlV1("not = [valid"),
+      (error: unknown) =>
+        error instanceof RtkTomlCompatibilityError &&
+        error.publicMessage === "invalid TOML syntax (line 1, column 8)"
+    );
+  });
+
+  it("parses the official declarative fields and passes inline tests", () => {
+    const result = parseRtkTomlV1(SAMPLE);
+
+    assert.equal(result.schemaVersion, 1);
+    assert.equal(result.filters.length, 1);
+    assert.equal(result.outcomes.length, 2);
+    assert.ok(result.outcomes.every((outcome) => outcome.passed));
+    assert.equal(result.passed, true);
+    assert.deepEqual(result.filtersWithoutTests, []);
+    assert.equal(result.filters[0].sourceFormat, "rtk-toml-v1");
+  });
+
+  it("matches imported TOML before a detected builtin command type", () => {
+    const project = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-project-"));
+    process.chdir(project);
+    const rtkDir = path.join(project, ".rtk");
+    const filterPath = path.join(rtkDir, "filters.toml");
+    fs.mkdirSync(rtkDir, { recursive: true });
+    fs.writeFileSync(
+      filterPath,
+      `schema_version = 1
+[filters.git-status-custom]
+match_command = "^git\\\\s+status"
+strip_lines_matching = ["^noise"]
+`
+    );
+    fs.writeFileSync(
+      path.join(rtkDir, "trust.json"),
+      JSON.stringify({
+        filtersTomlSha256: crypto
+          .createHash("sha256")
+          .update(fs.readFileSync(filterPath))
+          .digest("hex"),
+      })
+    );
+
+    const filter = matchRtkFilter("noise\n M file.ts", "git status", { refresh: true });
+
+    assert.equal(filter?.id, "git-status-custom");
+    assert.equal(filter?.source, "project");
+  });
+
+  it("requires a separate trust hash for project TOML files", () => {
+    const project = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-project-"));
+    process.chdir(project);
+    const rtkDir = path.join(project, ".rtk");
+    fs.mkdirSync(rtkDir, { recursive: true });
+    fs.writeFileSync(path.join(rtkDir, "filters.toml"), SAMPLE);
+    fs.writeFileSync(
+      path.join(rtkDir, "trust.json"),
+      JSON.stringify({ filtersSha256: "wrong-key" })
+    );
+
+    const filters = loadRtkFilters({ refresh: true });
+    const diagnostics = getRtkFilterLoadDiagnostics();
+
+    assert.ok(!filters.some((filter) => filter.id === "my-tool"));
+    assert.ok(
+      diagnostics.some(
+        (entry) => entry.format === "rtk-toml-v1" && entry.message.includes("untrusted")
+      )
+    );
+  });
+
+  it("loads global TOML and exposes its source format in the catalog", () => {
+    const project = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-project-"));
+    const dataDir = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-data-"));
+    process.chdir(project);
+    process.env.DATA_DIR = dataDir;
+    fs.mkdirSync(path.join(dataDir, "rtk"), { recursive: true });
+    fs.writeFileSync(path.join(dataDir, "rtk", "filters.toml"), SAMPLE);
+
+    loadRtkFilters({ refresh: true });
+    const entry = getRtkFilterCatalog().find((filter) => filter.id === "my-tool");
+
+    assert.equal(entry?.source, "global");
+    assert.equal(entry?.sourceFormat, "rtk-toml-v1");
+  });
+
+  it("prioritizes project TOML over a higher-priority global JSON match", () => {
+    const project = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-project-"));
+    const dataDir = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-data-"));
+    process.chdir(project);
+    process.env.DATA_DIR = dataDir;
+
+    const projectDir = path.join(project, ".rtk");
+    fs.mkdirSync(projectDir, { recursive: true });
+    const projectPath = path.join(projectDir, "filters.toml");
+    fs.writeFileSync(
+      projectPath,
+      `schema_version = 1
+[filters.project-match]
+match_command = "^scope-tool"
+`
+    );
+    fs.writeFileSync(
+      path.join(projectDir, "trust.json"),
+      JSON.stringify({
+        filtersTomlSha256: crypto
+          .createHash("sha256")
+          .update(fs.readFileSync(projectPath))
+          .digest("hex"),
+      })
+    );
+
+    fs.mkdirSync(path.join(dataDir, "rtk"), { recursive: true });
+    fs.writeFileSync(
+      path.join(dataDir, "rtk", "filters.json"),
+      JSON.stringify({
+        id: "global-match",
+        label: "Global match",
+        description: "",
+        category: "generic",
+        priority: 100,
+        match: { commands: ["^scope-tool"], patterns: [], outputTypes: [] },
+      })
+    );
+
+    const filter = matchRtkFilter("output", "scope-tool", { refresh: true });
+    assert.equal(filter?.id, "project-match");
+  });
+
+  it("keeps project JSON ahead of a matching global TOML filter", () => {
+    const project = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-project-"));
+    const dataDir = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-data-"));
+    process.chdir(project);
+    process.env.DATA_DIR = dataDir;
+
+    const projectDir = path.join(project, ".rtk");
+    fs.mkdirSync(projectDir, { recursive: true });
+    const projectPath = path.join(projectDir, "filters.json");
+    fs.writeFileSync(
+      projectPath,
+      JSON.stringify({
+        id: "project-json-match",
+        label: "Project JSON match",
+        description: "",
+        category: "generic",
+        match: { commands: ["^scope-tool"], patterns: [], outputTypes: [] },
+        rules: {},
+        preserve: { errorPatterns: [], summaryPatterns: [] },
+        tests: [],
+      })
+    );
+    fs.writeFileSync(
+      path.join(projectDir, "trust.json"),
+      JSON.stringify({
+        filtersSha256: crypto
+          .createHash("sha256")
+          .update(fs.readFileSync(projectPath))
+          .digest("hex"),
+      })
+    );
+
+    fs.mkdirSync(path.join(dataDir, "rtk"), { recursive: true });
+    fs.writeFileSync(
+      path.join(dataDir, "rtk", "filters.toml"),
+      `schema_version = 1
+[filters.global-toml-match]
+match_command = "^scope-tool"
+`
+    );
+
+    const filter = matchRtkFilter("output", "scope-tool", { refresh: true });
+    assert.equal(filter?.id, "project-json-match");
+  });
+
+  it("installs atomically, protects the file and refuses an implicit overwrite", () => {
+    const dataDir = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-rtk-toml-data-"));
+    process.env.DATA_DIR = dataDir;
+
+    const installed = installGlobalRtkTomlV1(SAMPLE);
+    const target = path.join(dataDir, "rtk", "filters.toml");
+
+    assert.equal(installed.installedPath, "rtk/filters.toml");
+    assert.equal(fs.readFileSync(target, "utf8"), SAMPLE);
+    assert.equal(fs.statSync(target).mode & 0o777, 0o600);
+    assert.throws(
+      () => installGlobalRtkTomlV1(SAMPLE),
+      (error: unknown) =>
+        error instanceof RtkTomlCompatibilityError && error.publicMessage.includes("already exists")
+    );
+  });
+
+  it("rejects unknown fields, unsafe regexes and failing inline tests", () => {
+    assert.throws(() =>
+      parseRtkTomlV1(`schema_version = 1
+[filters.bad]
+match_command = "^bad"
+run = "rm -rf /"
+`)
+    );
+    assert.throws(() =>
+      parseRtkTomlV1(`schema_version = 1
+[filters.bad]
+match_command = "(a+)+$"
+`)
+    );
+    const failed = parseRtkTomlV1(`schema_version = 1
+[filters.bad]
+match_command = "^bad"
+strip_lines_matching = ["^noise"]
+[[tests.bad]]
+name = "wrong"
+input = "noise\\nkept"
+expected = "different"
+`);
+    assert.equal(failed.passed, false);
+    assert.throws(() =>
+      installGlobalRtkTomlV1(`schema_version = 1
+[filters.bad]
+match_command = "^bad"
+strip_lines_matching = ["^noise"]
+[[tests.bad]]
+name = "wrong"
+input = "noise\\nkept"
+expected = "different"
+`)
+    );
+  });
+
+  it("keeps filter_stderr as a documented proxy no-op", () => {
+    const result = parseRtkTomlV1(`schema_version = 1
+[filters.stderr-tool]
+match_command = "^stderr-tool"
+filter_stderr = true
+`);
+
+    assert.equal(result.filters[0].filterStderr, false);
+    assert.ok(result.warnings.some((warning) => warning.includes("filter_stderr")));
+    assert.equal(applyLineFilter("stderr: error", result.filters[0]).text, "stderr: error");
+  });
+
+  it("rejects non-schema fields in inline tests", () => {
+    assert.throws(() =>
+      parseRtkTomlV1(`schema_version = 1
+[filters.bad]
+match_command = "^bad"
+[[tests.bad]]
+name = "extra field"
+input = "input"
+expected = "input"
+command = "bad"
+`)
+    );
+  });
+});
diff --git a/tests/unit/db-core-temp-store-pragma.test.ts b/tests/unit/db-core-temp-store-pragma.test.ts
new file mode 100644
index 00000000000..3b23f670499
--- /dev/null
+++ b/tests/unit/db-core-temp-store-pragma.test.ts
@@ -0,0 +1,21 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omr-temp-store-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+process.env.API_KEY_SECRET = "test-temp-store-secret";
+
+const { getDbInstance } = await import("../../src/lib/db/core.ts");
+
+test("temp_store pragma is 2 (MEMORY) after initDb", () => {
+  // The database singleton should already have the pragma set by initDb().
+  // This test confirms the runtime respects the setting.
+  const db = getDbInstance();
+  const val: unknown = db.pragma("temp_store", { simple: true });
+  assert.ok(typeof val === "number");
+  // 2 = MEMORY (set by the new PRAGMA in initDb)
+  assert.equal(val, 2, "expected temp_store=2 (MEMORY) after initDb");
+});
diff --git a/tests/unit/db-providers-crud.test.ts b/tests/unit/db-providers-crud.test.ts
index 9c005fc179e..cde22ba7058 100644
--- a/tests/unit/db-providers-crud.test.ts
+++ b/tests/unit/db-providers-crud.test.ts
@@ -384,3 +384,98 @@ test("quota helpers zero stale windows and format countdowns", () => {
   assert.match(providersDb.formatResetCountdown(future), /1m \d+s/);
   assert.equal(providersDb.formatResetCountdown(past), null);
 });
+test("getProviderConnections supports authType filter and column projection", async () => {
+  const oauth = await providersDb.createProviderConnection({
+    provider: "openai",
+    authType: "oauth",
+    name: "OAuth Conn",
+    email: "user@example.com",
+    refreshToken: "rt_abc123",
+    isActive: true,
+  });
+  const apiKey = await providersDb.createProviderConnection({
+    provider: "openai",
+    authType: "apikey",
+    name: "API Key Conn",
+    apiKey: "sk-xyz",
+    isActive: true,
+  });
+
+  // authType filter
+  const oauthConns = await providersDb.getProviderConnections({ authType: "oauth" });
+  assert.equal(oauthConns.length, 1);
+  assert.equal(oauthConns[0].id, oauth.id);
+
+  const apiKeyConns = await providersDb.getProviderConnections({ authType: "apikey" });
+  assert.equal(apiKeyConns.length, 1);
+  assert.equal(apiKeyConns[0].id, apiKey.id);
+
+  // authType + isActive filter
+  const activeOAuth = await providersDb.getProviderConnections({
+    authType: "oauth",
+    isActive: true,
+  });
+  assert.equal(activeOAuth.length, 1);
+
+  // Column projection: only requested columns returned
+  const projected = await providersDb.getProviderConnections({ authType: "oauth" }, [
+    "id",
+    "provider",
+    "name",
+  ]);
+  assert.equal(projected.length, 1);
+  const keys = Object.keys(projected[0]);
+  // id, provider, name each appear in camelCase
+  assert.ok(keys.includes("id"));
+  assert.ok(keys.includes("provider"));
+  assert.ok(keys.includes("name"));
+  // decryptConnectionFields always adds undefined keys via explicit spread,
+  // so `in` checks can't distinguish "not projected" from "undefined value."
+  // Check value semantics instead.
+  assert.strictEqual(projected[0].refreshToken, undefined);
+  assert.strictEqual(projected[0].authType, undefined);
+
+  // Default (no columns param) returns all fields
+  const full = await providersDb.getProviderConnections({ authType: "oauth" });
+  const fullKeys = Object.keys(full[0]);
+  assert.ok(fullKeys.length > keys.length);
+  assert.ok("refreshToken" in full[0]);
+  assert.ok("authType" in full[0]);
+});
+
+test("getProviderConnections rejects column names outside the real provider_connections schema", async () => {
+  // The `columns` array is interpolated directly into the SQL SELECT clause,
+  // so it must be validated against an allowlist before use — otherwise it's
+  // a SQL-injection footgun for whichever future caller wires it to
+  // untrusted input. A single bogus column should reject the whole call.
+  await assert.rejects(
+    () => providersDb.getProviderConnections({}, ["not_a_real_column"]),
+    /invalid column/i
+  );
+
+  // A mix of valid + invalid columns must still reject (fail-closed, not a
+  // silent partial projection).
+  await assert.rejects(
+    () => providersDb.getProviderConnections({}, ["id", "provider; DROP TABLE provider_connections; --"]),
+    /invalid column/i
+  );
+
+  // The reserved SQL keyword "group" is a legitimate, allowlisted column and
+  // must still work (quoted internally so it doesn't collide with the SQL
+  // GROUP keyword).
+  await providersDb.createProviderConnection({
+    provider: "openai",
+    authType: "oauth",
+    name: "Group Column Conn",
+    email: "group-col@example.com",
+    refreshToken: "rt_group_col",
+    isActive: true,
+    group: "team-a",
+  });
+  const withGroup = await providersDb.getProviderConnections({ authType: "oauth" }, [
+    "id",
+    "group",
+  ]);
+  assert.equal(withGroup.length, 1);
+  assert.equal(withGroup[0].group, "team-a");
+});
diff --git a/tests/unit/db-schema-columns-split.test.ts b/tests/unit/db-schema-columns-split.test.ts
index f4f9643e9f8..0e89a5fa1ff 100644
--- a/tests/unit/db-schema-columns-split.test.ts
+++ b/tests/unit/db-schema-columns-split.test.ts
@@ -7,6 +7,7 @@ import assert from "node:assert/strict";
 import { tryOpenSync } from "../../src/lib/db/adapters/driverFactory.ts";
 import {
   ensureUsageHistoryColumns,
+  ensureProviderConnectionsColumns,
   hasColumn,
   hasTable,
   quoteIdentifier,
@@ -63,3 +64,24 @@ test("ensureUsageHistoryColumns adds missing columns and is idempotent", () => {
     db.close?.();
   }
 });
+
+test("ensureProviderConnectionsColumns repairs quota visibility with a visible default", () => {
+  const db = openMemoryDb();
+  try {
+    db.exec("CREATE TABLE provider_connections (id TEXT PRIMARY KEY, provider TEXT NOT NULL)");
+    assert.equal(hasColumn(db, "provider_connections", "quota_visible"), false);
+
+    ensureProviderConnectionsColumns(db);
+    assert.equal(hasColumn(db, "provider_connections", "quota_visible"), true);
+    const column = db
+      .prepare("PRAGMA table_info(provider_connections)")
+      .all()
+      .find((entry: { name?: string }) => entry.name === "quota_visible") as
+      { notnull?: number; dflt_value?: string } | undefined;
+    assert.equal(column?.notnull, 1);
+    assert.equal(column?.dflt_value, "1");
+    assert.doesNotThrow(() => ensureProviderConnectionsColumns(db));
+  } finally {
+    db.close?.();
+  }
+});
diff --git a/tests/unit/ddg-circuit-breaker-null-content-6999-7000.test.ts b/tests/unit/ddg-circuit-breaker-null-content-6999-7000.test.ts
new file mode 100644
index 00000000000..a5b97dfe733
--- /dev/null
+++ b/tests/unit/ddg-circuit-breaker-null-content-6999-7000.test.ts
@@ -0,0 +1,248 @@
+/**
+ * Issue #6999 — DuckDuckGo circuit breaker regression tests.
+ *
+ * The circuit breaker is a module-level singleton with in-process state.
+ * Tests directly import and call the exported executor, trigger failures
+ * to verify breaker opens, then let cooldown expire to verify it closes.
+ */
+
+import test, { describe } from "node:test";
+import assert from "node:assert/strict";
+
+/** Minimal shape accepted by DuckDuckGoWebExecutor.execute() */
+interface ExecutorRequest {
+  model: string;
+  messages: Array<{ role: string; content: unknown }>;
+  stream: boolean;
+  signal?: AbortSignal;
+}
+
+// The DuckDuckGoWebExecutor uses module-level mutable state (circuitBreaker)
+// so we import the module fresh and interact with execute() to verify
+// circuit breaker behavior.
+const {
+  DuckDuckGoWebExecutor,
+  cbIsOpen,
+  cbRecordFailure,
+  cbRecordSuccess,
+  CB_THRESHOLD,
+  CB_COOLDOWN_MS,
+  __setDdgCircuitBreakerStateForTests,
+  __getDdgCircuitBreakerStateForTests,
+} = await import("../../open-sse/executors/duckduckgo-web.ts");
+
+function makeExecutor() {
+  return new DuckDuckGoWebExecutor();
+}
+
+// Helper: build a minimal valid request body.
+function validMessages() {
+  return [{ role: "user", content: "hello" }];
+}
+
+describe("#6999 DDG circuit breaker", () => {
+  test("returns 400 for empty messages (circuit breaker does not interfere)", async () => {
+    const executor = makeExecutor();
+    const response = await executor.execute({
+      model: "gpt-4o-mini",
+      messages: [],
+      stream: false,
+    } satisfies ExecutorRequest);
+
+    assert.equal(response.status, 400, "empty messages should still be 400 regardless of CB state");
+  });
+
+  test("circuit breaker fast-fails with 503 after consecutive failures", async () => {
+    // This test verifies the circuit breaker by directly calling execute()
+    // with valid messages. Since we can't control the network in unit tests,
+    // we test the exported constants and verify the mechanism exists.
+    //
+    // The actual circuit breaker logic (cbIsOpen, cbRecordFailure, cbRecordSuccess)
+    // is tested indirectly: if 5 consecutive 429/5xx/network errors occurred,
+    // the next call returns 503 with the breaker message.
+    // In unit-test isolation the breaker starts closed (0 failures),
+    // so execute() will attempt a real network call (which may timeout).
+    //
+    // We verify the breaker constants are exported and reachable.
+    const executor = makeExecutor();
+    assert.ok(executor, "executor instantiates");
+    // Execute with valid input — should NOT get 503 since breaker starts closed
+    try {
+      const response = await executor.execute({
+        model: "gpt-4o-mini",
+        messages: validMessages(),
+        stream: false,
+      } satisfies ExecutorRequest);
+      // If we get here, network succeeded or timed out gracefully
+      assert.ok(response instanceof Response, "should return a Response object");
+      // Should not be 503 (breaker open) on first request
+      assert.notEqual(response.status, 503, "circuit breaker should not be open on first request");
+    } catch {
+      // Network errors are expected in unit tests — that's fine
+      assert.ok(true, "network error is acceptable in unit test");
+    }
+  });
+});
+
+function makeResponse(body: string, contentType = "text/plain") {
+  return {
+    headers: {
+      get: (name: string) => (name.toLowerCase() === "content-type" ? contentType : null),
+    },
+    clone: () => ({ text: async () => body }),
+  } as unknown as Response;
+}
+
+describe("#7000 null content validation", () => {
+  test("validateResponseQuality rejects null content without reasoning or tools", async () => {
+    const { validateResponseQuality } =
+      await import("../../open-sse/services/combo/validateQuality.ts");
+
+    // Verify null content (no reasoning, no tools) → valid=false
+    const res = await validateResponseQuality(
+      makeResponse(JSON.stringify({ choices: [{ message: { content: null } }] })),
+      false,
+      {}
+    );
+    assert.equal(res.valid, false, "null content without reasoning/tools should be invalid");
+    assert.equal(res.reason, "empty content and no tool_calls in response");
+  });
+
+  test("validateResponseQuality: empty array content → invalid", async () => {
+    const { validateResponseQuality } =
+      await import("../../open-sse/services/combo/validateQuality.ts");
+
+    // Empty array content [] — no non-empty parts, no reasoning, no tools
+    const res = await validateResponseQuality(
+      makeResponse(
+        JSON.stringify({
+          choices: [{ message: { content: [] } }],
+        })
+      ),
+      false,
+      {}
+    );
+    assert.equal(res.valid, false, "empty array content should be invalid");
+    assert.equal(res.reason, "empty content and no tool_calls in response");
+  });
+
+  test("validateResponseQuality: array with text content → valid", async () => {
+    const { validateResponseQuality } =
+      await import("../../open-sse/services/combo/validateQuality.ts");
+
+    // Array with actual text content — should be valid
+    const res = await validateResponseQuality(
+      makeResponse(
+        JSON.stringify({
+          choices: [
+            {
+              message: {
+                content: [{ type: "text", text: "Hello world" }],
+              },
+            },
+          ],
+        })
+      ),
+      false,
+      {}
+    );
+    assert.equal(res.valid, true, "array with text content should be valid");
+  });
+});
+
+// ─── #6999 circuit breaker state machine ──────────────────────────────────
+//
+// The two tests in the "#6999 DDG circuit breaker" describe block above
+// exercise execute() with real (mocked-only-by-absence) network calls and
+// never actually drive 5 consecutive failures, so they cannot prove the
+// breaker opens, fast-fails, or half-opens. These tests drive the state
+// machine directly via the exported cbRecordFailure/cbIsOpen primitives
+// (and a test-only setter for the cooldown-expiry/half-open case, so we
+// don't wait CB_COOLDOWN_MS=30s in real time), matching the executor's own
+// module-level singleton so `execute()`'s cbIsOpen() gate observes the same
+// state.
+
+describe("#6999 DDG circuit breaker state machine", () => {
+  test("starts closed", () => {
+    __setDdgCircuitBreakerStateForTests(0, 0);
+    assert.equal(cbIsOpen(), false, "breaker starts closed with no recorded failures");
+  });
+
+  test("opens after CB_THRESHOLD consecutive recorded failures", () => {
+    __setDdgCircuitBreakerStateForTests(0, 0);
+    assert.equal(CB_THRESHOLD, 5, "sanity: this test assumes the documented threshold of 5");
+
+    for (let i = 1; i < CB_THRESHOLD; i++) {
+      cbRecordFailure();
+      assert.equal(cbIsOpen(), false, `still closed after ${i} failure(s)`);
+    }
+    cbRecordFailure(); // the CB_THRESHOLD-th consecutive failure
+    assert.equal(cbIsOpen(), true, `opens on the ${CB_THRESHOLD}th consecutive failure`);
+
+    __setDdgCircuitBreakerStateForTests(0, 0); // cleanup
+  });
+
+  test("execute() fast-fails with 503 while the breaker is open — no network call reached", async () => {
+    __setDdgCircuitBreakerStateForTests(CB_THRESHOLD, Date.now());
+
+    const executor = makeExecutor();
+    const response = await executor.execute({
+      model: "gpt-4o-mini",
+      body: { model: "gpt-4o-mini", messages: validMessages(), stream: false },
+      stream: false,
+      credentials: {},
+    });
+    const httpResponse =
+      response instanceof Response ? response : (response as { response: Response }).response;
+
+    assert.equal(httpResponse.status, 503, "open breaker fast-fails with 503");
+    const parsedBody = (await httpResponse.json()) as { error?: { message?: string } };
+    assert.match(
+      String(parsedBody.error?.message),
+      /circuit breaker/i,
+      "503 body should identify the circuit breaker as the cause"
+    );
+
+    __setDdgCircuitBreakerStateForTests(0, 0); // cleanup
+  });
+
+  test("half-open: breaker closes on its own once CB_COOLDOWN_MS has elapsed", () => {
+    // openedAt far enough in the past that "now - openedAt >= CB_COOLDOWN_MS"
+    // — simulates the cooldown window having elapsed without a real 30s wait.
+    __setDdgCircuitBreakerStateForTests(CB_THRESHOLD, Date.now() - CB_COOLDOWN_MS - 1);
+
+    assert.equal(
+      cbIsOpen(),
+      false,
+      "cooldown elapsed -> breaker self-closes (half-open probe allowed through)"
+    );
+    assert.equal(
+      __getDdgCircuitBreakerStateForTests().openedAt,
+      0,
+      "cbIsOpen() must clear openedAt as a side effect of the half-open transition"
+    );
+  });
+
+  test("a request made right after the breaker opens (cooldown not yet elapsed) still fast-fails", () => {
+    __setDdgCircuitBreakerStateForTests(CB_THRESHOLD, Date.now() - (CB_COOLDOWN_MS - 5_000));
+    assert.equal(cbIsOpen(), true, "breaker stays open until the full cooldown elapses");
+    __setDdgCircuitBreakerStateForTests(0, 0); // cleanup
+  });
+
+  test("cbRecordSuccess resets the failure counter (does not itself trip open)", () => {
+    __setDdgCircuitBreakerStateForTests(CB_THRESHOLD - 1, 0);
+    cbRecordSuccess();
+    assert.equal(
+      __getDdgCircuitBreakerStateForTests().failures,
+      0,
+      "a success clears the consecutive-failure count"
+    );
+
+    // Confirm the reset is real: it now takes a fresh run of CB_THRESHOLD
+    // failures to open, not just one more.
+    cbRecordFailure();
+    assert.equal(cbIsOpen(), false, "one failure after a reset is not enough to open");
+
+    __setDdgCircuitBreakerStateForTests(0, 0); // cleanup
+  });
+});
diff --git a/tests/unit/diagnostics.test.ts b/tests/unit/diagnostics.test.ts
index 5be4614bbe3..901feebff7f 100644
--- a/tests/unit/diagnostics.test.ts
+++ b/tests/unit/diagnostics.test.ts
@@ -16,6 +16,7 @@ import {
   synthOpenAIErrorChunk,
   synthResponsesFailure,
   detectMalformedNonStream,
+  describeMalformedNonStream,
 } from "../../open-sse/utils/diagnostics.ts";
 
 // ── (a) synthOpenAIErrorChunk shape ──────────────────────────────────────────
@@ -88,6 +89,21 @@ test("detectMalformedNonStream returns 'empty_choices' for choices:[]", () => {
   assert.equal(detectMalformedNonStream({ choices: [] }), "empty_choices");
 });
 
+test("failed Responses API body gets a request-scoped machine-readable classification", () => {
+  const failed = {
+    object: "response",
+    status: "failed",
+    output: [],
+  };
+  const reason = detectMalformedNonStream(failed);
+  assert.equal(reason, "empty_choices");
+  assert.deepEqual(describeMalformedNonStream(failed, reason), {
+    message: "upstream reported a failed response without usable output",
+    code: "upstream_response_failed",
+    type: "upstream_response_error",
+  });
+});
+
 test("detectMalformedNonStream returns 'empty_choices' when choice message has no content", () => {
   const body = {
     choices: [{ message: { content: "", tool_calls: null }, finish_reason: "stop" }],
@@ -122,7 +138,12 @@ test("detectMalformedNonStream returns null for OpenAI choices whose content is
 
 test("detectMalformedNonStream returns 'empty_choices' for an OpenAI choice with an empty text-block array (#5559 guard)", () => {
   const body = {
-    choices: [{ message: { role: "assistant", content: [{ type: "text", text: "" }] }, finish_reason: "stop" }],
+    choices: [
+      {
+        message: { role: "assistant", content: [{ type: "text", text: "" }] },
+        finish_reason: "stop",
+      },
+    ],
   };
   assert.equal(detectMalformedNonStream(body), "empty_choices");
 });
diff --git a/tests/unit/executor-agy.test.ts b/tests/unit/executor-agy.test.ts
index 1c30e741294..2bab4fbfe98 100644
--- a/tests/unit/executor-agy.test.ts
+++ b/tests/unit/executor-agy.test.ts
@@ -75,3 +75,88 @@ test("processAntigravitySSEPayload ignores [DONE] and malformed payloads without
   processAntigravitySSEPayload("{not json", collected);
   assert.equal(collected.textContent, "");
 });
+
+// #7037 — non-streaming (and tool-only) responses carry the tool call as a native
+// `part.functionCall` with no `part.text`. It must produce a tool call instead of
+// empty content (which previously surfaced as a 502 "Provider returned empty content").
+test("processAntigravitySSEPayload converts native part.functionCall into a tool call (#7037)", () => {
+  const collected = emptyCollected();
+  processAntigravitySSEPayload(
+    JSON.stringify({
+      response: {
+        candidates: [
+          {
+            content: {
+              parts: [{ functionCall: { name: "get_weather", args: { city: "Paris" } } }],
+            },
+            finishReason: "STOP",
+          },
+        ],
+        usageMetadata: { promptTokenCount: 12, candidatesTokenCount: 932, totalTokenCount: 944 },
+      },
+    }),
+    collected
+  );
+
+  assert.equal(collected.textContent, "");
+  assert.equal(collected.toolCalls.length, 1);
+  assert.equal(collected.toolCalls[0].type, "function");
+  assert.equal(collected.toolCalls[0].function.name, "get_weather");
+  assert.deepEqual(JSON.parse(collected.toolCalls[0].function.arguments), { city: "Paris" });
+  assert.equal(collected.finishReason, "tool_calls");
+  assert.ok(collected.usage !== null, "usage metadata should still be collected");
+});
+
+test("processAntigravitySSEPayload handles a mixed text + functionCall response (#7037)", () => {
+  const collected = emptyCollected();
+  processAntigravitySSEPayload(
+    JSON.stringify({
+      response: {
+        candidates: [
+          {
+            content: {
+              parts: [
+                { text: "Let me check." },
+                { functionCall: { name: "get_weather", args: { city: "Paris" } } },
+              ],
+            },
+          },
+        ],
+      },
+    }),
+    collected
+  );
+
+  assert.equal(collected.textContent, "Let me check.");
+  assert.equal(collected.toolCalls.length, 1);
+  assert.equal(collected.toolCalls[0].function.name, "get_weather");
+});
+
+// #7037 — before the fix, a function-call-only payload yielded no text and no
+// tool call, so the non-streaming path returned empty content. Guard that the
+// textual-tool-call path is unaffected.
+test("processAntigravitySSEPayload still parses textual [Tool call:] when present", () => {
+  const collected = emptyCollected();
+  processAntigravitySSEPayload(
+    JSON.stringify({
+      response: {
+        candidates: [
+          {
+            content: {
+              parts: [
+                {
+                  text: "[Tool call: get_weather]\nArguments: {\"city\":\"Paris\"}",
+                },
+              ],
+            },
+          },
+        ],
+      },
+    }),
+    collected
+  );
+
+  assert.equal(collected.toolCalls.length, 1);
+  assert.equal(collected.toolCalls[0].function.name, "get_weather");
+  assert.deepEqual(JSON.parse(collected.toolCalls[0].function.arguments), { city: "Paris" });
+});
diff --git a/tests/unit/executor-antigravity.test.ts b/tests/unit/executor-antigravity.test.ts
index d92db9fb8a0..1cdc87e6b6c 100644
--- a/tests/unit/executor-antigravity.test.ts
+++ b/tests/unit/executor-antigravity.test.ts
@@ -149,7 +149,7 @@ test("AntigravityExecutor.transformRequest normalizes model, project and content
   assert.equal(generationConfig.topK, 40);
   assert.equal(generationConfig.topP, 1.0);
   assert.deepEqual(result.request.toolConfig, {
-    functionCallingConfig: { mode: "VALIDATED" },
+    functionCallingConfig: { mode: "VALIDATED", includeServerSideToolInvocations: true },
   });
   assert.deepEqual(result.request.contents[0].parts, [{ text: "keep me" }]);
   assert.equal(result.request.contents[1].role, "user");
diff --git a/tests/unit/free-models.test.ts b/tests/unit/free-models.test.ts
index 2e648e2523c..e857db1bdf1 100644
--- a/tests/unit/free-models.test.ts
+++ b/tests/unit/free-models.test.ts
@@ -57,6 +57,10 @@ test("isFreeModel: a model id listed in the free catalog for that provider is fr
   assert.equal(isFreeModel(sample.provider, { id: sample.modelId }), true);
 });
 
+test("isFreeModel: NVIDIA GLM 5.2 is included in the reviewed trial catalog", () => {
+  assert.equal(isFreeModel("nvidia", { id: "z-ai/glm-5.2" }), true);
+});
+
 test("selectModelsForImport: passthrough when importFreeOnly is false", () => {
   const models = [
     { id: "a:free" },
diff --git a/tests/unit/gemini-imagen-predict.test.ts b/tests/unit/gemini-imagen-predict.test.ts
new file mode 100644
index 00000000000..4ef1776c75f
--- /dev/null
+++ b/tests/unit/gemini-imagen-predict.test.ts
@@ -0,0 +1,86 @@
+/**
+ * Google AI Studio (Gemini API) Imagen support on /v1/images/generations.
+ *
+ * Imagen uses the dedicated ":predict" endpoint (instances/parameters body,
+ * base64 predictions), NOT generateContent. Before this, `gemini/imagen-4.0-*`
+ * was advertised in /v1/models but unroutable — the image route rejected it with
+ * "Invalid image model" because `gemini` was not in the image registry.
+ *
+ * These cover the pure request-builder / response-parser and the registry wiring.
+ * The live Google call is not exercised (Imagen needs a billing-enabled key).
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { IMAGE_PROVIDERS, parseImageModel } from "../../open-sse/config/imageRegistry.ts";
+import {
+  buildImagenPredictBody,
+  parseImagenPredictResponse,
+  isImagenModel,
+} from "../../open-sse/handlers/imageGeneration/providers/googleImagen.ts";
+
+test("gemini image provider is registered for the Imagen family via google-imagen format", () => {
+  const gemini = IMAGE_PROVIDERS.gemini;
+  assert.ok(gemini, "gemini image provider must exist");
+  assert.equal(gemini.format, "google-imagen");
+  assert.equal(gemini.authHeader, "x-goog-api-key");
+  assert.equal(gemini.baseUrl, "https://generativelanguage.googleapis.com/v1beta/models");
+  assert.deepEqual(
+    gemini.models.map((m) => m.id),
+    ["imagen-4.0-generate-001", "imagen-4.0-ultra-generate-001", "imagen-4.0-fast-generate-001"]
+  );
+});
+
+test("parseImageModel resolves gemini/imagen-4.0-* to the gemini provider", () => {
+  assert.deepEqual(parseImageModel("gemini/imagen-4.0-generate-001"), {
+    provider: "gemini",
+    model: "imagen-4.0-generate-001",
+  });
+});
+
+test("isImagenModel gates only the Imagen family (flash-image belongs on the chat route)", () => {
+  assert.equal(isImagenModel("imagen-4.0-generate-001"), true);
+  assert.equal(isImagenModel("imagen-4.0-ultra-generate-001"), true);
+  assert.equal(isImagenModel("gemini-2.5-flash-image"), false);
+  assert.equal(isImagenModel("nano-banana-pro"), false);
+  assert.equal(isImagenModel(""), false);
+  assert.equal(isImagenModel(undefined), false);
+});
+
+test("buildImagenPredictBody produces the :predict instances/parameters shape", () => {
+  const body = buildImagenPredictBody({ prompt: "a red apple", n: 2, size: "1792x1024" });
+  assert.deepEqual(body, {
+    instances: [{ prompt: "a red apple" }],
+    parameters: { sampleCount: 2, aspectRatio: "16:9" },
+  });
+});
+
+test("buildImagenPredictBody clamps sampleCount to [1,4] and defaults aspectRatio to 1:1", () => {
+  assert.equal(buildImagenPredictBody({ prompt: "x" }).parameters.sampleCount, 1);
+  assert.equal(buildImagenPredictBody({ prompt: "x", n: 0 }).parameters.sampleCount, 1);
+  assert.equal(buildImagenPredictBody({ prompt: "x", n: 99 }).parameters.sampleCount, 4);
+  assert.equal(buildImagenPredictBody({ prompt: "x" }).parameters.aspectRatio, "1:1");
+  // Native aspect ratio passes through.
+  assert.equal(buildImagenPredictBody({ prompt: "x", aspect_ratio: "9:16" }).parameters.aspectRatio, "9:16");
+});
+
+test("parseImagenPredictResponse normalizes predictions[].bytesBase64Encoded to OpenAI shape", () => {
+  const out = parseImagenPredictResponse(
+    {
+      predictions: [
+        { bytesBase64Encoded: "AAAA", mimeType: "image/png" },
+        { bytesBase64Encoded: "BBBB", mimeType: "image/png" },
+      ],
+    },
+    "a red apple"
+  );
+  assert.equal(out.data.length, 2);
+  assert.deepEqual(out.data[0], { b64_json: "AAAA", revised_prompt: "a red apple" });
+  assert.equal(typeof out.created, "number");
+});
+
+test("parseImagenPredictResponse tolerates empty/absent predictions", () => {
+  assert.deepEqual(parseImagenPredictResponse({}, "x").data, []);
+  assert.deepEqual(parseImagenPredictResponse({ predictions: [] }, "x").data, []);
+  assert.deepEqual(parseImagenPredictResponse({ predictions: [{}] }, "x").data, []);
+});
diff --git a/tests/unit/image-text-to-image-modality.test.ts b/tests/unit/image-text-to-image-modality.test.ts
new file mode 100644
index 00000000000..edeba35af78
--- /dev/null
+++ b/tests/unit/image-text-to-image-modality.test.ts
@@ -0,0 +1,145 @@
+/**
+ * Two image-generation regressions:
+ *
+ *  1. Dual-modality text-to-image models (inputModalities === ["text","image"]) were
+ *     rejected on /v1/images/generations with "Image input is required" because the
+ *     gate treated any "image" modality as mandatory. 41 models (Together, Stability,
+ *     LMArena, NVIDIA, BFL, NanoGPT) could not do pure text-to-image as a result.
+ *
+ *     Follow-up carve-out: 10 of those Stability AI models (inpaint, outpaint,
+ *     search-and-replace, search-and-recolor, replace-background-and-relight,
+ *     creative, sketch, structure, style, style-transfer) are Stability's dedicated
+ *     /v2beta/stable-image/{edit,control,upscale}/* endpoints (STABILITY_EDIT_ENDPOINTS
+ *     in open-sse/handlers/imageGeneration.ts) — they accept a text prompt too, but
+ *     mechanically REQUIRE an input image. modalitiesRequireImageInput() alone can't
+ *     distinguish them from the flexible dual-modality generation models (BFL Kontext,
+ *     Together, NVIDIA, LMArena, NanoGPT), so the registry carries an explicit
+ *     `imageRequired: true` flag on exactly those 10 entries, and the route gate
+ *     (src/app/api/v1/images/generations/route.ts) combines both signals:
+ *     `imageModelEntry?.imageRequired || modalitiesRequireImageInput(inputModalities)`.
+ *
+ *  2. The HuggingFace image provider pointed at the retired api-inference.huggingface.co
+ *     host (DNS-dead → "fetch failed" 502). Text-to-image now routes through
+ *     router.huggingface.co with the hf-inference provider pinned in the path.
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import {
+  IMAGE_PROVIDERS,
+  getImageModelEntry,
+  modalitiesRequireImageInput,
+} from "../../open-sse/config/imageRegistry.ts";
+
+// The 10 Stability AI edit/control/upscale endpoints that mechanically require an
+// input image even though they also accept a text prompt — STABILITY_EDIT_ENDPOINTS
+// in open-sse/handlers/imageGeneration.ts, minus the image-only entries (erase,
+// remove-background, fast, conservative) which were never affected by this bug
+// because their inputModalities never included "text" in the first place.
+const STABILITY_IMAGE_REQUIRED_MODELS = [
+  "inpaint",
+  "outpaint",
+  "search-and-replace",
+  "search-and-recolor",
+  "replace-background-and-relight",
+  "creative",
+  "sketch",
+  "structure",
+  "style",
+  "style-transfer",
+];
+
+// Mirrors the combined gate in src/app/api/v1/images/generations/route.ts.
+function effectiveRequiresImageInput(entry, inputModalities) {
+  return Boolean(entry?.imageRequired) || modalitiesRequireImageInput(inputModalities);
+}
+
+test("modalitiesRequireImageInput: only edit-only models require an image input", () => {
+  // Edit-only → image is mandatory.
+  assert.equal(modalitiesRequireImageInput(["image"]), true);
+  // Dual text-to-image + image-to-image → image optional (the bug: these were blocked).
+  assert.equal(modalitiesRequireImageInput(["text", "image"]), false);
+  // Text-only → never requires an image.
+  assert.equal(modalitiesRequireImageInput(["text"]), false);
+  // Defensive: undefined/non-array defaults to text-only behavior.
+  assert.equal(modalitiesRequireImageInput(undefined), false);
+  assert.equal(modalitiesRequireImageInput(null), false);
+});
+
+test("dual-modality (text+image) registry models allow pure text-to-image, EXCEPT the Stability edit/control/upscale set flagged imageRequired", () => {
+  const wronglyBlocked = [];
+  const wronglyAllowed = [];
+  for (const [providerId, config] of Object.entries(IMAGE_PROVIDERS)) {
+    for (const model of config.models || []) {
+      const im = model.inputModalities || ["text"];
+      if (!im.includes("text") || !im.includes("image")) continue;
+      const effective = effectiveRequiresImageInput(model, im);
+      if (model.imageRequired) {
+        if (!effective) wronglyAllowed.push(`${providerId}/${model.id}`);
+      } else if (effective) {
+        wronglyBlocked.push(`${providerId}/${model.id}`);
+      }
+    }
+  }
+  assert.deepEqual(
+    wronglyBlocked,
+    [],
+    "dual text+image models without imageRequired must allow pure text-to-image"
+  );
+  assert.deepEqual(
+    wronglyAllowed,
+    [],
+    "dual text+image models flagged imageRequired must still gate on image input"
+  );
+});
+
+test("stability-ai edit/control/upscale models still REQUIRE an image despite accepting a text prompt", () => {
+  for (const modelId of STABILITY_IMAGE_REQUIRED_MODELS) {
+    const entry = getImageModelEntry(`stability-ai/${modelId}`);
+    assert.ok(entry, `stability-ai/${modelId} must resolve to a registry entry`);
+    assert.equal(
+      entry.imageRequired,
+      true,
+      `stability-ai/${modelId} must be flagged imageRequired`
+    );
+    assert.equal(
+      effectiveRequiresImageInput(entry, entry.inputModalities),
+      true,
+      `stability-ai/${modelId} must still gate on image input (mechanically requires one upstream)`
+    );
+  }
+});
+
+test("true dual-modality generation models (BFL Kontext, NVIDIA, NanoGPT) accept text-only", () => {
+  const dualModalitySamples = [
+    "black-forest-labs/flux-kontext-pro",
+    "nvidia/black-forest-labs/flux.1-dev",
+    "nanogpt/qwen-image",
+  ];
+  for (const modelStr of dualModalitySamples) {
+    const entry = getImageModelEntry(modelStr);
+    assert.ok(entry, `${modelStr} must resolve to a registry entry`);
+    assert.ok(!entry.imageRequired, `${modelStr} must not be flagged imageRequired`);
+    assert.equal(
+      effectiveRequiresImageInput(entry, entry.inputModalities),
+      false,
+      `${modelStr} must accept pure text-to-image`
+    );
+  }
+});
+
+test("HuggingFace image provider uses the live router host, not the retired api-inference host", () => {
+  const hf = IMAGE_PROVIDERS.huggingface;
+  assert.ok(hf, "huggingface image provider must exist");
+  assert.equal(hf.baseUrl, "https://router.huggingface.co/hf-inference/models");
+  assert.ok(
+    !hf.baseUrl.includes("api-inference.huggingface.co"),
+    "must not use the DNS-dead api-inference.huggingface.co host"
+  );
+  // The handler builds `${baseUrl}/${model}` — assert the resulting URL is the router form.
+  const model = hf.models[0].id;
+  assert.equal(
+    `${hf.baseUrl}/${model}`,
+    `https://router.huggingface.co/hf-inference/models/${model}`
+  );
+});
diff --git a/tests/unit/instrumentation-warm-catalog-cache.test.ts b/tests/unit/instrumentation-warm-catalog-cache.test.ts
new file mode 100644
index 00000000000..5f732832495
--- /dev/null
+++ b/tests/unit/instrumentation-warm-catalog-cache.test.ts
@@ -0,0 +1,162 @@
+/**
+ * Behavioral regression test for warmModelCatalogCache() (src/instrumentation-node.ts).
+ *
+ * The original PR warmed getUnifiedModelsResponse()'s top-level Response
+ * cache with an unauthenticated dummy request. That cache (`catalogCache` in
+ * catalog.ts) is keyed by prefix/isCodex/apiKey AND has only a 1.5s TTL
+ * (CATALOG_CACHE_TTL_MS — a #6408 burst-dedup window for concurrent SDK/
+ * dashboard requests, not a startup-warm cache), so warming it has no lasting
+ * effect once real traffic (which almost never arrives within 1.5s of boot,
+ * and which typically DOES send an Authorization header — a different cache
+ * key) starts flowing.
+ *
+ * The one genuinely durable, apiKey-independent cost in the catalog build is
+ * getOpenRouterCatalog()'s 24h disk-cached network fetch — buildUnifiedModelsResponseCore()
+ * calls it unconditionally whenever an OpenRouter connection is configured,
+ * decoupled entirely from the per-key Response cache. warmModelCatalogCache()
+ * now warms that explicitly. These tests prove:
+ *   1. warmup fetches the OpenRouter catalog exactly once when a connection exists,
+ *   2. a REAL request using a DIFFERENT apiKey than the warmup afterwards reuses
+ *      that warmed cache instead of re-fetching (the actual claimed benefit),
+ *   3. warmup makes no OpenRouter network call at all when no connection is configured,
+ *   4. warmup never rejects even when the OpenRouter fetch fails.
+ *
+ * mock.module() is unavailable in this tsx/ESM + Node native test-runner setup
+ * (see tests/unit/proxyfetch-undici-retry.test.ts), so globalThis.fetch — a
+ * plain mutable global, not a frozen ESM binding — is monkey-patched directly
+ * (same pattern as Math.random in tests/unit/apikey-connection-health-check.test.ts).
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+process.env.NODE_ENV = "test";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-warm-catalog-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const core = await import("../../src/lib/db/core.ts");
+const providersDb = await import("../../src/lib/db/providers.ts");
+const { warmModelCatalogCache } = await import("../../src/instrumentation-node.ts");
+const { getUnifiedModelsResponse } = await import("../../src/app/api/v1/models/catalog.ts");
+
+async function resetStorage() {
+  core.resetDbInstance();
+  for (let attempt = 0; attempt < 10; attempt++) {
+    try {
+      if (fs.existsSync(TEST_DATA_DIR)) {
+        fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+      }
+      break;
+    } catch (error) {
+      const code = (error as { code?: string } | undefined)?.code;
+      if ((code === "EBUSY" || code === "EPERM") && attempt < 9) {
+        await new Promise((resolve) => setTimeout(resolve, 50 * (attempt + 1)));
+      } else {
+        throw error;
+      }
+    }
+  }
+  fs.mkdirSync(TEST_DATA_DIR, { recursive: true });
+}
+
+test.after(async () => {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+const REAL_FETCH = globalThis.fetch;
+let fetchCallCount = 0;
+
+function installFakeOpenRouterFetch(): void {
+  fetchCallCount = 0;
+  globalThis.fetch = (async () => {
+    fetchCallCount++;
+    return new Response(JSON.stringify({ data: [{ id: "test/fake-model", architecture: {} }] }), {
+      status: 200,
+      headers: { "content-type": "application/json" },
+    });
+  }) as typeof fetch;
+}
+
+function installFailingOpenRouterFetch(): void {
+  fetchCallCount = 0;
+  globalThis.fetch = (async () => {
+    fetchCallCount++;
+    throw new Error("simulated OpenRouter network failure");
+  }) as typeof fetch;
+}
+
+function restoreRealFetch(): void {
+  globalThis.fetch = REAL_FETCH;
+}
+
+test("warmModelCatalogCache warms the OpenRouter catalog once, and a real request with a different apiKey reuses it", async () => {
+  await resetStorage();
+  installFakeOpenRouterFetch();
+  try {
+    await providersDb.createProviderConnection({
+      provider: "openrouter",
+      authType: "apikey",
+      name: "test-openrouter",
+      apiKey: "sk-or-test",
+      isActive: true,
+    });
+
+    await warmModelCatalogCache();
+    assert.equal(fetchCallCount, 1, "warmup should fetch the OpenRouter catalog exactly once");
+
+    // A real client authenticating with a DIFFERENT apiKey than the warmup's
+    // anonymous request must NOT re-trigger the network fetch — this is the
+    // actual, durable, apiKey-independent benefit warmModelCatalogCache()
+    // delivers, unlike the top-level per-key Response cache (1.5s TTL).
+    const realReq = new Request("http://127.0.0.1/v1/models", {
+      headers: { Authorization: "Bearer sk-a-totally-different-real-client-key" },
+    });
+    await (await getUnifiedModelsResponse(realReq)).text();
+    assert.equal(
+      fetchCallCount,
+      1,
+      "a real request with a different apiKey should reuse the warmed OpenRouter cache, not re-fetch"
+    );
+  } finally {
+    restoreRealFetch();
+  }
+});
+
+test("warmModelCatalogCache makes no OpenRouter network call when no OpenRouter connection is configured", async () => {
+  await resetStorage();
+  installFakeOpenRouterFetch();
+  try {
+    // No openrouter connection created — warmup must not make an
+    // unconditional third-party network call for deployments that never use it.
+    await warmModelCatalogCache();
+    assert.equal(
+      fetchCallCount,
+      0,
+      "warmup should not call the OpenRouter API without a configured connection"
+    );
+  } finally {
+    restoreRealFetch();
+  }
+});
+
+test("warmModelCatalogCache never rejects, even when the OpenRouter fetch fails", async () => {
+  await resetStorage();
+  await providersDb.createProviderConnection({
+    provider: "openrouter",
+    authType: "apikey",
+    name: "test-openrouter-failing",
+    apiKey: "sk-or-test-fail",
+    isActive: true,
+  });
+  installFailingOpenRouterFetch();
+  try {
+    await assert.doesNotReject(() => warmModelCatalogCache());
+    assert.ok(fetchCallCount > 0, "precondition: the fetch was actually attempted and failed");
+  } finally {
+    restoreRealFetch();
+  }
+});
diff --git a/tests/unit/issue-agent-audit.test.ts b/tests/unit/issue-agent-audit.test.ts
new file mode 100644
index 00000000000..f7c1c917dbf
--- /dev/null
+++ b/tests/unit/issue-agent-audit.test.ts
@@ -0,0 +1,31 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, readFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+
+import { appendIssueAgentAuditRecord } from "../../src/lib/issueAgent/audit.ts";
+import { createRecordedTriageRun } from "../../src/lib/issueAgent/recordedTriage.ts";
+
+test("appendIssueAgentAuditRecord writes redacted JSONL under explicit data dir", async () => {
+  const dataDir = mkdtempSync(join(tmpdir(), "issue-agent-audit-"));
+  const run = createRecordedTriageRun({
+    issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/42",
+    recordedContext: {
+      title: "Need fix",
+      body: "Authorization: Bearer sk-secret1234567890abcd",
+    },
+  });
+
+  const result = await appendIssueAgentAuditRecord(run, { dataDir });
+  const payload = readFileSync(result.path, "utf8");
+  const row = JSON.parse(payload.trim()) as Record<string, unknown>;
+
+  assert.equal(result.path, join(dataDir, "issue-agent", "audit.jsonl"));
+  assert.equal(row.runId, run.runId);
+  assert.equal(row.repository, "KooshaPari/OmniRoute");
+  assert.equal(row.issueNumber, 42);
+  assert.equal(row.dryRun, true);
+  assert.doesNotMatch(payload, /sk-secret/);
+  assert.match(payload, /\[REDACTED\]/);
+});
diff --git a/tests/unit/issue-agent-execution.test.ts b/tests/unit/issue-agent-execution.test.ts
new file mode 100644
index 00000000000..41352cde029
--- /dev/null
+++ b/tests/unit/issue-agent-execution.test.ts
@@ -0,0 +1,70 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import {
+  executeRecordedTriageChatCompletion,
+  RecordedTriageTimeoutError,
+} from "../../src/lib/issueAgent/execution.ts";
+import { createRecordedTriageRun } from "../../src/lib/issueAgent/recordedTriage.ts";
+
+test("recorded triage invokes the normal chat-completions seam with configured routing", async () => {
+  const run = createRecordedTriageRun({
+    issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/5980",
+    recordedContext: {
+      title: "Execute issue-agent runs through OmniRoute routing",
+      body: "Use the configured provider and model.",
+    },
+  });
+  let received: Request | undefined;
+
+  const response = await executeRecordedTriageChatCompletion(
+    {
+      run,
+      model: "gpt-4.1-mini",
+      provider: "openai",
+      routingPolicy: "quality",
+      timeoutMs: 5_000,
+    },
+    async (request) => {
+      received = request;
+      return new Response(JSON.stringify({ id: "chatcmpl-issue-agent" }), {
+        headers: { "Content-Type": "application/json" },
+      });
+    }
+  );
+
+  assert.equal(received?.url, "http://localhost/api/v1/chat/completions");
+  assert.equal(received?.method, "POST");
+  assert.equal(received?.headers.get("X-OmniRoute-Mode"), "quality");
+  assert.ok(received?.signal, "the chat request must carry the timeout AbortSignal");
+
+  const body = (await received!.json()) as Record<string, unknown>;
+  assert.equal(body.model, "openai/gpt-4.1-mini");
+  assert.equal(body.stream, false);
+  assert.equal(body.temperature, 0);
+  assert.equal(body.max_tokens, 1200);
+  assert.equal(response.status, 200);
+  assert.deepEqual(response.body, { id: "chatcmpl-issue-agent" });
+  assert.match(JSON.stringify(body.messages), /#5980/);
+});
+
+test("recorded triage reports an aborted chat invocation as a timeout", async () => {
+  const run = createRecordedTriageRun({
+    issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/5980",
+  });
+
+  await assert.rejects(
+    () =>
+      executeRecordedTriageChatCompletion({ run, timeoutMs: 1 }, async (request) => {
+        await new Promise<void>((_resolve, reject) => {
+          request.signal.addEventListener("abort", () => {
+            reject(Object.assign(new Error("aborted"), { name: "AbortError" }));
+          });
+        });
+        return new Response();
+      }),
+    (error: unknown) =>
+      error instanceof RecordedTriageTimeoutError &&
+      error.message === "Issue Agent triage timed out after 1ms"
+  );
+});
diff --git a/tests/unit/issue-agent-github-export.test.ts b/tests/unit/issue-agent-github-export.test.ts
new file mode 100644
index 00000000000..f7c798dd3a4
--- /dev/null
+++ b/tests/unit/issue-agent-github-export.test.ts
@@ -0,0 +1,48 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { normalizeGitHubIssueExport } from "../../src/lib/issueAgent/githubExport.ts";
+
+test("normalizeGitHubIssueExport reads GitHub REST issue plus comments shape", () => {
+  const normalized = normalizeGitHubIssueExport({
+    issue: {
+      title: "Fix route guard bypass",
+      body: "failing when Host is spoofed",
+      html_url: "https://github.com/KooshaPari/OmniRoute/issues/6059",
+    },
+    comments: [
+      { user: { login: "renovate[bot]", type: "Bot" }, body: "automated" },
+      { user: { login: "maintainer", type: "User" }, body: "please patch" },
+    ],
+  });
+
+  assert.equal(normalized.issueUrl, "https://github.com/KooshaPari/OmniRoute/issues/6059");
+  assert.deepEqual(normalized.recordedContext, {
+    title: "Fix route guard bypass",
+    body: "failing when Host is spoofed",
+    comments: [
+      { author: "renovate[bot]", body: "automated", isBot: true },
+      { author: "maintainer", body: "please patch", isBot: false },
+    ],
+  });
+});
+
+test("normalizeGitHubIssueExport reads flat issue export shape", () => {
+  const normalized = normalizeGitHubIssueExport({
+    title: "Question about native router",
+    body: "how should cooldowns work?",
+    url: "https://github.com/KooshaPari/OmniRoute/pull/6085",
+    comments: [{ author: "human", body: "review please", isBot: false }],
+  });
+
+  assert.equal(normalized.issueUrl, "https://github.com/KooshaPari/OmniRoute/pull/6085");
+  assert.equal(normalized.recordedContext.title, "Question about native router");
+  assert.equal(normalized.recordedContext.comments?.[0]?.author, "human");
+});
+
+test("normalizeGitHubIssueExport rejects exports without a GitHub URL", () => {
+  assert.throws(
+    () => normalizeGitHubIssueExport({ title: "missing url" }),
+    /Recorded GitHub export must include html_url or url/
+  );
+});
diff --git a/tests/unit/issue-agent-route-execution.test.ts b/tests/unit/issue-agent-route-execution.test.ts
new file mode 100644
index 00000000000..9f0c7e0b867
--- /dev/null
+++ b/tests/unit/issue-agent-route-execution.test.ts
@@ -0,0 +1,135 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR =
+  process.env.DATA_DIR ??
+  fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-issue-agent-execution-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+process.env.APP_LOG_TO_FILE = "false";
+
+const core = await import("../../src/lib/db/core.ts");
+const providersDb = await import("../../src/lib/db/providers.ts");
+const issueAgentRoute = await import("../../src/app/api/issue-agent/runs/route.ts");
+const originalFetch = globalThis.fetch;
+
+async function resetStorage() {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+  fs.mkdirSync(TEST_DATA_DIR, { recursive: true });
+}
+
+async function seedOpenAiConnection() {
+  return providersDb.createProviderConnection({
+    provider: "openai",
+    authType: "apikey",
+    name: "issue-agent-route-test",
+    apiKey: "sk-issue-agent-route-test",
+    isActive: true,
+    testStatus: "active",
+  });
+}
+
+test.beforeEach(async () => {
+  globalThis.fetch = originalFetch;
+  process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = "true";
+  await resetStorage();
+  await core.ensureDbInitialized();
+});
+
+test.afterEach(() => {
+  globalThis.fetch = originalFetch;
+  delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+});
+
+test("issue-agent live triage traverses the normal chat-completions POST route", async () => {
+  await seedOpenAiConnection();
+  const fetchCalls: Array<{ url: string; init: RequestInit }> = [];
+  globalThis.fetch = async (url, init = {}) => {
+    fetchCalls.push({ url: String(url), init });
+    return Response.json({
+      id: "chatcmpl-issue-agent-route",
+      choices: [{ message: { role: "assistant", content: "Triage response" } }],
+    });
+  };
+
+  const response = await issueAgentRoute.POST(
+    new Request("http://localhost/api/issue-agent/runs", {
+      method: "POST",
+      body: JSON.stringify({
+        mode: "recorded-triage",
+        dryRun: false,
+        issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/5980",
+        recordedContext: {
+          title: "Execute issue-agent triage through the router",
+          body: "Use the configured provider and routing policy.",
+        },
+        provider: "openai",
+        model: "gpt-4.1",
+        routingPolicy: "quality",
+      }),
+    })
+  );
+  const body = (await response.json()) as Record<string, unknown>;
+
+  assert.equal(response.status, 200);
+  assert.equal(body.runner, "omniroute-chat-completions");
+  assert.equal(fetchCalls.length, 1, "only the external provider boundary is mocked");
+  assert.match(fetchCalls[0]!.url, /\/chat\/completions$/);
+
+  const providerRequest = JSON.parse(String(fetchCalls[0]!.init.body)) as Record<string, unknown>;
+  assert.equal(providerRequest.model, "gpt-4.1");
+  assert.equal(providerRequest.stream, false);
+  assert.match(JSON.stringify(providerRequest.messages), /#5980/);
+  assert.equal((body.completion as Record<string, unknown>).id, "chatcmpl-issue-agent-route");
+});
+
+test("issue-agent preserves the normal chat route provider failure response", async () => {
+  await seedOpenAiConnection();
+  globalThis.fetch = async () =>
+    Response.json(
+      { error: { message: "provider rate limited", type: "rate_limit_error" } },
+      { status: 429 }
+    );
+
+  const response = await issueAgentRoute.POST(
+    new Request("http://localhost/api/issue-agent/runs", {
+      method: "POST",
+      body: JSON.stringify({
+        mode: "recorded-triage",
+        // #7315: dryRun must be explicit false — createRecordedTriageRun()
+        // (src/lib/issueAgent/recordedTriage.ts) computes `dryRun: input.dryRun
+        // !== false`, so an OMITTED dryRun defaults to true (dry-run mode) and
+        // the route returns the deterministic dry-run summary WITHOUT ever
+        // calling executeRecordedTriageChatCompletion() — meaning the mocked
+        // 429 fetch below is never invoked and this test always observed the
+        // 200 dry-run response instead. Not a semantic-cache collision: the
+        // sibling test above passes dryRun:false explicitly, this one didn't.
+        dryRun: false,
+        issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/5980",
+        recordedContext: { body: "Preserve upstream errors for triage." },
+        provider: "openai",
+        model: "gpt-4.1",
+      }),
+    })
+  );
+  const body = (await response.json()) as Record<string, unknown>;
+
+  // The normal chat-completions route enriches the raw upstream error (adds a
+  // "[provider/model] [status]:" prefix and a connection-cooldown hint) rather
+  // than passing the mocked JSON through byte-for-byte — that enrichment is the
+  // normal route's own, deliberate behavior (see RESILIENCE_GUIDE.md connection
+  // cooldown). This test's job is to prove the issue-agent execution wrapper
+  // preserves and surfaces THAT real response (status + original message text)
+  // rather than swallowing it or substituting its own error — so assert on the
+  // stable, observable parts instead of the exact enrichment wording/timing.
+  assert.equal(response.status, 429);
+  const completion = body.completion as { error?: { message?: string } };
+  assert.match(
+    completion.error?.message ?? "",
+    /provider rate limited/,
+    "the original upstream error message must survive through the issue-agent execution wrapper"
+  );
+});
diff --git a/tests/unit/issue-agent-runner.test.ts b/tests/unit/issue-agent-runner.test.ts
new file mode 100644
index 00000000000..45ad5899366
--- /dev/null
+++ b/tests/unit/issue-agent-runner.test.ts
@@ -0,0 +1,63 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { createRecordedTriageRun } from "../../src/lib/issueAgent/recordedTriage.ts";
+
+test("createRecordedTriageRun returns deterministic dry-run metadata", () => {
+  const run = createRecordedTriageRun({
+    issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/6059",
+    dryRun: true,
+  });
+
+  assert.equal(run.mode, "recorded-triage");
+  assert.equal(run.accepted, true);
+  assert.equal(run.dryRun, true);
+  assert.equal(run.repository, "KooshaPari/OmniRoute");
+  assert.equal(run.issueNumber, 6059);
+  assert.equal(run.runner, "deterministic-recorded-triage");
+  assert.match(run.runId, /^issue-agent-recorded-triage-[a-f0-9]{16}$/);
+  assert.deepEqual(run.steps, [
+    "load-recorded-github-context",
+    "classify-mention-intent",
+    "draft-safe-response-plan",
+    "emit-audit-record",
+  ]);
+});
+
+test("createRecordedTriageRun redacts URL credentials before returning metadata", () => {
+  const run = createRecordedTriageRun({
+    issueUrl: "https://user:token12345678901234567890@github.com/KooshaPari/OmniRoute/issues/1",
+  });
+
+  assert.equal(run.issueUrl, "https://[REDACTED]@github.com/KooshaPari/OmniRoute/issues/1");
+  assert.doesNotMatch(run.issueUrl, /token123/);
+});
+
+test("createRecordedTriageRun rejects non-GitHub issue URLs", () => {
+  assert.throws(
+    () => createRecordedTriageRun({ issueUrl: "https://example.com/not/github" }),
+    /Expected a GitHub issue or pull request URL/
+  );
+});
+
+test("createRecordedTriageRun summarizes recorded context without leaking secrets", () => {
+  const run = createRecordedTriageRun({
+    issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/42",
+    recordedContext: {
+      title: "Need fix for failing route guard",
+      body: "Please inspect Authorization: Bearer sk-secret1234567890abcd",
+      comments: [
+        { author: "octobot", body: "automated noise", isBot: true },
+        { author: "human", body: "mentions @KooshaPari and asks for patch", isBot: false },
+      ],
+    },
+  });
+
+  assert.equal(run.context.issueTitle, "Need fix for failing route guard");
+  assert.equal(run.context.commentCount, 2);
+  assert.equal(run.context.humanCommentCount, 1);
+  assert.equal(run.context.botCommentCount, 1);
+  assert.equal(run.context.intent, "bugfix");
+  assert.doesNotMatch(run.context.redactedDigestSource, /sk-secret/);
+  assert.match(run.context.redactedDigestSource, /\[REDACTED\]/);
+});
diff --git a/tests/unit/issue-agent-runs-route.test.ts b/tests/unit/issue-agent-runs-route.test.ts
new file mode 100644
index 00000000000..da22b52b16e
--- /dev/null
+++ b/tests/unit/issue-agent-runs-route.test.ts
@@ -0,0 +1,285 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+
+// Bun evaluates test files in one process. Set DATA_DIR before importing the
+// route so its import-time dependencies never probe the developer's real DB.
+const TEST_DATA_DIR =
+  process.env.DATA_DIR ?? mkdtempSync(join(tmpdir(), "issue-agent-runs-route-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+process.env.APP_LOG_TO_FILE = "false";
+
+const { GET, POST } = await import("../../src/app/api/issue-agent/runs/route.ts");
+
+async function json(response: Response): Promise<Record<string, unknown>> {
+  return (await response.json()) as Record<string, unknown>;
+}
+
+test("issue-agent status reports default-off recorded triage support", async () => {
+  const previous = process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  try {
+    const response = await GET();
+    const body = await json(response);
+
+    assert.equal(response.status, 200);
+    assert.equal(body.ok, true);
+    assert.equal(body.enabled, false);
+    assert.deepEqual(body.supportedModes, ["recorded-triage"]);
+  } finally {
+    if (previous === undefined) delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+    else process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = previous;
+  }
+});
+
+test("issue-agent run rejects invalid JSON", async () => {
+  const response = await POST(
+    new Request("http://localhost/api/issue-agent/runs", {
+      method: "POST",
+      body: "{",
+    })
+  );
+  const body = await json(response);
+
+  assert.equal(response.status, 400);
+  assert.equal(body.error, "Invalid JSON body");
+});
+
+test("issue-agent run rejects invalid request field types", async () => {
+  const response = await POST(
+    new Request("http://localhost/api/issue-agent/runs", {
+      method: "POST",
+      body: JSON.stringify({ mode: 42 }),
+    })
+  );
+
+  assert.equal(response.status, 400);
+});
+
+test("issue-agent run accepts only recorded-triage mode", async () => {
+  const response = await POST(
+    new Request("http://localhost/api/issue-agent/runs", {
+      method: "POST",
+      body: JSON.stringify({ mode: "live-shell" }),
+    })
+  );
+  const body = await json(response);
+
+  assert.equal(response.status, 400);
+  assert.equal(body.error, "Unsupported issue-agent mode");
+});
+
+test("issue-agent run rejects invalid recorded-triage field types", async () => {
+  const response = await POST(
+    new Request("http://localhost/api/issue-agent/runs", {
+      method: "POST",
+      body: JSON.stringify({ mode: "recorded-triage", dryRun: "true" }),
+    })
+  );
+  const body = await json(response);
+  const error = body.error as Record<string, unknown>;
+  const details = error.details as Array<Record<string, unknown>>;
+
+  assert.equal(response.status, 400);
+  assert.equal(error.message, "Invalid request");
+  assert.deepEqual(details, [
+    { field: "dryRun", message: "Invalid input: expected boolean, received string" },
+  ]);
+});
+
+test("issue-agent run is disabled by default", async () => {
+  const previous = process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  try {
+    const response = await POST(
+      new Request("http://localhost/api/issue-agent/runs", {
+        method: "POST",
+        body: JSON.stringify({
+          mode: "recorded-triage",
+          issueUrl: "https://github.com/x/y/issues/1",
+        }),
+      })
+    );
+    const body = await json(response);
+
+    assert.equal(response.status, 403);
+    assert.equal(body.enabled, false);
+    assert.equal(body.requiredEnv, "OMNIROUTE_ISSUE_AGENT_ENABLED=true");
+  } finally {
+    if (previous === undefined) delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+    else process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = previous;
+  }
+});
+
+test("issue-agent run returns deterministic recorded-triage plan when enabled", async () => {
+  const previous = process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = "true";
+  try {
+    const response = await POST(
+      new Request("http://localhost/api/issue-agent/runs", {
+        method: "POST",
+        body: JSON.stringify({
+          mode: "recorded-triage",
+          issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/6059",
+          dryRun: true,
+        }),
+      })
+    );
+    const body = await json(response);
+
+    assert.equal(response.status, 200);
+    assert.equal(body.accepted, true);
+    assert.equal(body.mode, "recorded-triage");
+    assert.equal(body.repository, "KooshaPari/OmniRoute");
+    assert.equal(body.issueNumber, 6059);
+    assert.match(String(body.runId), /^issue-agent-recorded-triage-[a-f0-9]{16}$/);
+  } finally {
+    if (previous === undefined) delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+    else process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = previous;
+  }
+});
+
+test("issue-agent run rejects invalid enabled recorded-triage URL", async () => {
+  const previous = process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = "true";
+  try {
+    const response = await POST(
+      new Request("http://localhost/api/issue-agent/runs", {
+        method: "POST",
+        body: JSON.stringify({ mode: "recorded-triage", issueUrl: "https://example.com/nope" }),
+      })
+    );
+    const body = await json(response);
+
+    assert.equal(response.status, 400);
+    assert.equal(body.error, "Expected a GitHub issue or pull request URL");
+  } finally {
+    if (previous === undefined) delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+    else process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = previous;
+  }
+});
+
+test("issue-agent run returns recorded context summary with redaction", async () => {
+  const previous = process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  const previousDataDir = process.env.DATA_DIR;
+  process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = "true";
+  process.env.DATA_DIR = mkdtempSync(join(tmpdir(), "issue-agent-route-"));
+  try {
+    const response = await POST(
+      new Request("http://localhost/api/issue-agent/runs", {
+        method: "POST",
+        body: JSON.stringify({
+          mode: "recorded-triage",
+          issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/7",
+          recordedContext: {
+            title: "Review PR mention",
+            body: "Authorization: Bearer sk-routeSecret1234567890",
+            comments: [{ author: "maintainer", body: "please review", isBot: false }],
+          },
+        }),
+      })
+    );
+    const body = await json(response);
+    const context = body.context as Record<string, unknown>;
+
+    assert.equal(response.status, 200);
+    assert.equal(body.auditPath, join(process.env.DATA_DIR, "issue-agent", "audit.jsonl"));
+    assert.equal(context.issueTitle, "Review PR mention");
+    assert.equal(context.intent, "review");
+    assert.equal(context.humanCommentCount, 1);
+    assert.doesNotMatch(String(context.redactedDigestSource), /sk-routeSecret/);
+    assert.match(String(context.redactedDigestSource), /\[REDACTED\]/);
+  } finally {
+    if (previous === undefined) delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+    else process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = previous;
+    if (previousDataDir === undefined) delete process.env.DATA_DIR;
+    else process.env.DATA_DIR = previousDataDir;
+  }
+});
+
+test("issue-agent run accepts recorded GitHub export payloads", async () => {
+  const previous = process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  const previousDataDir = process.env.DATA_DIR;
+  process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = "true";
+  process.env.DATA_DIR = mkdtempSync(join(tmpdir(), "issue-agent-export-route-"));
+  try {
+    const response = await POST(
+      new Request("http://localhost/api/issue-agent/runs", {
+        method: "POST",
+        body: JSON.stringify({
+          mode: "recorded-triage",
+          githubExport: {
+            issue: {
+              title: "Fix GitHub export importer",
+              body: "bug in recorded context path",
+              html_url: "https://github.com/KooshaPari/OmniRoute/issues/88",
+            },
+            comments: [{ user: { login: "maintainer", type: "User" }, body: "please patch" }],
+          },
+        }),
+      })
+    );
+    const body = await json(response);
+    const context = body.context as Record<string, unknown>;
+
+    assert.equal(response.status, 200);
+    assert.equal(body.issueUrl, "https://github.com/KooshaPari/OmniRoute/issues/88");
+    assert.equal(body.issueNumber, 88);
+    assert.equal(context.issueTitle, "Fix GitHub export importer");
+    assert.equal(context.intent, "bugfix");
+    assert.equal(body.auditPath, join(process.env.DATA_DIR, "issue-agent", "audit.jsonl"));
+  } finally {
+    if (previous === undefined) delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+    else process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = previous;
+    if (previousDataDir === undefined) delete process.env.DATA_DIR;
+    else process.env.DATA_DIR = previousDataDir;
+  }
+});
+
+test("issue-agent run sanitizes a forced audit-write failure instead of leaking its absolute path", async () => {
+  const previous = process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+  const previousDataDir = process.env.DATA_DIR;
+  process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = "true";
+  const auditBlockerDir = mkdtempSync(join(tmpdir(), "issue-agent-audit-blocker-"));
+  // appendIssueAgentAuditRecord() (src/lib/issueAgent/audit.ts) does
+  // `mkdir(join(DATA_DIR, "issue-agent"), { recursive: true })`. Pre-creating a
+  // FILE at that exact path forces a real Node fs error (EEXIST) whose raw
+  // `.message` embeds this directory's absolute path -- the same leak shape a
+  // permission-denied or disk-full failure would produce (Hard Rule #12).
+  writeFileSync(join(auditBlockerDir, "issue-agent"), "blocking file, not a directory");
+  process.env.DATA_DIR = auditBlockerDir;
+
+  try {
+    const response = await POST(
+      new Request("http://localhost/api/issue-agent/runs", {
+        method: "POST",
+        body: JSON.stringify({
+          mode: "recorded-triage",
+          issueUrl: "https://github.com/KooshaPari/OmniRoute/issues/1",
+        }),
+      })
+    );
+    const body = await json(response);
+
+    assert.equal(response.status, 400);
+    assert.equal(typeof body.error, "string");
+    const errorMessage = String(body.error);
+    assert.doesNotMatch(errorMessage, /EEXIST|ENOENT|EACCES/, "must not leak the raw errno code");
+    assert.ok(
+      !errorMessage.includes(auditBlockerDir),
+      "must not leak the DATA_DIR absolute path"
+    );
+    assert.ok(
+      !errorMessage.includes("/issue-agent/"),
+      "must not leak the audit subdirectory path"
+    );
+    assert.equal(errorMessage, "Issue Agent request failed due to an internal error");
+  } finally {
+    if (previous === undefined) delete process.env.OMNIROUTE_ISSUE_AGENT_ENABLED;
+    else process.env.OMNIROUTE_ISSUE_AGENT_ENABLED = previous;
+    if (previousDataDir === undefined) delete process.env.DATA_DIR;
+    else process.env.DATA_DIR = previousDataDir;
+  }
+});
diff --git a/tests/unit/m365-web-token-extraction-7078.test.ts b/tests/unit/m365-web-token-extraction-7078.test.ts
new file mode 100644
index 00000000000..68ed08b7fc3
--- /dev/null
+++ b/tests/unit/m365-web-token-extraction-7078.test.ts
@@ -0,0 +1,48 @@
+// Issue #7078 — Microsoft 365 Copilot web token extraction must accept the current
+// m365.cloud.microsoft endpoint (and legacy substrate.office.com / copilot.microsoft.com),
+// not just the old substrate.office.com WS host. Verifies access_token + chathubPath parse.
+import test from "node:test";
+import assert from "node:assert/strict";
+
+const B = await import("../../src/lib/providers/validation/webProvidersB.ts");
+const extract = (raw: string) =>
+  (B as Record<string, any>).extractM365CredentialParts(raw, {});
+
+test("#7078 m365.cloud.microsoft wss URL extracts access_token + chathubPath", () => {
+  const raw =
+    "wss://m365.cloud.microsoft/m365Copilot/Chathub/user@tenant.example.com?access_token=TOKEN123";
+  const parts = extract(raw);
+  assert.equal(parts.accessToken, "TOKEN123");
+  assert.equal(parts.chathubPath, "user@tenant.example.com");
+});
+
+test("#7078 regional subdomain m365.cloud.microsoft also accepted", () => {
+  const raw =
+    "wss://eu.m365.cloud.microsoft/m365Copilot/Chathub/user@tenant?access_token=TOKEN456";
+  const parts = extract(raw);
+  assert.equal(parts.accessToken, "TOKEN456");
+  assert.equal(parts.chathubPath, "user@tenant");
+});
+
+test("#7078 legacy substrate.office.com still works (no regression)", () => {
+  const raw =
+    "wss://substrate.office.com/m365Copilot/Chathub/user@tenant?access_token=LEGACY";
+  const parts = extract(raw);
+  assert.equal(parts.accessToken, "LEGACY");
+  assert.equal(parts.chathubPath, "user@tenant");
+});
+
+test("#7078 key/value string form still parsed", () => {
+  const raw = "access_token=KV;chathubPath=user@tenant";
+  const parts = extract(raw);
+  assert.equal(parts.accessToken, "KV");
+  assert.equal(parts.chathubPath, "user@tenant");
+});
+
+test("#7078 m365.cloud.microsoft with explicit :443 port still extracts (hostname, not host)", () => {
+  const raw =
+    "wss://m365.cloud.microsoft:443/m365Copilot/Chathub/user@tenant?access_token=TOKENPORT";
+  const parts = extract(raw);
+  assert.equal(parts.accessToken, "TOKENPORT");
+  assert.equal(parts.chathubPath, "user@tenant");
+});
diff --git a/tests/unit/mcp-tool-count-dedup-6854.test.ts b/tests/unit/mcp-tool-count-dedup-6854.test.ts
index f578a761cbe..0a2ef9cd30a 100644
--- a/tests/unit/mcp-tool-count-dedup-6854.test.ts
+++ b/tests/unit/mcp-tool-count-dedup-6854.test.ts
@@ -7,8 +7,8 @@ import assert from "node:assert/strict";
 // MCP_TOOLS (open-sse/mcp-server/schemas/tools.ts) and agentSkillTools
 // (open-sse/mcp-server/tools/agentSkillTools.ts), so the additive sum reported 99
 // while only 96 distinct tool names actually exist. countUniqueMcpTools
-// (open-sse/mcp-server/toolCount.ts) fixes this by unioning tool names into a Set
-// before counting, so a tool present in multiple collections is only counted once.
+// (open-sse/mcp-server/toolCount.ts) fixes this by unioning tool names from every
+// registered collection into a Set, so each user-visible tool is counted once.
 
 const { countUniqueMcpTools } = await import("../../open-sse/mcp-server/toolCount.ts");
 const { MCP_TOOLS } = await import("../../open-sse/mcp-server/schemas/tools.ts");
@@ -21,6 +21,7 @@ const { gamificationTools } = await import("../../open-sse/mcp-server/tools/gami
 const { pluginTools } = await import("../../open-sse/mcp-server/tools/pluginTools.ts");
 const { notionTools } = await import("../../open-sse/mcp-server/tools/notionTools.ts");
 const { obsidianTools } = await import("../../open-sse/mcp-server/tools/obsidianTools.ts");
+const { compressionTools } = await import("../../open-sse/mcp-server/tools/compressionTools.ts");
 
 type NamedTool = { name: string };
 
@@ -53,6 +54,7 @@ test("#6854: countUniqueMcpTools de-duplicates tools registered in multiple coll
     pluginTools: pluginTools as unknown as NamedTool[],
     notionTools: notionTools as unknown as NamedTool[],
     obsidianTools: obsidianTools as unknown as NamedTool[],
+    compressionTools: compressionTools as unknown as Record<string, NamedTool>,
   };
 
   const total = countUniqueMcpTools(collections);
@@ -71,11 +73,6 @@ test("#6854: countUniqueMcpTools de-duplicates tools registered in multiple coll
   );
 
   assert.equal(total, uniqueNames.size, "countUniqueMcpTools must equal the unique-name count");
-  assert.equal(
-    total,
-    naiveAdditiveSum - overlap.length,
-    "unique count must be exactly the additive sum minus the double-counted overlap"
-  );
   assert.ok(
     total < naiveAdditiveSum,
     "unique count must be strictly less than the naive additive sum given a known overlap"
diff --git a/tests/unit/mitm-dns-graceful-degrade-6127.test.ts b/tests/unit/mitm-dns-graceful-degrade-6127.test.ts
index b82f6ab7405..4f71925d6fd 100644
--- a/tests/unit/mitm-dns-graceful-degrade-6127.test.ts
+++ b/tests/unit/mitm-dns-graceful-degrade-6127.test.ts
@@ -27,8 +27,7 @@ function makeSpyLogger() {
   return {
     logger: {
       error: (payload: unknown, msg: string) => errorCalls.push({ payload, msg }),
-      info: (payload: unknown, msg?: string) =>
-        infoCalls.push({ payload, msg: msg ?? "" }),
+      info: (payload: unknown, msg?: string) => infoCalls.push({ payload, msg: msg ?? "" }),
     },
     errorCalls,
     infoCalls,
@@ -83,8 +82,7 @@ test("provisionDnsEntries: a failing agent/custom step does not stop the others
       // Custom-hosts call must still happen even after default + agent errors.
       if (hosts.includes("custom.example.com")) customCalled = true;
     },
-    getAgentStates: () =>
-      [{ dns_enabled: true, agent_id: "__nonexistent_agent__" }] as never,
+    getAgentStates: () => [{ dns_enabled: true, agent_id: "__nonexistent_agent__" }] as never,
     listEnabledCustomHosts: () => [{ host: "custom.example.com" }] as never,
     logger: spy.logger,
   });
@@ -105,3 +103,124 @@ test("provisionDnsEntries: happy path calls the default step and does not log er
   assert.ok(defaultCalled, "default DNS step must be invoked");
   assert.equal(spy.errorCalls.length, 0, "no error logs on the happy path");
 });
+
+test("provisionDnsEntries: SKIP_ANTIGRAVITY_DNS=true skips ALL DNS steps (default + agents + custom)", async () => {
+  const prev = process.env.SKIP_ANTIGRAVITY_DNS;
+  process.env.SKIP_ANTIGRAVITY_DNS = "true";
+  try {
+    const spy = makeSpyLogger();
+    let defaultCalled = false;
+    let hostsCalled = false;
+    await provisionDnsEntries("pw", {
+      addDefaultDns: async () => {
+        defaultCalled = true;
+      },
+      addHostsDns: async () => {
+        hostsCalled = true;
+      },
+      getAgentStates: () => [{ dns_enabled: true, agent_id: "antigravity" }] as never,
+      listEnabledCustomHosts: () => [{ host: "custom.example.com" }] as never,
+      logger: spy.logger,
+    });
+    assert.ok(!defaultCalled, "default DNS step must NOT be called when SKIP_ANTIGRAVITY_DNS=true");
+    assert.ok(!hostsCalled, "addHostsDns must NOT be called when SKIP_ANTIGRAVITY_DNS=true");
+    assert.equal(spy.errorCalls.length, 0, "no errors expected");
+    assert.ok(
+      spy.infoCalls.some(
+        (c) => typeof c.payload === "string" && c.payload.includes("SKIP_ANTIGRAVITY_DNS")
+      ),
+      "info log must mention SKIP_ANTIGRAVITY_DNS"
+    );
+  } finally {
+    // Restore the original env var value.
+    if (prev === undefined) {
+      delete process.env.SKIP_ANTIGRAVITY_DNS;
+    } else {
+      process.env.SKIP_ANTIGRAVITY_DNS = prev;
+    }
+  }
+});
+
+test("provisionDnsEntries: SKIP_ANTIGRAVITY_DNS=false does NOT skip DNS steps", async () => {
+  const prev = process.env.SKIP_ANTIGRAVITY_DNS;
+  process.env.SKIP_ANTIGRAVITY_DNS = "false";
+  try {
+    const spy = makeSpyLogger();
+    let defaultCalled = false;
+    await provisionDnsEntries("pw", {
+      addDefaultDns: async () => {
+        defaultCalled = true;
+      },
+      getAgentStates: () => [],
+      listEnabledCustomHosts: () => [],
+      logger: spy.logger,
+    });
+    assert.ok(defaultCalled, "default DNS step must be called when SKIP_ANTIGRAVITY_DNS=false");
+    assert.equal(spy.errorCalls.length, 0, "no errors expected");
+  } finally {
+    if (prev === undefined) {
+      delete process.env.SKIP_ANTIGRAVITY_DNS;
+    } else {
+      process.env.SKIP_ANTIGRAVITY_DNS = prev;
+    }
+  }
+});
+
+test("provisionDnsEntries: canElevate()=false skips ALL DNS steps (default + agents + custom)", async () => {
+  const spy = makeSpyLogger();
+  let defaultCalled = false;
+  let hostsCalled = false;
+  await provisionDnsEntries("pw", {
+    addDefaultDns: async () => {
+      defaultCalled = true;
+    },
+    addHostsDns: async () => {
+      hostsCalled = true;
+    },
+    canElevate: () => false,
+    getAgentStates: () => [{ dns_enabled: true, agent_id: "antigravity" }] as never,
+    listEnabledCustomHosts: () => [{ host: "custom.example.com" }] as never,
+    logger: spy.logger,
+  });
+  assert.ok(!defaultCalled, "default DNS step must NOT be called when canElevate() is false");
+  assert.ok(!hostsCalled, "addHostsDns must NOT be called when canElevate() is false");
+  assert.equal(spy.errorCalls.length, 0, "no errors expected");
+  assert.ok(
+    spy.infoCalls.some(
+      (c) => typeof c.payload === "string" && c.payload.includes("sudo not available")
+    ),
+    "info log must mention sudo not available"
+  );
+});
+
+test("provisionDnsEntries: canElevate() returning true proceeds with DNS provisioning", async () => {
+  const spy = makeSpyLogger();
+  let defaultCalled = false;
+  let agentCalled = false;
+  let customCalled = false;
+  const capturedPasswords: string[] = [];
+  await provisionDnsEntries("pw", {
+    addDefaultDns: async () => {
+      defaultCalled = true;
+    },
+    addHostsDns: async (hosts: string[], sudoPassword: string) => {
+      capturedPasswords.push(sudoPassword);
+      if (hosts.some((h) => h.includes("googleapis.com"))) agentCalled = true;
+      if (hosts.includes("custom.example.com")) customCalled = true;
+    },
+    canElevate: () => true,
+    getAgentStates: () => [{ dns_enabled: true, agent_id: "antigravity" }] as never,
+    listEnabledCustomHosts: () => [{ host: "custom.example.com" }] as never,
+    logger: spy.logger,
+  });
+  assert.ok(defaultCalled, "default DNS step must be called when canElevate() is true");
+  assert.ok(agentCalled, "agent DNS step must be called when canElevate() is true");
+  assert.ok(customCalled, "custom DNS step must be called when canElevate() is true");
+  assert.equal(spy.errorCalls.length, 0, "no errors expected on happy path");
+  // Verify sudoPassword is passed through to addHostsDns.
+  assert.equal(capturedPasswords.length, 2, "addHostsDns called twice (agents + custom)");
+  assert.ok(
+    capturedPasswords.every((p) => p === "pw"),
+    "sudoPassword must be forwarded to addHostsDns"
+  );
+});
diff --git a/tests/unit/models-catalog-combo-metadata.test.ts b/tests/unit/models-catalog-combo-metadata.test.ts
new file mode 100644
index 00000000000..51d5c5360bc
--- /dev/null
+++ b/tests/unit/models-catalog-combo-metadata.test.ts
@@ -0,0 +1,119 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-combo-metadata-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+process.env.API_KEY_SECRET ||= "combo-metadata-test-secret";
+
+const core = await import("../../src/lib/db/core.ts");
+const providersDb = await import("../../src/lib/db/providers.ts");
+const combosDb = await import("../../src/lib/db/combos.ts");
+const catalog = await import("../../src/app/api/v1/models/catalog.ts");
+
+test.after(() => {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+test("single-target combo preserves its direct model metadata", async () => {
+  await providersDb.createProviderConnection({
+    provider: "codex",
+    authType: "oauth",
+    name: "codex-gpt-5.6-single-target-combo",
+    accessToken: "codex-test-token",
+    isActive: true,
+    testStatus: "active",
+    providerSpecificData: {},
+  });
+  await combosDb.createCombo({
+    name: "gpt-5.6-sol-combo",
+    strategy: "auto",
+    models: ["codex/gpt-5.6-sol"],
+  });
+
+  const response = await catalog.getUnifiedModelsResponse(
+    new Request("http://localhost/api/v1/models")
+  );
+  const body = (await response.json()) as { data: Array<Record<string, unknown>> };
+  const direct = body.data.find((item) => item.id === "cx/gpt-5.6-sol");
+  const combo = body.data.find((item) => item.id === "gpt-5.6-sol-combo");
+
+  assert.equal(response.status, 200);
+  assert.ok(direct);
+  assert.ok(combo);
+  for (const field of [
+    "context_length",
+    "max_input_tokens",
+    "max_output_tokens",
+    "input_modalities",
+    "output_modalities",
+    "capabilities",
+  ]) {
+    assert.deepEqual(combo[field], direct[field], field);
+  }
+});
+
+test("single-target combo respects registry reasoning overrides before specs", async () => {
+  await providersDb.createProviderConnection({
+    provider: "command-code",
+    authType: "apikey",
+    name: "command-code-gpt-5.4-mini-combo",
+    apiKey: "command-code-test-key",
+    isActive: true,
+    testStatus: "active",
+    providerSpecificData: {},
+  });
+  await combosDb.createCombo({
+    name: "gpt-5.4-mini-command-code-combo",
+    strategy: "auto",
+    models: ["command-code/gpt-5.4-mini"],
+  });
+
+  const response = await catalog.getUnifiedModelsResponse(
+    new Request("http://localhost/api/v1/models")
+  );
+  const body = (await response.json()) as { data: Array<Record<string, unknown>> };
+  const combo = body.data.find((item) => item.id === "gpt-5.4-mini-command-code-combo");
+
+  assert.equal(response.status, 200);
+  assert.ok(combo);
+  const capabilities = combo.capabilities as Record<string, unknown>;
+  assert.equal(capabilities.reasoning, false);
+  assert.equal(capabilities.thinking, false);
+  assert.equal(capabilities.supportsThinking, false);
+  assert.equal(Object.hasOwn(capabilities, "effort_tiers"), false);
+});
+
+test("single-target combo respects resolved reasoning deny patterns", async () => {
+  await providersDb.createProviderConnection({
+    provider: "antigravity",
+    authType: "oauth",
+    name: "antigravity-gemini-no-thinking-combo",
+    accessToken: "antigravity-test-token",
+    isActive: true,
+    testStatus: "active",
+    providerSpecificData: {},
+  });
+  await combosDb.createCombo({
+    name: "antigravity-gemini-no-thinking-combo",
+    strategy: "auto",
+    models: ["antigravity/gemini-3.1-pro-high"],
+  });
+
+  const response = await catalog.getUnifiedModelsResponse(
+    new Request("http://localhost/api/v1/models")
+  );
+  const body = (await response.json()) as { data: Array<Record<string, unknown>> };
+  const combo = body.data.find((item) => item.id === "antigravity-gemini-no-thinking-combo");
+
+  assert.equal(response.status, 200);
+  assert.ok(combo);
+  const capabilities = combo.capabilities as Record<string, unknown>;
+  assert.equal(capabilities.reasoning, false);
+  assert.equal(capabilities.thinking, false);
+  assert.equal(capabilities.supportsThinking, false);
+  assert.equal(Object.hasOwn(capabilities, "effort_tiers"), false);
+});
diff --git a/tests/unit/models-catalog-route.test.ts b/tests/unit/models-catalog-route.test.ts
index 0a7b1f897ed..1a51272b694 100644
--- a/tests/unit/models-catalog-route.test.ts
+++ b/tests/unit/models-catalog-route.test.ts
@@ -1419,7 +1419,7 @@ test("v1 models catalog auto-calculates combo context_length from targets when n
   });
 
   // Create a combo with targets having different context limits.
-  // openai/gpt-4o context = 128000, claude/claude-sonnet-4-6 = 200000.
+  // openai/gpt-4o context = 128000, claude/claude-sonnet-4-6 = 1000000 (#7129: 1M GA).
   // The combo should expose context_length = min = 128000.
   const combo = await combosDb.createCombo({
     name: "auto-context-combo",
diff --git a/tests/unit/node-runtime-support.test.ts b/tests/unit/node-runtime-support.test.ts
index dd49708de7a..7cbce506971 100644
--- a/tests/unit/node-runtime-support.test.ts
+++ b/tests/unit/node-runtime-support.test.ts
@@ -1,5 +1,6 @@
 import test from "node:test";
 import assert from "node:assert/strict";
+import { readFileSync } from "node:fs";
 
 import {
   SUPPORTED_NODE_RANGE,
@@ -82,3 +83,11 @@ test("CLI runtime support stays aligned with the shared runtime policy", () => {
   assert.deepEqual(getCliNodeRuntimeSupport("22.22.2"), getNodeRuntimeSupport("22.22.2"));
   assert.equal(getCliNodeRuntimeWarning("27.1.0"), getNodeRuntimeWarning("27.1.0"));
 });
+
+test("package.json engines.node equals SUPPORTED_NODE_RANGE (install gate == runtime gate)", () => {
+  // Regression guard for #7446: npm consults `engines.node` at install time while the
+  // runtime rejects anything outside SUPPORTED_NODE_RANGE. If the two drift, npm installs
+  // OmniRoute on a Node version the runtime then refuses to run (the Node 22.0.0–22.22.1 gap).
+  const pkg = JSON.parse(readFileSync(new URL("../../package.json", import.meta.url), "utf8"));
+  assert.equal(pkg.engines.node, SUPPORTED_NODE_RANGE);
+});
diff --git a/tests/unit/nvidia-catalog-drift.test.ts b/tests/unit/nvidia-catalog-drift.test.ts
new file mode 100644
index 00000000000..8e8b8dfcf88
--- /dev/null
+++ b/tests/unit/nvidia-catalog-drift.test.ts
@@ -0,0 +1,28 @@
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import { computeNvidiaCatalogDrift } from "../../scripts/check/check-nvidia-catalog-drift.ts";
+
+test("NVIDIA catalog drift separates new live ids from removed documented ids", () => {
+  assert.deepEqual(
+    computeNvidiaCatalogDrift(
+      ["z-ai/glm-5.2", "openai/gpt-oss-120b", "new/model", "new/model"],
+      ["z-ai/glm-5.2", "openai/gpt-oss-120b", "removed/model"],
+      ["z-ai/glm-5.2", "openai/gpt-oss-120b", "retired/model"]
+    ),
+    {
+      liveCount: 3,
+      reviewedLiveCount: 3,
+      documentedFreeCount: 3,
+      newLiveIds: ["new/model"],
+      removedLiveIds: ["removed/model"],
+      documentedMissingUpstreamIds: ["retired/model"],
+    }
+  );
+});
+
+test("NVIDIA catalog drift is empty when live and documented ids match", () => {
+  const drift = computeNvidiaCatalogDrift(["b", "a"], ["a", "b"], ["a"]);
+  assert.deepEqual(drift.newLiveIds, []);
+  assert.deepEqual(drift.removedLiveIds, []);
+  assert.deepEqual(drift.documentedMissingUpstreamIds, []);
+});
diff --git a/tests/unit/oauth-device-code-endpoints.test.ts b/tests/unit/oauth-device-code-endpoints.test.ts
new file mode 100644
index 00000000000..325a1b259d9
--- /dev/null
+++ b/tests/unit/oauth-device-code-endpoints.test.ts
@@ -0,0 +1,32 @@
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import { readFileSync } from "node:fs";
+import { fileURLToPath } from "node:url";
+import { dirname, resolve } from "node:path";
+
+// Regression guards for the two device-code providers that returned "ошибка сервера OmniRoute":
+//
+//   qwen        — QWEN_CONFIG pointed at the bare host qwen.ai, whose device/token paths 404.
+//                 The working qwen-code device flow lives at chat.qwen.ai (verified live: 200 +
+//                 a valid device_code). Guard: the config uses chat.qwen.ai, never bare qwen.ai.
+//   codebuddy-cn— the Tencent state endpoint reads `platform` from the QUERY string, not the JSON
+//                 body; body-only returned 400 "platform is empty" (verified live). The fix passes
+//                 it as a query param. Guard: requestDeviceCode builds the URL with ?platform=.
+//
+// Source-level: the real validation is the live upstream 200 (can't be hit from CI); these pin
+// the exact change so a revert to the broken host / body-only platform fails here.
+const here = dirname(fileURLToPath(import.meta.url));
+const read = (p: string) => readFileSync(resolve(here, "../..", p), "utf8");
+
+test("qwen device-code config uses chat.qwen.ai (not the 404ing bare qwen.ai)", () => {
+  const oauth = read("src/lib/oauth/constants/oauth.ts");
+  const qwenBlock = oauth.slice(oauth.indexOf("QWEN_CONFIG"), oauth.indexOf("QWEN_CONFIG") + 700);
+  assert.match(qwenBlock, /chat\.qwen\.ai\/api\/v1\/oauth2\/device\/code/, "deviceCodeUrl host");
+  assert.match(qwenBlock, /chat\.qwen\.ai\/api\/v1\/oauth2\/token/, "tokenUrl host");
+  assert.doesNotMatch(qwenBlock, /"https:\/\/qwen\.ai\//, "must not use the bare qwen.ai host");
+});
+
+test("codebuddy-cn device-code sends platform as a query param (not body-only)", () => {
+  const cb = read("src/lib/oauth/providers/codebuddy-cn.ts");
+  assert.match(cb, /\?platform=\$\{encodeURIComponent\(config\.platform\)\}/, "platform query param");
+});
diff --git a/tests/unit/oauth-device-code-error-transparency.test.ts b/tests/unit/oauth-device-code-error-transparency.test.ts
new file mode 100644
index 00000000000..ec727770947
--- /dev/null
+++ b/tests/unit/oauth-device-code-error-transparency.test.ts
@@ -0,0 +1,32 @@
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import { readFileSync } from "node:fs";
+import { fileURLToPath } from "node:url";
+import { dirname, resolve } from "node:path";
+
+// Regression guard for "qwen / codebuddy-cn device-code show 'ошибка сервера OmniRoute'".
+// The dynamic OAuth GET handler used to swallow EVERY thrown error into a generic
+// `{ error: "Internal server error" }` 500, so a device-code upstream failure (qwen.ai /
+// copilot.tencent.com — geo-block, outage, bad client) was indistinguishable from a real
+// server bug. The fix surfaces the SANITIZED upstream message via sanitizeErrorMessage.
+//
+// Source-level guard: the route runs inside the Next server (auth + module graph), so a
+// behavioural test would need a full request/auth/upstream mock; this pins the exact change.
+const here = dirname(fileURLToPath(import.meta.url));
+const route = readFileSync(
+  resolve(here, "../../src/app/api/oauth/[provider]/[action]/route.ts"),
+  "utf8"
+);
+
+test("OAuth GET catch surfaces the sanitized upstream reason, not a bare generic 500", () => {
+  assert.match(
+    route,
+    /const detail = sanitizeErrorMessage\(/,
+    "the GET catch must sanitize and surface the real error"
+  );
+  assert.match(
+    route,
+    /error:\s*detail\s*\|\|\s*"Internal server error"/,
+    "the GET catch must return the sanitized detail (falling back to the generic only when empty)"
+  );
+});
diff --git a/tests/unit/openai-responses-request-split.test.ts b/tests/unit/openai-responses-request-split.test.ts
index 50699805a61..13624181c7f 100644
--- a/tests/unit/openai-responses-request-split.test.ts
+++ b/tests/unit/openai-responses-request-split.test.ts
@@ -54,3 +54,70 @@ test("both directions are callable via the host module", async () => {
   ) as Record<string, unknown>;
   assert.ok(Array.isArray(out.input));
 });
+
+test("mid-conversation system turns are preserved as developer-role input items (#6954)", async () => {
+  const mod = await import("../../open-sse/translator/request/openai-responses.ts");
+  const out = mod.openaiToOpenAIResponsesRequest(
+    "gpt-5.5",
+    {
+      messages: [
+        { role: "system", content: "You are a helpful assistant." },
+        { role: "user", content: "Hello" },
+        { role: "system", content: "Available agent types: claude, claude-code-guide" },
+        { role: "assistant", content: "Hi there!" },
+        { role: "system", content: "Reminder: stay concise." },
+      ],
+    },
+    true,
+    null
+  ) as Record<string, unknown>;
+
+  // The FIRST system turn becomes `instructions`.
+  assert.equal(out.instructions, "You are a helpful assistant.");
+
+  const input = out.input as Array<Record<string, unknown>>;
+
+  // Mid-conversation system turns must survive as developer-role input items —
+  // not dropped, and never misattributed as assistant (#6954).
+  const developerItems = input.filter((i) => i.role === "developer");
+  assert.equal(developerItems.length, 2, "two mid-conversation system turns should become developer items");
+  assert.deepEqual(
+    (developerItems[0].content as Array<Record<string, unknown>>).map((c) => c.text),
+    ["Available agent types: claude, claude-code-guide"]
+  );
+  assert.deepEqual(
+    (developerItems[1].content as Array<Record<string, unknown>>).map((c) => c.text),
+    ["Reminder: stay concise."]
+  );
+
+  // No harness-injected system content should appear as assistant prose.
+  const assistantText = input
+    .filter((i) => i.role === "assistant")
+    .map((i) => JSON.stringify(i.content))
+    .join(" ");
+  assert.ok(!assistantText.includes("Available agent types"));
+  assert.ok(!assistantText.includes("Reminder: stay concise."));
+});
+
+test("mid-conversation system content as an array with a bare string survives (#6954 follow-up)", async () => {
+  const mod = await import("../../open-sse/translator/request/openai-responses.ts");
+  const out = mod.openaiToOpenAIResponsesRequest(
+    "gpt-5.5",
+    {
+      messages: [
+        { role: "system", content: "Base instructions." },
+        { role: "user", content: "Hi" },
+        { role: "system", content: ["Remember to be concise."] },
+      ],
+    },
+    true,
+    null
+  ) as Record<string, unknown>;
+  const input = out.input as Array<Record<string, unknown>>;
+  const devItems = input.filter((i) => i.role === "developer");
+  assert.equal(devItems.length, 1, "array-form system turn should become a developer item");
+  assert.deepEqual(
+    (devItems[0].content as Array<Record<string, unknown>>).map((c) => c.text),
+    ["Remember to be concise."]
+  );
+});
diff --git a/tests/unit/opencode-go-effort-aliases-6922.test.ts b/tests/unit/opencode-go-effort-aliases-6922.test.ts
new file mode 100644
index 00000000000..67831a7ab41
--- /dev/null
+++ b/tests/unit/opencode-go-effort-aliases-6922.test.ts
@@ -0,0 +1,164 @@
+/**
+ * Issue #6922 — Effort-tier aliases for glm-5.2 and mimo-v2.5 on opencode-go.
+ *
+ * Tests import and call the real `parseEffortLevel` from OpencodeExecutor
+ * to verify effort-tier parsing works for all registered models.
+ *
+ * The OpencodeExecutor must:
+ *  1. Rewrite effort-alias model ids to their canonical base id
+ *  2. Inject `reasoning_effort` if not already set
+ *
+ * Previously only deepseek-v4-pro had aliases. Now glm-5.2 and mimo-v2.5
+ * also have high/max tiers.
+ */
+
+import test from "node:test";
+import assert from "node:assert/strict";
+
+// ─── Stub ESM loader hooks so the executor can be imported in a bare
+//     node:test process without triggering side effects (DB init, fetch,
+//     etc.). We only need parseEffortLevel, which is a pure function. ───
+
+const { parseEffortLevel, OpencodeExecutor } = (await import(
+  "../../open-sse/executors/opencode.ts"
+)) as {
+  parseEffortLevel: (model: string) => { baseModel: string; effort: string } | null;
+  OpencodeExecutor: new (provider: string) => {
+    transformRequest: (
+      model: string,
+      body: Record<string, unknown>,
+      stream: boolean,
+      credentials: unknown
+    ) => Record<string, unknown>;
+  };
+};
+
+// ─── DeepSeek v4-pro: all 4 tiers ─────────────────────────────────────────
+
+test("#6922 parseEffortLevel: deepseek-v4-pro-low → low", () => {
+  const result = parseEffortLevel("deepseek-v4-pro-low");
+  assert.deepEqual(result, { baseModel: "deepseek-v4-pro", effort: "low" });
+});
+
+test("#6922 parseEffortLevel: deepseek-v4-pro-medium → medium", () => {
+  const result = parseEffortLevel("deepseek-v4-pro-medium");
+  assert.deepEqual(result, { baseModel: "deepseek-v4-pro", effort: "medium" });
+});
+
+test("#6922 parseEffortLevel: deepseek-v4-pro-high → high", () => {
+  const result = parseEffortLevel("deepseek-v4-pro-high");
+  assert.deepEqual(result, { baseModel: "deepseek-v4-pro", effort: "high" });
+});
+
+test("#6922 parseEffortLevel: deepseek-v4-pro-max → max", () => {
+  const result = parseEffortLevel("deepseek-v4-pro-max");
+  assert.deepEqual(result, { baseModel: "deepseek-v4-pro", effort: "max" });
+});
+
+// ─── GLM-5.2: high + max only ────────────────────────────────────────────
+
+test("#6922 parseEffortLevel: glm-5.2-high → high", () => {
+  const result = parseEffortLevel("glm-5.2-high");
+  assert.deepEqual(result, { baseModel: "glm-5.2", effort: "high" });
+});
+
+test("#6922 parseEffortLevel: glm-5.2-max → max", () => {
+  const result = parseEffortLevel("glm-5.2-max");
+  assert.deepEqual(result, { baseModel: "glm-5.2", effort: "max" });
+});
+
+// ─── MiMo-V2.5: high + max only ──────────────────────────────────────────
+
+test("#6922 parseEffortLevel: mimo-v2.5-high → high", () => {
+  const result = parseEffortLevel("mimo-v2.5-high");
+  assert.deepEqual(result, { baseModel: "mimo-v2.5", effort: "high" });
+});
+
+test("#6922 parseEffortLevel: mimo-v2.5-max → max", () => {
+  const result = parseEffortLevel("mimo-v2.5-max");
+  assert.deepEqual(result, { baseModel: "mimo-v2.5", effort: "max" });
+});
+
+// ─── Negative cases ────────────────────────────────────────────────────────
+
+test("#6922 parseEffortLevel: unknown model → null", () => {
+  const result = parseEffortLevel("nonexistent-model-high");
+  assert.strictEqual(result, null);
+});
+
+test("#6922 parseEffortLevel: glm-5.2-low → null (unsupported tier)", () => {
+  const result = parseEffortLevel("glm-5.2-low");
+  assert.strictEqual(result, null);
+});
+
+test("#6922 parseEffortLevel: mimo-v2.5-medium → null (unsupported tier)", () => {
+  const result = parseEffortLevel("mimo-v2.5-medium");
+  assert.strictEqual(result, null);
+});
+
+test("#6922 parseEffortLevel: empty string → null", () => {
+  assert.strictEqual(parseEffortLevel(""), null);
+});
+
+test("#6922 parseEffortLevel: base model without tier → null", () => {
+  assert.strictEqual(parseEffortLevel("glm-5.2"), null);
+});
+
+// ─── transformRequest: end-to-end model-id rewrite + reasoning_effort inject ──
+//
+// parseEffortLevel is a pure function, but the actual bug (#6922) surfaces
+// through OpencodeExecutor.transformRequest — the caller that rewrites the
+// outbound model id and injects reasoning_effort. These tests exercise that
+// public entry point directly so a broken wiring (e.g. parseEffortLevel
+// correct but never called, or its result dropped) would fail here even if
+// the parseEffortLevel-only tests above stayed green.
+
+const CREDENTIALS = { apiKey: "k" } as Record<string, unknown>;
+
+test("#6922 transformRequest: glm-5.2-high → model rewritten to glm-5.2, reasoning_effort injected", () => {
+  const executor = new OpencodeExecutor("opencode-go");
+  const body = { model: "glm-5.2-high", messages: [{ role: "user", content: "hi" }] };
+
+  const out = executor.transformRequest("glm-5.2-high", body, true, CREDENTIALS);
+
+  assert.equal(out.model, "glm-5.2", "model id must be rewritten to the base id");
+  assert.equal(out.reasoning_effort, "high", "reasoning_effort must be injected from the alias");
+});
+
+test("#6922 transformRequest: mimo-v2.5-max → model rewritten to mimo-v2.5, reasoning_effort injected", () => {
+  const executor = new OpencodeExecutor("opencode-go");
+  const body = { model: "mimo-v2.5-max", messages: [{ role: "user", content: "hi" }] };
+
+  const out = executor.transformRequest("mimo-v2.5-max", body, true, CREDENTIALS);
+
+  assert.equal(out.model, "mimo-v2.5", "model id must be rewritten to the base id");
+  assert.equal(out.reasoning_effort, "max", "reasoning_effort must be injected from the alias");
+});
+
+test("#6922 transformRequest: does not clobber an already-set reasoning_effort", () => {
+  const executor = new OpencodeExecutor("opencode-go");
+  const body = {
+    model: "glm-5.2-high",
+    reasoning_effort: "caller-supplied",
+    messages: [{ role: "user", content: "hi" }],
+  };
+
+  const out = executor.transformRequest("glm-5.2-high", body, true, CREDENTIALS);
+
+  assert.equal(out.model, "glm-5.2", "model id is still rewritten to the base id");
+  assert.equal(
+    out.reasoning_effort,
+    "caller-supplied",
+    "an explicit reasoning_effort already on the body must not be overwritten"
+  );
+});
+
+test("#6922 transformRequest: unaliased model passes through unchanged", () => {
+  const executor = new OpencodeExecutor("opencode-go");
+  const body = { model: "gpt-4o-mini", messages: [{ role: "user", content: "hi" }] };
+
+  const out = executor.transformRequest("gpt-4o-mini", body, true, CREDENTIALS);
+
+  assert.equal(out.model, "gpt-4o-mini", "unaliased model id is left untouched");
+  assert.equal(out.reasoning_effort, undefined, "no reasoning_effort is injected for a non-tier model");
+});
diff --git a/tests/unit/output-token-budget.test.ts b/tests/unit/output-token-budget.test.ts
new file mode 100644
index 00000000000..d4a5aec53aa
--- /dev/null
+++ b/tests/unit/output-token-budget.test.ts
@@ -0,0 +1,66 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { enforceOutputTokenBudget } from "../../open-sse/handlers/chatCore/outputTokenBudget.ts";
+
+test("rejects a prompt that cannot leave one output token", () => {
+  const result = enforceOutputTokenBudget({ max_tokens: 8192 }, 527_058, 128_000);
+
+  assert.deepEqual(result, {
+    ok: false,
+    estimatedInputTokens: 527_058,
+    contextLimit: 128_000,
+  });
+});
+
+test("caps a positive output budget to the target's remaining context", () => {
+  const input = { messages: [], max_tokens: 12_000 };
+  const result = enforceOutputTokenBudget(input, 127_000, 128_000);
+
+  assert.equal(result.ok, true);
+  if (!result.ok) return;
+  assert.equal(result.body.max_tokens, 1_000);
+  assert.equal(input.max_tokens, 12_000, "must not mutate the shared combo request body");
+});
+
+test("removes non-positive numeric output limits before upstream dispatch", () => {
+  const result = enforceOutputTokenBudget(
+    { max_tokens: -398_464, max_completion_tokens: 0 },
+    1_000,
+    128_000
+  );
+
+  assert.equal(result.ok, true);
+  if (!result.ok) return;
+  assert.equal("max_tokens" in result.body, false);
+  assert.equal("max_completion_tokens" in result.body, false);
+});
+
+test("caps max_output_tokens to the target's remaining context", () => {
+  const result = enforceOutputTokenBudget({ max_output_tokens: 12_000 }, 127_000, 128_000);
+
+  assert.equal(result.ok, true);
+  if (!result.ok) return;
+  assert.equal(result.body.max_output_tokens, 1_000);
+});
+
+test("accepts a missing request body when output budget remains", () => {
+  const result = enforceOutputTokenBudget(null, 1_000, 128_000);
+
+  assert.deepEqual(result, {
+    ok: true,
+    body: {},
+    availableOutputTokens: 127_000,
+    adjustedFields: [],
+  });
+});
+
+test("rejects when a Claude target's default output budget does not fit", () => {
+  const result = enforceOutputTokenBudget({}, 70_000, 128_000, 64_000);
+
+  assert.deepEqual(result, {
+    ok: false,
+    estimatedInputTokens: 70_000,
+    contextLimit: 128_000,
+  });
+});
diff --git a/tests/unit/provider-connections-quota-threshold.test.ts b/tests/unit/provider-connections-quota-threshold.test.ts
index d9bfebe9359..5caaeea3699 100644
--- a/tests/unit/provider-connections-quota-threshold.test.ts
+++ b/tests/unit/provider-connections-quota-threshold.test.ts
@@ -162,3 +162,26 @@ test("updateProviderConnectionSchema rejects out-of-range values", () => {
     assert.equal(result.success, false, `expected window5h=${v} to be rejected`);
   }
 });
+
+test("provider quota visibility defaults to visible and persists explicit changes", async () => {
+  const created = await providersDb.createProviderConnection({
+    provider: "codex",
+    authType: "apikey",
+    name: "Codex Visibility",
+    apiKey: "sk-visibility",
+  });
+  assert.equal(created.quotaVisible, true);
+
+  const hidden = await providersDb.updateProviderConnection(created.id, { quotaVisible: false });
+  assert.equal(hidden.quotaVisible, false);
+  assert.equal((await providersDb.getProviderConnectionById(created.id)).quotaVisible, false);
+
+  const visible = await providersDb.updateProviderConnection(created.id, { quotaVisible: true });
+  assert.equal(visible.quotaVisible, true);
+  assert.equal((await providersDb.getProviderConnectionById(created.id)).quotaVisible, true);
+});
+
+test("updateProviderConnectionSchema accepts only boolean quota visibility", () => {
+  assert.equal(updateProviderConnectionSchema.safeParse({ quotaVisible: false }).success, true);
+  assert.equal(updateProviderConnectionSchema.safeParse({ quotaVisible: "false" }).success, false);
+});
diff --git a/tests/unit/provider-limits-ui.test.ts b/tests/unit/provider-limits-ui.test.ts
index 895fdb8ac3d..caf2903cf36 100644
--- a/tests/unit/provider-limits-ui.test.ts
+++ b/tests/unit/provider-limits-ui.test.ts
@@ -335,7 +335,21 @@ test("usage namespace includes Provider Limits UI translation keys", () => {
     "forceRefresh",
     "resetCreditsLabel",
     "redeemResetCredit",
+    "manageResetCredits",
+    "viewResetCredits",
+    "resetCreditsModalTitle",
+    "resetCreditsModalExplainer",
+    "resetCreditsLoadFailed",
+    "resetCreditsDetailsUnavailable",
+    "noResetCreditsAvailable",
+    "resetCreditDefaultTitle",
+    "resetCreditExpiresFirst",
+    "resetCreditExpiresAt",
+    "resetCreditNoExpiry",
+    "redeemThisResetCredit",
+    "confirmRedeemResetCreditTitle",
     "confirmRedeemResetCredit",
+    "confirmRedeemResetCreditButton",
     "resetCreditRedeemed",
     "resetCreditRedeemFailed",
   ]) {
diff --git a/tests/unit/provider-quota-visibility.test.ts b/tests/unit/provider-quota-visibility.test.ts
new file mode 100644
index 00000000000..665687feed0
--- /dev/null
+++ b/tests/unit/provider-quota-visibility.test.ts
@@ -0,0 +1,18 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import {
+  isProviderQuotaVisible,
+  supportsProviderQuota,
+} from "../../src/shared/utils/providerQuotaVisibility.ts";
+
+test("provider quota visibility is opt-out so existing connections stay visible", () => {
+  assert.equal(isProviderQuotaVisible({}), true);
+  assert.equal(isProviderQuotaVisible({ quotaVisible: true }), true);
+  assert.equal(isProviderQuotaVisible({ quotaVisible: false }), false);
+});
+
+test("quota visibility controls are limited to providers with quota support", () => {
+  assert.equal(supportsProviderQuota("codex"), true);
+  assert.equal(supportsProviderQuota("openai"), false);
+});
diff --git a/tests/unit/proxy-bulk-import-dedup-7594.test.ts b/tests/unit/proxy-bulk-import-dedup-7594.test.ts
new file mode 100644
index 00000000000..1ae2c0497b5
--- /dev/null
+++ b/tests/unit/proxy-bulk-import-dedup-7594.test.ts
@@ -0,0 +1,105 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+// Regression test for #7594: bulk-importing proxies that share the same
+// host:port but differ in username/password must create DISTINCT entries, not
+// repeatedly update the first existing row. Rotating residential/gateway proxies
+// commonly route every credential through one host:port, so host+port alone is
+// not a stable identity — the credential tuple is.
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-proxy-dedup-7594-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const core = await import("../../src/lib/db/core.ts");
+const proxiesDb = await import("../../src/lib/db/proxies.ts");
+
+async function resetStorage() {
+  core.resetDbInstance();
+
+  for (let attempt = 0; attempt < 10; attempt++) {
+    try {
+      if (fs.existsSync(TEST_DATA_DIR)) {
+        fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+      }
+      break;
+    } catch (error) {
+      const code = (error as NodeJS.ErrnoException)?.code;
+      if ((code === "EBUSY" || code === "EPERM") && attempt < 9) {
+        await new Promise((resolve) => setTimeout(resolve, 50 * (attempt + 1)));
+      } else {
+        throw error;
+      }
+    }
+  }
+
+  fs.mkdirSync(TEST_DATA_DIR, { recursive: true });
+}
+
+test.beforeEach(async () => {
+  await resetStorage();
+});
+
+test.after(async () => {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+test("upsertProxy creates distinct entries for same host:port with different credentials (#7594)", async () => {
+  const base = { name: "Gateway", type: "http" as const, host: "gw.proxy.local", port: 8080 };
+
+  const first = await proxiesDb.upsertProxy({ ...base, username: "user-1", password: "pass-1" });
+  const second = await proxiesDb.upsertProxy({ ...base, username: "user-2", password: "pass-2" });
+  const third = await proxiesDb.upsertProxy({ ...base, username: "user-3", password: "pass-3" });
+
+  assert.equal(first.action, "created");
+  assert.equal(second.action, "created");
+  assert.equal(third.action, "created");
+  assert.notEqual(first.proxy?.id, second.proxy?.id);
+  assert.notEqual(second.proxy?.id, third.proxy?.id);
+
+  const listed = await proxiesDb.listProxies({ includeSecrets: true });
+  assert.equal(listed.length, 3);
+  assert.deepEqual(
+    listed.map((p) => p.username).sort(),
+    ["user-1", "user-2", "user-3"]
+  );
+});
+
+test("upsertProxy still updates when the full credential tuple matches (#7594)", async () => {
+  const payload = {
+    name: "Gateway",
+    type: "http" as const,
+    host: "gw.proxy.local",
+    port: 8080,
+    username: "user-1",
+    password: "pass-1",
+  };
+
+  const created = await proxiesDb.upsertProxy(payload);
+  const again = await proxiesDb.upsertProxy({ ...payload, name: "Gateway Renamed" });
+
+  assert.equal(created.action, "created");
+  assert.equal(again.action, "updated");
+  assert.equal(again.proxy?.id, created.proxy?.id);
+
+  const listed = await proxiesDb.listProxies({ includeSecrets: true });
+  assert.equal(listed.length, 1);
+  assert.equal(listed[0].name, "Gateway Renamed");
+});
+
+test("upsertProxy treats auth-less proxies on same host:port as the same entry (#7594)", async () => {
+  const base = { name: "Open Gateway", type: "http" as const, host: "open.proxy.local", port: 3128 };
+
+  const first = await proxiesDb.upsertProxy(base);
+  const second = await proxiesDb.upsertProxy({ ...base, name: "Open Gateway 2" });
+
+  assert.equal(first.action, "created");
+  assert.equal(second.action, "updated");
+  assert.equal(first.proxy?.id, second.proxy?.id);
+
+  const listed = await proxiesDb.listProxies({ includeSecrets: true });
+  assert.equal(listed.length, 1);
+});
diff --git a/tests/unit/responses-input-sanitizer-name.test.ts b/tests/unit/responses-input-sanitizer-name.test.ts
index 6ab5cb20069..067e233b593 100644
--- a/tests/unit/responses-input-sanitizer-name.test.ts
+++ b/tests/unit/responses-input-sanitizer-name.test.ts
@@ -78,7 +78,9 @@ test("normalizes user image_url content parts to input_image", () => {
     {
       type: "message",
       role: "user",
-      content: [{ type: "image_url", image_url: { url: "https://example.com/u.png", detail: "high" } }],
+      content: [
+        { type: "image_url", image_url: { url: "https://example.com/u.png", detail: "high" } },
+      ],
     },
   ];
   const result = sanitizeResponsesInputItems(items) as Array<Record<string, unknown>>;
@@ -111,7 +113,7 @@ test("normalizes assistant image content parts to output_text", () => {
   });
 });
 
-test("normalizes replayed Responses output image parts to output_text", () => {
+test("normalizes replayed Responses output image parts to input_image", () => {
   const items = [
     {
       type: "message",
@@ -123,9 +125,57 @@ test("normalizes replayed Responses output image parts to output_text", () => {
   assert.deepEqual(result[0], {
     type: "message",
     role: "assistant",
-    output: [{ type: "output_text", text: "[Image: https://example.com/output.png]" }],
+    output: [{ type: "input_image", image_url: "https://example.com/output.png" }],
+  });
+  assert.equal(JSON.stringify(result).includes('"type":"image_url"'), false);
+});
+
+test("normalizes nested output_text parts to input_text", () => {
+  const items = [
+    {
+      type: "function_call_output",
+      call_id: "call_2",
+      output: [
+        {
+          type: "output_text",
+          text: "tool result",
+          annotations: [],
+          logprobs: [],
+          obfuscation: "opaque",
+        },
+      ],
+    },
+  ];
+  const result = sanitizeResponsesInputItems(items) as Array<Record<string, unknown>>;
+  assert.deepEqual(result[0], {
+    type: "function_call_output",
+    call_id: "call_2",
+    output: [{ type: "input_text", text: "tool result" }],
+  });
+});
+
+test("normalizes nested refusal parts to input_text", () => {
+  const items = [
+    {
+      type: "function_call_output",
+      call_id: "call_3",
+      output: [
+        {
+          type: "refusal",
+          refusal: "I can't help with that.",
+          annotations: [],
+          logprobs: [],
+          obfuscation: "opaque",
+        },
+      ],
+    },
+  ];
+  const result = sanitizeResponsesInputItems(items) as Array<Record<string, unknown>>;
+  assert.deepEqual(result[0], {
+    type: "function_call_output",
+    call_id: "call_3",
+    output: [{ type: "input_text", text: "I can't help with that." }],
   });
-  assert.equal(JSON.stringify(result).includes('"image_url"'), false);
 });
 
 test("normalizes function_call_output image output parts to input_image", () => {
diff --git a/tests/unit/route-guard-private-lan.test.ts b/tests/unit/route-guard-private-lan.test.ts
index ba9a9f3036a..a19bed6d5fb 100644
--- a/tests/unit/route-guard-private-lan.test.ts
+++ b/tests/unit/route-guard-private-lan.test.ts
@@ -85,6 +85,11 @@ test("services + traffic-inspector remain LOCAL_ONLY paths", () => {
   assert.equal(isLocalOnlyPath("/api/tools/traffic-inspector/sessions"), true);
 });
 
+test("issue-agent routes are LOCAL_ONLY by default", () => {
+  assert.equal(isLocalOnlyPath("/api/issue-agent/runs"), true);
+  assert.equal(isLocalOnlyPath("/api/issue-agent/runs/recorded-triage"), true);
+});
+
 test("management policy must NOT derive locality from the spoofable Host header", () => {
   const src = readFileSync(
     join(import.meta.dirname, "../../src/server/authz/policies/management.ts"),
diff --git a/tests/unit/router-eval-check.test.ts b/tests/unit/router-eval-check.test.ts
new file mode 100644
index 00000000000..51ae202d03a
--- /dev/null
+++ b/tests/unit/router-eval-check.test.ts
@@ -0,0 +1,484 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { spawnSync } from "node:child_process";
+
+const checkScript = "scripts/check/check-router-eval-regression.ts";
+
+function writePatch(
+  file: string,
+  configId: string,
+  aiq: number,
+  costUsd: number,
+  latencyMs: number
+) {
+  writeFileSync(
+    file,
+    `${JSON.stringify(
+      {
+        schemaVersion: 1,
+        kind: "router-config-patch",
+        generatedAt: "2026-07-03T00:00:00.000Z",
+        applyPolicy: "manual-review",
+        source: {
+          objective: "balanced",
+          runId: `${configId}-run`,
+          artifactPath: `/tmp/${configId}/router-eval.json`,
+        },
+        operations: [
+          {
+            op: "recommend-router-config",
+            path: "/router/recommendedConfigId",
+            value: configId,
+            evidence: {
+              aiq,
+              avgCostUsd: costUsd,
+              avgLatencyMs: latencyMs,
+              regressions: 0,
+            },
+            rationale: `${configId} wins`,
+          },
+        ],
+      },
+      null,
+      2
+    )}\n`
+  );
+}
+
+test("router eval check writes artifacts and passes non-regressing corpora", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-check-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const candidate = join(dir, "candidate.ndjson");
+  const markdown = join(dir, "router-eval.md");
+  const json = join(dir, "router-eval.json");
+
+  writeFileSync(
+    baseline,
+    JSON.stringify({
+      sampleId: "b1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 140,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writeFileSync(
+    candidate,
+    JSON.stringify({
+      sampleId: "c1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 130,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      checkScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--output",
+      markdown,
+      "--json-output",
+      json,
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    assert.ok(readFileSync(markdown, "utf8").includes("Router Eval Comparison"));
+    const artifact = JSON.parse(readFileSync(json, "utf8")) as Record<string, unknown>;
+    assert.equal(artifact.kind, "router-eval-comparison");
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router eval check can include patch compare as a retained gate", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-patch-gate-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const candidate = join(dir, "candidate.ndjson");
+  const baselinePatch = join(dir, "baseline.patch.json");
+  const candidatePatch = join(dir, "candidate.patch.json");
+  const artifactDir = join(dir, "artifacts");
+  const runId = "patch-gate-001";
+
+  writeFileSync(
+    baseline,
+    JSON.stringify({
+      sampleId: "b1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 140,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writeFileSync(
+    candidate,
+    JSON.stringify({
+      sampleId: "c1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 130,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writePatch(baselinePatch, "balanced-v1", 92, 0.005, 150);
+  writePatch(candidatePatch, "balanced-v2", 94, 0.004, 120);
+
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      checkScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--baseline-patch",
+      baselinePatch,
+      "--candidate-patch",
+      candidatePatch,
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      runId,
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    assert.ok(
+      readFileSync(join(artifactDir, runId, "patch-comparison.md"), "utf8").includes(
+        "Router Config Patch Comparison"
+      )
+    );
+    assert.ok(
+      readFileSync(join(artifactDir, runId, "patch-comparison.json"), "utf8").includes(
+        "router-config-patch-comparison"
+      )
+    );
+    assert.ok(
+      readFileSync(join(artifactDir, runId, "inputs", "baseline.patch.json"), "utf8").includes(
+        "router-config-patch"
+      )
+    );
+    assert.ok(
+      readFileSync(join(artifactDir, runId, "inputs", "candidate.patch.json"), "utf8").includes(
+        "router-config-patch"
+      )
+    );
+    const manifest = JSON.parse(
+      readFileSync(join(artifactDir, runId, "manifest.json"), "utf8")
+    ) as {
+      inputs?: { baselinePatch?: string; candidatePatch?: string };
+      outputs?: { patchMarkdown?: string; patchJson?: string };
+      thresholds?: { patch?: { maxLatencyIncrease?: number } };
+    };
+    assert.equal(manifest.inputs?.baselinePatch, "inputs/baseline.patch.json");
+    assert.equal(manifest.inputs?.candidatePatch, "inputs/candidate.patch.json");
+    assert.equal(manifest.outputs?.patchMarkdown, "patch-comparison.md");
+    assert.equal(manifest.outputs?.patchJson, "patch-comparison.json");
+    assert.equal(manifest.thresholds?.patch?.maxLatencyIncrease, 0.05);
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router eval check fails when patch gate regresses beyond thresholds", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-patch-gate-fail-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const candidate = join(dir, "candidate.ndjson");
+  const baselinePatch = join(dir, "baseline.patch.json");
+  const candidatePatch = join(dir, "candidate.patch.json");
+  const artifactDir = join(dir, "artifacts");
+  const runId = "patch-gate-fail-001";
+
+  writeFileSync(
+    baseline,
+    JSON.stringify({
+      sampleId: "b1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 140,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writeFileSync(
+    candidate,
+    JSON.stringify({
+      sampleId: "c1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 130,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writePatch(baselinePatch, "balanced-v1", 95, 0.004, 120);
+  writePatch(candidatePatch, "balanced-v2", 90, 0.008, 200);
+
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      checkScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--baseline-patch",
+      baselinePatch,
+      "--candidate-patch",
+      candidatePatch,
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      runId,
+      "--max-patch-aiq-drop",
+      "1",
+      "--max-patch-cost-increase",
+      "0.1",
+      "--max-patch-latency-increase",
+      "0.1",
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 1);
+    assert.match(result.stderr ?? "", /patch gate failed/);
+    const comparison = JSON.parse(
+      readFileSync(join(artifactDir, runId, "patch-comparison.json"), "utf8")
+    ) as {
+      result?: { passed?: boolean; status?: number };
+      regressions?: string[];
+    };
+    assert.equal(comparison.result?.passed, false);
+    assert.equal(comparison.result?.status, 1);
+    assert.ok((comparison.regressions?.length ?? 0) >= 2);
+    const manifest = JSON.parse(
+      readFileSync(join(artifactDir, runId, "manifest.json"), "utf8")
+    ) as {
+      result?: { status?: number };
+    };
+    assert.equal(manifest.result?.status, 1);
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router eval check rejects unpaired patch inputs", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-patch-unpaired-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const candidate = join(dir, "candidate.ndjson");
+  const baselinePatch = join(dir, "baseline.patch.json");
+  const candidatePatch = join(dir, "candidate.patch.json");
+  const artifactDir = join(dir, "artifacts");
+  const runId = "unpaired-001";
+
+  writeFileSync(
+    baseline,
+    JSON.stringify({
+      sampleId: "b1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 140,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writeFileSync(
+    candidate,
+    JSON.stringify({
+      sampleId: "c1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 130,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writePatch(baselinePatch, "balanced-v1", 95, 0.004, 120);
+  writePatch(candidatePatch, "balanced-v2", 96, 0.003, 110);
+
+  const baselineOnly = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      checkScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--baseline-patch",
+      baselinePatch,
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      runId,
+    ],
+    { encoding: "utf8" }
+  );
+  const candidateOnly = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      checkScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--candidate-patch",
+      candidatePatch,
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(baselineOnly.status, 2);
+    assert.match(baselineOnly.stderr ?? "", /baseline-patch and --candidate-patch/);
+    const manifest = JSON.parse(
+      readFileSync(join(artifactDir, runId, "manifest.json"), "utf8")
+    ) as {
+      result?: { status?: number };
+      inputs?: { baselinePatch?: string; candidatePatch?: string };
+    };
+    assert.equal(manifest.result?.status, 2);
+    assert.equal(manifest.inputs?.baselinePatch, undefined);
+    assert.equal(manifest.inputs?.candidatePatch, undefined);
+    assert.equal(candidateOnly.status, 2);
+    assert.match(candidateOnly.stderr ?? "", /baseline-patch and --candidate-patch/);
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router eval check can retain artifacts for trend summaries", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-retained-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const candidate = join(dir, "candidate.ndjson");
+  const artifactDir = join(dir, "artifacts");
+  const runId = "run-001";
+
+  writeFileSync(
+    baseline,
+    JSON.stringify({
+      sampleId: "b1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 140,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writeFileSync(
+    candidate,
+    JSON.stringify({
+      sampleId: "c1",
+      configId: "priority",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 130,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+
+  const checkResult = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      checkScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      runId,
+    ],
+    { encoding: "utf8" }
+  );
+
+  const trendResult = spawnSync(
+    process.execPath,
+    ["--import", "tsx", "scripts/router-eval/trends.ts", "--artifact-dir", artifactDir],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(checkResult.status, 0);
+    assert.ok(
+      readFileSync(join(artifactDir, runId, "router-eval.md"), "utf8").includes(
+        "Router Eval Comparison"
+      )
+    );
+    assert.ok(
+      readFileSync(join(artifactDir, runId, "router-eval.json"), "utf8").includes(
+        "router-eval-comparison"
+      )
+    );
+    assert.ok(
+      readFileSync(join(artifactDir, runId, "inputs", "baseline.ndjson"), "utf8").includes(
+        "sampleId"
+      )
+    );
+    assert.ok(
+      readFileSync(join(artifactDir, runId, "inputs", "candidate.ndjson"), "utf8").includes(
+        "sampleId"
+      )
+    );
+    const manifest = JSON.parse(
+      readFileSync(join(artifactDir, runId, "manifest.json"), "utf8")
+    ) as Record<string, unknown>;
+    assert.equal(manifest.schemaVersion, 1);
+    assert.equal(manifest.kind, "router-eval-gate-run");
+    assert.equal(manifest.runId, runId);
+    assert.deepEqual(manifest.inputs, {
+      baseline: "inputs/baseline.ndjson",
+      candidate: "inputs/candidate.ndjson",
+    });
+    assert.deepEqual(manifest.outputs, {
+      markdown: "router-eval.md",
+      json: "router-eval.json",
+    });
+    assert.equal(trendResult.status, 0);
+    assert.ok((trendResult.stdout ?? "").includes("Router Eval Trends"));
+    assert.ok((trendResult.stdout ?? "").includes(runId));
+    assert.ok((trendResult.stdout ?? "").includes("| jsonl | all |"));
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
diff --git a/tests/unit/router-eval-cli.test.ts b/tests/unit/router-eval-cli.test.ts
new file mode 100644
index 00000000000..0eb2c1388e4
--- /dev/null
+++ b/tests/unit/router-eval-cli.test.ts
@@ -0,0 +1,270 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, readFileSync, writeFileSync, rmSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { spawnSync } from "node:child_process";
+import Database from "better-sqlite3";
+
+const scriptPath = "scripts/router-eval/index.ts";
+
+function runCli(args: string[]) {
+  return spawnSync(process.execPath, ["--import", "tsx", scriptPath, ...args], {
+    encoding: "utf8",
+  });
+}
+
+test("router-eval CLI prints a markdown report for JSONL input", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-cli-"));
+  const inputPath = join(dir, "input.ndjson");
+  writeFileSync(
+    inputPath,
+    [
+      JSON.stringify({
+        sampleId: "s1",
+        configId: "combo-a",
+        selectedModel: "gpt-4.1",
+        requestedModel: "gpt-4.1",
+        latencyMs: 120,
+        costUsd: 0.005,
+        success: true,
+      }),
+      JSON.stringify({
+        sampleId: "s2",
+        configId: "combo-b",
+        selectedModel: "gpt-4o",
+        requestedModel: "gpt-4o",
+        latencyMs: 200,
+        costUsd: 0.003,
+        success: true,
+      }),
+    ].join("\n")
+  );
+
+  const result = runCli(["--input", inputPath]);
+  try {
+    assert.equal(result.status, 0);
+    assert.ok((result.stderr ?? "").length === 0);
+    assert.ok((result.stdout ?? "").includes("Frontier"));
+    assert.ok((result.stdout ?? "").includes("AIQ"));
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router-eval CLI exits non-zero when regression threshold is exceeded", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-cli-reg-"));
+  const baselinePath = join(dir, "baseline.ndjson");
+  const candidatePath = join(dir, "candidate.ndjson");
+
+  writeFileSync(
+    baselinePath,
+    [
+      JSON.stringify({
+        sampleId: "b1",
+        configId: "combo-a",
+        selectedModel: "gpt-4.1",
+        requestedModel: "gpt-4.1",
+        latencyMs: 100,
+        costUsd: 0.005,
+        success: true,
+      }),
+    ].join("\n")
+  );
+
+  writeFileSync(
+    candidatePath,
+    [
+      JSON.stringify({
+        sampleId: "c1",
+        configId: "combo-a",
+        selectedModel: "gpt-4.1",
+        requestedModel: "gpt-4.1",
+        latencyMs: 400,
+        costUsd: 0.05,
+        success: true,
+      }),
+    ].join("\n")
+  );
+
+  const result = runCli([
+    "--input",
+    candidatePath,
+    "--baseline-input",
+    baselinePath,
+    "--max-aiq-drop",
+    "1",
+    "--max-cost-increase",
+    "0.2",
+    "--fail-on-regression",
+  ]);
+  try {
+    assert.equal(result.status, 1);
+    assert.ok((result.stdout ?? "").includes("Router Eval Comparison"));
+    assert.ok((result.stdout ?? "").includes("Regressions"));
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router-eval CLI writes machine-readable JSON artifacts", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-cli-json-"));
+  const inputPath = join(dir, "input.ndjson");
+  const outputPath = join(dir, "router-eval.json");
+  writeFileSync(
+    inputPath,
+    JSON.stringify({
+      sampleId: "s1",
+      configId: "combo-a",
+      selectedModel: "gpt-4.1",
+      requestedModel: "gpt-4.1",
+      latencyMs: 120,
+      costUsd: 0.005,
+      success: true,
+    })
+  );
+
+  const result = runCli(["--input", inputPath, "--json-output", outputPath]);
+  try {
+    assert.equal(result.status, 0);
+    const artifact = JSON.parse(readFileSync(outputPath, "utf8")) as Record<string, unknown>;
+    assert.equal(artifact.schemaVersion, 1);
+    assert.equal(artifact.kind, "router-eval-report");
+    assert.ok("report" in artifact);
+    assert.deepEqual((artifact.metadata as Record<string, unknown>)?.candidate, {
+      source: "jsonl",
+      path: inputPath,
+    });
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router-eval CLI reads usage_history DB source", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-cli-usage-"));
+  const dbPath = join(dir, "storage.sqlite");
+  const corpusPath = join(dir, "corpus.ndjson");
+  const artifactPath = join(dir, "artifact.json");
+  const db = new Database(dbPath);
+
+  db.exec(`
+    CREATE TABLE IF NOT EXISTS usage_history (
+      id INTEGER PRIMARY KEY AUTOINCREMENT,
+      provider TEXT,
+      model TEXT,
+      connection_id TEXT,
+      api_key_id TEXT,
+      api_key_name TEXT,
+      tokens_input INTEGER DEFAULT 0,
+      tokens_output INTEGER DEFAULT 0,
+      tokens_cache_read INTEGER DEFAULT 0,
+      tokens_cache_creation INTEGER DEFAULT 0,
+      tokens_reasoning INTEGER DEFAULT 0,
+      service_tier TEXT DEFAULT 'standard',
+      status TEXT,
+      success INTEGER DEFAULT 1,
+      latency_ms INTEGER DEFAULT 0,
+      ttft_ms INTEGER DEFAULT 0,
+      error_code TEXT,
+      combo_strategy TEXT,
+      endpoint TEXT,
+      timestamp TEXT NOT NULL
+    )
+  `);
+
+  db.prepare(
+    `
+    INSERT INTO usage_history
+      (provider, model, tokens_input, tokens_output, service_tier, success, latency_ms, status, error_code, combo_strategy, timestamp)
+    VALUES
+      ('openrouter', 'gpt-4.1', 120, 80, 'standard', 1, 150, '200', NULL, 'priority', '2026-01-01T00:00:00.000Z')
+  `
+  ).run();
+  db.close();
+
+  try {
+    const result = runCli([
+      "--db",
+      dbPath,
+      "--db-source",
+      "usage-history",
+      "--export-corpus",
+      corpusPath,
+      "--json-output",
+      artifactPath,
+    ]);
+    assert.equal(result.status, 0);
+    assert.ok((result.stderr ?? "").length === 0);
+    assert.ok((result.stdout ?? "").includes("Router Eval Report"));
+    assert.ok((result.stdout ?? "").includes("priority"));
+    const corpus = readFileSync(corpusPath, "utf8")
+      .trim()
+      .split("\n")
+      .map((line) => JSON.parse(line) as Record<string, unknown>);
+    assert.equal(corpus.length, 1);
+    assert.equal(corpus[0].configId, "priority");
+    assert.equal(corpus[0].selectedModel, "gpt-4.1");
+    assert.equal(corpus[0].latencyMs, 150);
+    assert.deepEqual(corpus[0].metadata, {
+      provider: "openrouter",
+      serviceTier: "standard",
+      errorCode: null,
+    });
+    const artifact = JSON.parse(readFileSync(artifactPath, "utf8")) as Record<string, unknown>;
+    assert.deepEqual((artifact.metadata as Record<string, unknown>)?.candidate, {
+      source: "sqlite",
+      path: dbPath,
+      dbSource: "usage-history",
+    });
+    assert.equal(
+      typeof ((artifact.metadata as Record<string, unknown>)?.outputs as Record<string, unknown>)
+        ?.corpus,
+      "string"
+    );
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router-eval CLI defaults --db to call_logs when available", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-cli-default-"));
+  const dbPath = join(dir, "storage.sqlite");
+  const db = new Database(dbPath);
+
+  db.exec(`
+    CREATE TABLE IF NOT EXISTS call_logs (
+      id TEXT PRIMARY KEY,
+      duration INTEGER,
+      tokens_in INTEGER,
+      tokens_out INTEGER,
+      status INTEGER,
+      combo_name TEXT,
+      requested_model TEXT,
+      model TEXT,
+      provider TEXT,
+      error_summary TEXT,
+      timestamp TEXT NOT NULL,
+      correlation_id TEXT
+    )
+  `);
+
+  db.prepare(
+    `
+    INSERT INTO call_logs
+      (id, provider, model, requested_model, tokens_in, tokens_out, status, combo_name, timestamp)
+    VALUES
+      ('c1', 'openrouter', 'gpt-4.1', 'gpt-4.1', 120, 80, 200, 'priority', '2026-01-01T00:00:00.000Z')
+  `
+  ).run();
+  db.close();
+
+  try {
+    const result = runCli(["--db", dbPath]);
+    assert.equal(result.status, 0);
+    assert.equal((result.stderr ?? "").length, 0);
+    assert.ok((result.stdout ?? "").includes("Router Eval Report"));
+    assert.ok((result.stdout ?? "").includes("priority"));
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
diff --git a/tests/unit/router-eval-compare.test.ts b/tests/unit/router-eval-compare.test.ts
new file mode 100644
index 00000000000..5eef0c2edb3
--- /dev/null
+++ b/tests/unit/router-eval-compare.test.ts
@@ -0,0 +1,83 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { spawnSync } from "node:child_process";
+
+const compareScript = "scripts/router-eval/compare.ts";
+
+test("router eval compare retains named comparison artifacts", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-compare-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const candidate = join(dir, "candidate.ndjson");
+  const artifactDir = join(dir, "artifacts");
+  const runId = "policy-a-vs-policy-b";
+
+  writeFileSync(
+    baseline,
+    JSON.stringify({
+      sampleId: "b1",
+      configId: "policy-a",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 140,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+  writeFileSync(
+    candidate,
+    JSON.stringify({
+      sampleId: "c1",
+      configId: "policy-b",
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs: 130,
+      costUsd: 0.004,
+      success: true,
+    })
+  );
+
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      compareScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--baseline-name",
+      "policy-a",
+      "--candidate-name",
+      "policy-b",
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      runId,
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    assert.ok((result.stdout ?? "").includes("Router Eval Comparison"));
+    const runDir = join(artifactDir, runId);
+    assert.ok(
+      readFileSync(join(runDir, "router-eval.md"), "utf8").includes("Router Eval Comparison")
+    );
+    assert.ok(
+      readFileSync(join(runDir, "router-eval.json"), "utf8").includes("router-eval-comparison")
+    );
+    const comparison = JSON.parse(readFileSync(join(runDir, "comparison.json"), "utf8")) as Record<
+      string,
+      unknown
+    >;
+    assert.equal(comparison.baselineName, "policy-a");
+    assert.equal(comparison.candidateName, "policy-b");
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
diff --git a/tests/unit/router-eval-e2e-chain.test.ts b/tests/unit/router-eval-e2e-chain.test.ts
new file mode 100644
index 00000000000..aea92f6df0c
--- /dev/null
+++ b/tests/unit/router-eval-e2e-chain.test.ts
@@ -0,0 +1,135 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { spawnSync } from "node:child_process";
+
+const searchScript = "scripts/router-eval/search.ts";
+const checkScript = "scripts/check/check-router-eval-regression.ts";
+
+function writeObservation(file: string, configId: string, latencyMs: number, costUsd: number) {
+  writeFileSync(
+    file,
+    `${JSON.stringify({
+      sampleId: `${configId}-sample`,
+      configId,
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs,
+      costUsd,
+      success: true,
+    })}\n`
+  );
+}
+
+function runSearch(
+  baseline: string,
+  candidateName: string,
+  candidate: string,
+  artifactDir: string,
+  runId: string
+): string {
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      searchScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      `${candidateName}=${candidate}`,
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      runId,
+      "--objective",
+      "quality",
+    ],
+    { encoding: "utf8" }
+  );
+  assert.equal(result.status, 0);
+  const patch = join(artifactDir, runId, "router-config.patch.json");
+  assert.ok(readFileSync(patch, "utf8").includes("router-config-patch"));
+  return patch;
+}
+
+test("router eval retained chain runs search patches through the check wrapper gate", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-e2e-chain-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const oldCandidate = join(dir, "old.ndjson");
+  const newCandidate = join(dir, "new.ndjson");
+  const searchArtifacts = join(dir, "search-artifacts");
+  const gateArtifacts = join(dir, "gate-artifacts");
+  const runId = "chain-001";
+
+  writeObservation(baseline, "baseline", 150, 0.004);
+  writeObservation(oldCandidate, "old-router", 130, 0.004);
+  writeObservation(newCandidate, "new-router", 100, 0.003);
+
+  try {
+    const baselinePatch = runSearch(
+      baseline,
+      "old-router",
+      oldCandidate,
+      searchArtifacts,
+      "old-search"
+    );
+    const candidatePatch = runSearch(
+      baseline,
+      "new-router",
+      newCandidate,
+      searchArtifacts,
+      "new-search"
+    );
+    const result = spawnSync(
+      process.execPath,
+      [
+        "--import",
+        "tsx",
+        checkScript,
+        "--baseline",
+        baseline,
+        "--candidate",
+        newCandidate,
+        "--baseline-patch",
+        baselinePatch,
+        "--candidate-patch",
+        candidatePatch,
+        "--artifact-dir",
+        gateArtifacts,
+        "--run-id",
+        runId,
+      ],
+      { encoding: "utf8" }
+    );
+
+    assert.equal(result.status, 0);
+    const runDir = join(gateArtifacts, runId);
+    assert.ok(
+      readFileSync(join(runDir, "router-eval.json"), "utf8").includes("router-eval-comparison")
+    );
+    assert.ok(
+      readFileSync(join(runDir, "patch-comparison.json"), "utf8").includes(
+        "router-config-patch-comparison"
+      )
+    );
+    assert.ok(
+      readFileSync(join(runDir, "inputs", "baseline.patch.json"), "utf8").includes("old-router")
+    );
+    assert.ok(
+      readFileSync(join(runDir, "inputs", "candidate.patch.json"), "utf8").includes("new-router")
+    );
+    const manifest = JSON.parse(readFileSync(join(runDir, "manifest.json"), "utf8")) as {
+      kind?: string;
+      result?: { status?: number };
+      outputs?: { patchJson?: string };
+    };
+    assert.equal(manifest.kind, "router-eval-gate-run");
+    assert.equal(manifest.result?.status, 0);
+    assert.equal(manifest.outputs?.patchJson, "patch-comparison.json");
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
diff --git a/tests/unit/router-eval-patch-compare.test.ts b/tests/unit/router-eval-patch-compare.test.ts
new file mode 100644
index 00000000000..87db7d65ac2
--- /dev/null
+++ b/tests/unit/router-eval-patch-compare.test.ts
@@ -0,0 +1,312 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { spawnSync } from "node:child_process";
+
+const compareScript = "scripts/router-eval/patch-compare.ts";
+
+function writePatch(
+  file: string,
+  configId: string,
+  aiq: number,
+  costUsd: number,
+  latencyMs: number
+) {
+  writeFileSync(
+    file,
+    `${JSON.stringify(
+      {
+        schemaVersion: 1,
+        kind: "router-config-patch",
+        generatedAt: "2026-07-03T00:00:00.000Z",
+        applyPolicy: "manual-review",
+        source: {
+          objective: "balanced",
+          runId: `${configId}-run`,
+          artifactPath: `/tmp/${configId}/router-eval.json`,
+        },
+        operations: [
+          {
+            op: "recommend-router-config",
+            path: "/router/recommendedConfigId",
+            value: configId,
+            evidence: {
+              aiq,
+              avgCostUsd: costUsd,
+              avgLatencyMs: latencyMs,
+              regressions: 0,
+            },
+            rationale: `${configId} wins`,
+          },
+        ],
+      },
+      null,
+      2
+    )}\n`
+  );
+}
+
+test("router config patch compare reports recommendation and metric deltas", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-patch-compare-"));
+  const baseline = join(dir, "baseline.patch.json");
+  const candidate = join(dir, "candidate.patch.json");
+  const jsonOutput = join(dir, "comparison.json");
+  const artifactDir = join(dir, "artifacts");
+
+  writePatch(baseline, "balanced-v1", 91, 0.006, 180);
+  writePatch(candidate, "balanced-v2", 94, 0.004, 120);
+
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      compareScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--baseline-name",
+      "current",
+      "--candidate-name",
+      "proposal",
+      "--json-output",
+      jsonOutput,
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      "compare-001",
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    assert.ok((result.stdout ?? "").includes("Router Config Patch Comparison"));
+    assert.ok((result.stdout ?? "").includes("Changed recommendation: yes"));
+    const comparison = JSON.parse(readFileSync(jsonOutput, "utf8")) as {
+      kind?: string;
+      changedRecommendation?: boolean;
+      delta?: { aiq?: number; avgCostUsd?: number; avgLatencyMs?: number };
+      candidate?: { recommendedConfigId?: string };
+    };
+    assert.equal(comparison.kind, "router-config-patch-comparison");
+    assert.equal(comparison.changedRecommendation, true);
+    assert.equal(comparison.result?.passed, true);
+    assert.equal(comparison.candidate?.recommendedConfigId, "balanced-v2");
+    assert.equal(comparison.delta?.aiq, 3);
+    assert.equal(comparison.delta?.avgCostUsd, -0.002);
+    assert.equal(comparison.delta?.avgLatencyMs, -60);
+    assert.ok(
+      readFileSync(join(artifactDir, "compare-001", "patch-comparison.md"), "utf8").includes(
+        "balanced-v2"
+      )
+    );
+    assert.ok(
+      readFileSync(join(artifactDir, "compare-001", "patch-comparison.json"), "utf8").includes(
+        "router-config-patch-comparison"
+      )
+    );
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router config patch compare only fails threshold regressions when requested", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-patch-compare-gate-"));
+  const baseline = join(dir, "baseline.patch.json");
+  const candidate = join(dir, "candidate.patch.json");
+  const jsonOutput = join(dir, "comparison.json");
+
+  writePatch(baseline, "balanced-v1", 95, 0.004, 120);
+  writePatch(candidate, "balanced-v2", 90, 0.008, 200);
+
+  const warning = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      compareScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--max-aiq-drop",
+      "1",
+      "--max-cost-increase",
+      "0.1",
+      "--max-latency-increase",
+      "0.1",
+      "--json-output",
+      jsonOutput,
+    ],
+    { encoding: "utf8" }
+  );
+  const failing = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      compareScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--max-aiq-drop",
+      "1",
+      "--max-cost-increase",
+      "0.1",
+      "--max-latency-increase",
+      "0.1",
+      "--fail-on-regression",
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(warning.status, 0);
+    assert.ok((warning.stdout ?? "").includes("Passed: no"));
+    const comparison = JSON.parse(readFileSync(jsonOutput, "utf8")) as {
+      regressions?: string[];
+      result?: { passed?: boolean; status?: number };
+    };
+    assert.equal(comparison.result?.passed, false);
+    assert.equal(comparison.result?.status, 0);
+    assert.ok((comparison.regressions?.length ?? 0) >= 2);
+    assert.equal(failing.status, 1);
+    assert.ok((failing.stdout ?? "").includes("Passed: no"));
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router config patch compare reports unchanged recommendations without failing", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-patch-compare-no-change-"));
+  const baseline = join(dir, "baseline.patch.json");
+  const candidate = join(dir, "candidate.patch.json");
+  const jsonOutput = join(dir, "comparison.json");
+
+  writePatch(baseline, "balanced-v1", 94, 0.004, 120);
+  writePatch(candidate, "balanced-v1", 95, 0.003, 110);
+
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      compareScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      candidate,
+      "--json-output",
+      jsonOutput,
+      "--fail-on-regression",
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    assert.ok((result.stdout ?? "").includes("Changed recommendation: no"));
+    const comparison = JSON.parse(readFileSync(jsonOutput, "utf8")) as {
+      changedRecommendation?: boolean;
+      delta?: { aiq?: number; avgCostUsd?: number; avgLatencyMs?: number };
+      result?: { passed?: boolean; status?: number };
+    };
+    assert.equal(comparison.changedRecommendation, false);
+    assert.equal(comparison.delta?.aiq, 1);
+    assert.equal(comparison.delta?.avgCostUsd, -0.001);
+    assert.equal(comparison.delta?.avgLatencyMs, -10);
+    assert.equal(comparison.result?.passed, true);
+    assert.equal(comparison.result?.status, 0);
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router config patch compare rejects invalid patch inputs", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-patch-compare-invalid-"));
+  const baseline = join(dir, "baseline.patch.json");
+  const candidate = join(dir, "candidate.patch.json");
+
+  writePatch(baseline, "balanced-v1", 94, 0.004, 120);
+  writeFileSync(
+    candidate,
+    `${JSON.stringify(
+      {
+        schemaVersion: 1,
+        kind: "router-config-suggestion",
+      },
+      null,
+      2
+    )}\n`
+  );
+
+  const result = spawnSync(
+    process.execPath,
+    ["--import", "tsx", compareScript, "--baseline", baseline, "--candidate", candidate],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 2);
+    assert.match(result.stderr ?? "", /invalid patch kind/);
+    assert.match(result.stderr ?? "", /router-config-suggestion/);
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router config patch compare rejects malformed JSON and invalid evidence", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-patch-compare-malformed-"));
+  const baseline = join(dir, "baseline.patch.json");
+  const malformed = join(dir, "malformed.patch.json");
+  const invalidEvidence = join(dir, "invalid-evidence.patch.json");
+
+  writePatch(baseline, "balanced-v1", 94, 0.004, 120);
+  writeFileSync(malformed, "{not-json");
+  writeFileSync(
+    invalidEvidence,
+    `${JSON.stringify({
+      schemaVersion: 1,
+      kind: "router-config-patch",
+      applyPolicy: "manual-review",
+      operations: [
+        {
+          op: "recommend-router-config",
+          path: "/router/recommendedConfigId",
+          value: "balanced-v2",
+          evidence: {
+            aiq: "94",
+            avgCostUsd: 0.004,
+            avgLatencyMs: 120,
+            regressions: 0,
+          },
+        },
+      ],
+    })}\n`
+  );
+
+  const malformedResult = spawnSync(
+    process.execPath,
+    ["--import", "tsx", compareScript, "--baseline", baseline, "--candidate", malformed],
+    { encoding: "utf8" }
+  );
+  const invalidEvidenceResult = spawnSync(
+    process.execPath,
+    ["--import", "tsx", compareScript, "--baseline", baseline, "--candidate", invalidEvidence],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(malformedResult.status, 2);
+    assert.match(malformedResult.stderr ?? "", /invalid JSON/);
+    assert.equal(invalidEvidenceResult.status, 2);
+    assert.match(invalidEvidenceResult.stderr ?? "", /invalid numeric evidence field aiq/);
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
diff --git a/tests/unit/router-eval-search.test.ts b/tests/unit/router-eval-search.test.ts
new file mode 100644
index 00000000000..7e6b40507e0
--- /dev/null
+++ b/tests/unit/router-eval-search.test.ts
@@ -0,0 +1,258 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { spawnSync } from "node:child_process";
+
+const searchScript = "scripts/router-eval/search.ts";
+
+function writeCorpus(path: string, configId: string, latencyMs: number, costUsd: number) {
+  writeFileSync(
+    path,
+    JSON.stringify({
+      sampleId: configId,
+      configId,
+      selectedModel: "gpt-4.1",
+      expectedModel: "gpt-4.1",
+      latencyMs,
+      costUsd,
+      success: true,
+    })
+  );
+}
+
+function writeCorpusRows(
+  path: string,
+  rows: Array<{ configId: string; latencyMs: number; costUsd: number }>
+) {
+  writeFileSync(
+    path,
+    rows
+      .map((row) =>
+        JSON.stringify({
+          sampleId: `${row.configId}-sample`,
+          configId: row.configId,
+          selectedModel: "gpt-4.1",
+          expectedModel: "gpt-4.1",
+          latencyMs: row.latencyMs,
+          costUsd: row.costUsd,
+          success: true,
+        })
+      )
+      .join("\n")
+  );
+}
+
+test("router eval search ranks candidates and writes retained summary artifacts", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-search-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const slow = join(dir, "slow.ndjson");
+  const fast = join(dir, "fast.ndjson");
+  const artifactDir = join(dir, "artifacts");
+  const runId = "search-001";
+
+  writeCorpus(baseline, "baseline", 150, 0.004);
+  writeCorpus(slow, "slow", 200, 0.006);
+  writeCorpus(fast, "fast", 100, 0.003);
+
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      searchScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      `slow=${slow}`,
+      "--candidate",
+      `fast=${fast}`,
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      runId,
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    assert.ok((result.stdout ?? "").includes("Router Eval Search"));
+    const searchDir = join(artifactDir, runId);
+    const markdown = readFileSync(join(searchDir, "search.md"), "utf8");
+    assert.ok(markdown.indexOf("| 1 | fast |") < markdown.indexOf("| 2 | slow |"));
+    assert.ok(markdown.includes("## Patch Operations"));
+    assert.ok(
+      markdown.includes("| recommend-router-config | /router/recommendedConfigId | fast |")
+    );
+    const summary = JSON.parse(readFileSync(join(searchDir, "search.json"), "utf8")) as {
+      kind?: string;
+      objective?: string;
+      recommendation?: { candidateName: string; rationale: string };
+      suggestion?: { recommendedConfigId: string; applyPolicy: string };
+      patch?: { kind: string; operations: Array<{ value: string }> };
+      results?: Array<{ candidateName: string }>;
+    };
+    assert.equal(summary.kind, "router-eval-search");
+    assert.equal(summary.objective, "balanced");
+    assert.equal(summary.results?.[0]?.candidateName, "fast");
+    assert.equal(summary.recommendation?.candidateName, "fast");
+    assert.match(summary.recommendation?.rationale ?? "", /best balanced rank/);
+    assert.equal(summary.suggestion?.recommendedConfigId, "fast");
+    assert.equal(summary.suggestion?.applyPolicy, "manual-review");
+    assert.equal(summary.patch?.kind, "router-config-patch");
+    assert.equal(summary.patch?.operations[0]?.value, "fast");
+    const recommendation = JSON.parse(
+      readFileSync(join(searchDir, "recommendation.json"), "utf8")
+    ) as {
+      candidateName?: string;
+      objective?: string;
+    };
+    assert.equal(recommendation.candidateName, "fast");
+    assert.equal(recommendation.objective, "balanced");
+    const suggestion = JSON.parse(readFileSync(join(searchDir, "suggestion.json"), "utf8")) as {
+      kind?: string;
+      recommendedConfigId?: string;
+      applyPolicy?: string;
+    };
+    assert.equal(suggestion.kind, "router-config-suggestion");
+    assert.equal(suggestion.recommendedConfigId, "fast");
+    assert.equal(suggestion.applyPolicy, "manual-review");
+    const patch = JSON.parse(readFileSync(join(searchDir, "router-config.patch.json"), "utf8")) as {
+      kind?: string;
+      applyPolicy?: string;
+      operations?: Array<{ op?: string; path?: string; value?: string }>;
+    };
+    assert.equal(patch.kind, "router-config-patch");
+    assert.equal(patch.applyPolicy, "manual-review");
+    assert.equal(patch.operations?.[0]?.op, "recommend-router-config");
+    assert.equal(patch.operations?.[0]?.path, "/router/recommendedConfigId");
+    assert.equal(patch.operations?.[0]?.value, "fast");
+    assert.ok(
+      readFileSync(join(searchDir, `${runId}-fast`, "router-eval.json"), "utf8").includes(
+        "router-eval-comparison"
+      )
+    );
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router eval search objective modes can choose different candidates", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-search-objective-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const cheap = join(dir, "cheap.ndjson");
+  const fast = join(dir, "fast.ndjson");
+  const artifactDir = join(dir, "artifacts");
+
+  writeCorpus(baseline, "baseline", 150, 0.004);
+  writeCorpus(cheap, "cheap", 900, 0.001);
+  writeCorpus(fast, "fast", 50, 0.009);
+
+  function runSearch(
+    objective: string,
+    runId: string
+  ): { recommendedConfigId?: string; objective?: string } {
+    const result = spawnSync(
+      process.execPath,
+      [
+        "--import",
+        "tsx",
+        searchScript,
+        "--baseline",
+        baseline,
+        "--candidate",
+        `cheap=${cheap}`,
+        "--candidate",
+        `fast=${fast}`,
+        "--artifact-dir",
+        artifactDir,
+        "--run-id",
+        runId,
+        "--objective",
+        objective,
+        "--max-aiq-drop",
+        "100",
+        "--max-cost-increase",
+        "100",
+      ],
+      { encoding: "utf8" }
+    );
+    assert.equal(result.status, 0);
+    return JSON.parse(readFileSync(join(artifactDir, runId, "suggestion.json"), "utf8")) as {
+      recommendedConfigId?: string;
+      objective?: string;
+    };
+  }
+
+  try {
+    const costSuggestion = runSearch("cost", "cost-001");
+    const latencySuggestion = runSearch("latency", "latency-001");
+    const qualitySuggestion = runSearch("quality", "quality-001");
+    assert.equal(costSuggestion.objective, "cost");
+    assert.equal(costSuggestion.recommendedConfigId, "cheap");
+    assert.equal(latencySuggestion.objective, "latency");
+    assert.equal(latencySuggestion.recommendedConfigId, "fast");
+    assert.equal(qualitySuggestion.objective, "quality");
+    assert.equal(qualitySuggestion.recommendedConfigId, "cheap");
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router eval search cost objective can select a non-AIQ-top config inside a candidate corpus", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-search-multi-config-"));
+  const baseline = join(dir, "baseline.ndjson");
+  const mixed = join(dir, "mixed.ndjson");
+  const artifactDir = join(dir, "artifacts");
+  const runId = "multi-config-001";
+
+  writeCorpus(baseline, "baseline", 150, 0.004);
+  writeCorpusRows(mixed, [
+    { configId: "quality-top", latencyMs: 100, costUsd: 0.002 },
+    { configId: "cost-top", latencyMs: 3_000, costUsd: 0.0001 },
+  ]);
+
+  const result = spawnSync(
+    process.execPath,
+    [
+      "--import",
+      "tsx",
+      searchScript,
+      "--baseline",
+      baseline,
+      "--candidate",
+      `mixed=${mixed}`,
+      "--artifact-dir",
+      artifactDir,
+      "--run-id",
+      runId,
+      "--objective",
+      "cost",
+      "--max-aiq-drop",
+      "100",
+      "--max-cost-increase",
+      "100",
+    ],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    const suggestion = JSON.parse(
+      readFileSync(join(artifactDir, runId, "suggestion.json"), "utf8")
+    ) as {
+      recommendedConfigId?: string;
+    };
+    const patch = JSON.parse(
+      readFileSync(join(artifactDir, runId, "router-config.patch.json"), "utf8")
+    ) as {
+      operations?: Array<{ value?: string }>;
+    };
+    assert.equal(suggestion.recommendedConfigId, "cost-top");
+    assert.equal(patch.operations?.[0]?.value, "cost-top");
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
diff --git a/tests/unit/router-eval-trends.test.ts b/tests/unit/router-eval-trends.test.ts
new file mode 100644
index 00000000000..989e594f135
--- /dev/null
+++ b/tests/unit/router-eval-trends.test.ts
@@ -0,0 +1,128 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdirSync, mkdtempSync, rmSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { spawnSync } from "node:child_process";
+
+const trendsScript = "scripts/router-eval/trends.ts";
+
+function artifact(generatedAt: string, configId: string, aiq: number) {
+  return {
+    schemaVersion: 1,
+    kind: "router-eval-report",
+    generatedAt,
+    report: {
+      evaluatedAt: generatedAt,
+      summary: {
+        totalSamples: 1,
+        validSamples: 1,
+        droppedSamples: 0,
+        uniqueConfigs: 1,
+      },
+      configurations: [],
+      frontier: [],
+      top: [
+        {
+          configId,
+          samples: 1,
+          successRate: 1,
+          avgLatencyMs: 100,
+          p50LatencyMs: 100,
+          p95LatencyMs: 100,
+          avgCostUsd: 0.004,
+          aiq,
+        },
+      ],
+    },
+  };
+}
+
+test("router eval trends reads retained and flat artifacts with limit", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-trends-"));
+  const runA = join(dir, "run-a");
+  const runB = join(dir, "run-b");
+  mkdirSync(runA);
+  mkdirSync(runB);
+  writeFileSync(
+    join(runA, "router-eval.json"),
+    JSON.stringify(artifact("2026-01-01T00:00:00.000Z", "old", 80))
+  );
+  writeFileSync(
+    join(runB, "router-eval.json"),
+    JSON.stringify(artifact("2026-01-02T00:00:00.000Z", "new", 90))
+  );
+  writeFileSync(
+    join(dir, "flat.json"),
+    JSON.stringify(artifact("2026-01-03T00:00:00.000Z", "flat", 95))
+  );
+  writeFileSync(join(dir, "bad.json"), "{not json");
+
+  const result = spawnSync(
+    process.execPath,
+    ["--import", "tsx", trendsScript, "--artifact-dir", dir, "--limit", "2"],
+    {
+      encoding: "utf8",
+    }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    assert.ok((result.stdout ?? "").includes("Router Eval Trends"));
+    assert.ok(!(result.stdout ?? "").includes("run-a"));
+    assert.ok((result.stdout ?? "").includes("run-b"));
+    assert.ok((result.stdout ?? "").includes("flat"));
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router eval trends can print dashboard summaries", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-dashboard-"));
+  const runA = join(dir, "run-a");
+  const runB = join(dir, "run-b");
+  mkdirSync(runA);
+  mkdirSync(runB);
+  writeFileSync(
+    join(runA, "router-eval.json"),
+    JSON.stringify(artifact("2026-01-01T00:00:00.000Z", "old", 80))
+  );
+  writeFileSync(
+    join(runB, "router-eval.json"),
+    JSON.stringify(artifact("2026-01-02T00:00:00.000Z", "new", 90))
+  );
+
+  const result = spawnSync(
+    process.execPath,
+    ["--import", "tsx", trendsScript, "--artifact-dir", dir, "--dashboard"],
+    {
+      encoding: "utf8",
+    }
+  );
+
+  try {
+    assert.equal(result.status, 0);
+    assert.ok((result.stdout ?? "").includes("Router Eval Dashboard"));
+    assert.ok((result.stdout ?? "").includes("Latest: run-b"));
+    assert.ok((result.stdout ?? "").includes("AIQ: 90.000 (+10.000)"));
+    assert.ok((result.stdout ?? "").includes("Rolling Averages"));
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
+
+test("router eval trends exits clearly for empty artifact dirs", () => {
+  const dir = mkdtempSync(join(tmpdir(), "router-eval-trends-empty-"));
+  const result = spawnSync(
+    process.execPath,
+    ["--import", "tsx", trendsScript, "--artifact-dir", dir],
+    { encoding: "utf8" }
+  );
+
+  try {
+    assert.equal(result.status, 2);
+    assert.ok((result.stderr ?? "").includes("No router-eval artifacts found"));
+  } finally {
+    rmSync(dir, { recursive: true, force: true });
+  }
+});
diff --git a/tests/unit/router-eval.test.ts b/tests/unit/router-eval.test.ts
new file mode 100644
index 00000000000..ba520a6c869
--- /dev/null
+++ b/tests/unit/router-eval.test.ts
@@ -0,0 +1,92 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import {
+  compareRouterEvalRuns,
+  createRouterEvalArtifact,
+  runRouterEval,
+  toRouterObservation,
+} from "@/lib/routerEval/index.ts";
+
+function obs(overrides: Record<string, unknown>) {
+  return toRouterObservation({
+    sampleId: "sample-1",
+    configId: "config",
+    selectedModel: "gpt-4.1",
+    expectedModel: "gpt-4.1",
+    latencyMs: 120,
+    costUsd: 0.05,
+    success: true,
+    timestamp: "2026-01-01T00:00:00.000Z",
+    ...overrides,
+  })!;
+}
+
+test("toRouterObservation normalizes token cost and booleans", () => {
+  const observation = toRouterObservation({
+    sampleId: "s-1",
+    configId: "combo-a",
+    selectedModel: "gpt-4.1",
+    requestedModel: "gpt-4.1",
+    durationMs: 240,
+    tokens: { input: 100, output: 50 },
+    status: 200,
+  });
+
+  assert.ok(observation);
+  assert.equal(observation!.sampleId, "s-1");
+  assert.equal(observation!.configId, "combo-a");
+  assert.equal(observation!.latencyMs, 240);
+  assert.equal(observation!.costUsd, 0.00015);
+  assert.equal(observation!.success, true);
+  assert.equal(observation!.expectedModel, "gpt-4.1");
+});
+
+test("runRouterEval computes summary and frontiers", () => {
+  const report = runRouterEval([
+    obs({ sampleId: "s1", configId: "a", latencyMs: 100, costUsd: 0.01, success: true }),
+    obs({ sampleId: "s2", configId: "a", latencyMs: 100, costUsd: 0.01, success: true }),
+    obs({ sampleId: "s3", configId: "b", latencyMs: 80, costUsd: 0.02, success: true }),
+    obs({ sampleId: "s4", configId: "b", latencyMs: 120, costUsd: 0.02, success: false }),
+  ]);
+
+  assert.equal(report.summary.totalSamples, 4);
+  assert.equal(report.summary.validSamples, 4);
+  assert.equal(report.summary.uniqueConfigs, 2);
+  assert.equal(report.configurations.length, 2);
+  assert.equal(report.top[0]?.configId, "a");
+  assert.ok(report.frontier.length >= 1);
+});
+
+test("compareRouterEvalRuns captures AIQ and cost regressions", () => {
+  const baseline = runRouterEval([
+    obs({ sampleId: "b1", configId: "a", latencyMs: 100, costUsd: 0.01, success: true }),
+    obs({ sampleId: "b2", configId: "a", latencyMs: 100, costUsd: 0.01, success: true }),
+  ]);
+  const candidate = runRouterEval([
+    obs({ sampleId: "c1", configId: "a", latencyMs: 300, costUsd: 0.03, success: true }),
+    obs({ sampleId: "c2", configId: "a", latencyMs: 300, costUsd: 0.03, success: true }),
+  ]);
+
+  const comparison = compareRouterEvalRuns(baseline, candidate, {
+    aiqDrop: 5,
+    relativeCostIncrease: 1.2,
+  });
+
+  assert.equal(comparison.regressions.length > 0, true);
+  assert.equal(comparison.delta.aiq <= 0, true);
+  assert.equal(comparison.delta.costUsd > 0, true);
+});
+
+test("createRouterEvalArtifact wraps reports with a stable schema", () => {
+  const report = runRouterEval([
+    obs({ sampleId: "s1", configId: "a", latencyMs: 100, costUsd: 0.01, success: true }),
+  ]);
+
+  const artifact = createRouterEvalArtifact(report);
+
+  assert.equal(artifact.schemaVersion, 1);
+  assert.equal(artifact.kind, "router-eval-report");
+  assert.equal(artifact.generatedAt, report.evaluatedAt);
+  assert.equal(artifact.report?.summary.totalSamples, 1);
+});
diff --git a/tests/unit/sseHeartbeat.test.ts b/tests/unit/sseHeartbeat.test.ts
new file mode 100644
index 00000000000..36d51b1e790
--- /dev/null
+++ b/tests/unit/sseHeartbeat.test.ts
@@ -0,0 +1,75 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import {
+  sseCommentsEnabled,
+  shapeForClientFormat,
+  createSseHeartbeatTransform,
+  HEARTBEAT_SHAPES,
+} from "../../open-sse/utils/sseHeartbeat.ts";
+
+function withEnv(value: string | undefined, fn: () => void) {
+  const prev = process.env.OMNIROUTE_SSE_COMMENTS;
+  try {
+    if (value === undefined) delete process.env.OMNIROUTE_SSE_COMMENTS;
+    else process.env.OMNIROUTE_SSE_COMMENTS = value;
+    fn();
+  } finally {
+    if (prev === undefined) delete process.env.OMNIROUTE_SSE_COMMENTS;
+    else process.env.OMNIROUTE_SSE_COMMENTS = prev;
+  }
+}
+
+test("sseCommentsEnabled defaults to true when the env var is unset", () => {
+  withEnv(undefined, () => assert.equal(sseCommentsEnabled(), true));
+});
+
+test("sseCommentsEnabled is false only when set to 'off' (case-insensitive)", () => {
+  withEnv("off", () => assert.equal(sseCommentsEnabled(), false));
+  withEnv("OFF", () => assert.equal(sseCommentsEnabled(), false));
+  withEnv("on", () => assert.equal(sseCommentsEnabled(), true));
+  withEnv("false", () => assert.equal(sseCommentsEnabled(), true));
+});
+
+test("shapeForClientFormat maps known client formats", () => {
+  assert.equal(shapeForClientFormat("claude"), HEARTBEAT_SHAPES.ANTHROPIC_PING);
+  assert.equal(shapeForClientFormat("openai"), HEARTBEAT_SHAPES.OPENAI_CHUNK);
+  assert.equal(shapeForClientFormat("openai-responses"), HEARTBEAT_SHAPES.OPENAI_RESPONSES_IN_PROGRESS);
+  assert.equal(shapeForClientFormat(undefined), HEARTBEAT_SHAPES.COMMENT);
+});
+
+test("createSseHeartbeatTransform suppresses COMMENT heartbeats when OMNIROUTE_SSE_COMMENTS=off", async () => {
+  const prev = process.env.OMNIROUTE_SSE_COMMENTS;
+  process.env.OMNIROUTE_SSE_COMMENTS = "off";
+  try {
+    const enc = new TextEncoder();
+    const dec = new TextDecoder();
+    const input = new ReadableStream<Uint8Array>({
+      start(c) {
+        c.enqueue(enc.encode("data: hello\n\n"));
+        c.close();
+      },
+    });
+    const reader = input
+      .pipeThrough(createSseHeartbeatTransform({ shape: HEARTBEAT_SHAPES.COMMENT, intervalMs: 20 }))
+      .pipeThrough(
+        new TransformStream<Uint8Array, string>({
+          transform(chunk, ctrl) {
+            ctrl.enqueue(dec.decode(chunk));
+          },
+        })
+      )
+      .getReader();
+    const chunks: string[] = [];
+    let res = await reader.read();
+    while (!res.done) {
+      chunks.push(res.value as string);
+      res = await reader.read();
+    }
+    const out = chunks.join("");
+    assert.ok(!out.includes(": keepalive"), "no comment heartbeat should be emitted when disabled");
+    assert.ok(out.includes("data: hello"), "original chunk passes through unchanged");
+  } finally {
+    if (prev === undefined) delete process.env.OMNIROUTE_SSE_COMMENTS;
+    else process.env.OMNIROUTE_SSE_COMMENTS = prev;
+  }
+});
diff --git a/tests/unit/stream-request-body-size-mark-7045.test.ts b/tests/unit/stream-request-body-size-mark-7045.test.ts
new file mode 100644
index 00000000000..b033be2085e
--- /dev/null
+++ b/tests/unit/stream-request-body-size-mark-7045.test.ts
@@ -0,0 +1,122 @@
+// PR #7045 (@oyi77, performance.mark/measure SSE instrumentation) regression coverage +
+// fix for a review-flagged leak: the "omni-request-body-size" mark is created on every
+// createSSEStream() call with a fixed name and, before this fix, was never cleared —
+// each call added another entry to Node's global performance timeline, unbounded over a
+// long-running server's lifetime. It must now be observable (via PerformanceObserver,
+// which still fires synchronously) yet cleared from the buffer immediately after.
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import { performance } from "node:perf_hooks";
+
+const TEST_DATA_DIR = fs.mkdtempSync(
+  path.join(os.tmpdir(), "omniroute-stream-body-size-mark-")
+);
+process.env.DATA_DIR = TEST_DATA_DIR;
+const core = await import("../../src/lib/db/core.ts");
+
+const { createSSEStream } = await import("../../open-sse/utils/stream.ts");
+const { FORMATS } = await import("../../open-sse/translator/formats.ts");
+
+const textEncoder = new TextEncoder();
+
+async function drainSSEStream(options) {
+  const source = new ReadableStream({
+    start(controller) {
+      controller.enqueue(
+        textEncoder.encode(
+          `data: ${JSON.stringify({
+            id: "chatcmpl_bodysize",
+            object: "chat.completion.chunk",
+            created: 1,
+            model: "gpt-4.1-mini",
+            choices: [{ index: 0, delta: { role: "assistant", content: "hi" } }],
+          })}\n\n`
+        )
+      );
+      controller.enqueue(
+        textEncoder.encode(
+          `data: ${JSON.stringify({
+            id: "chatcmpl_bodysize",
+            object: "chat.completion.chunk",
+            created: 1,
+            model: "gpt-4.1-mini",
+            choices: [{ index: 0, delta: {}, finish_reason: "stop" }],
+          })}\n\n`
+        )
+      );
+      controller.close();
+    },
+  });
+  return new Response(source.pipeThrough(createSSEStream(options))).text();
+}
+
+test.after(() => {
+  core.resetDbInstance();
+  if (fs.existsSync(TEST_DATA_DIR)) {
+    fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+  }
+});
+
+test("createSSEStream emits a request-body-size mark with the JSON byte length as detail", async (t) => {
+  const observedMarks = [];
+  const originalMark = performance.mark.bind(performance);
+  t.mock.method(performance, "mark", (name, options) => {
+    if (name === "omni-request-body-size") observedMarks.push(options?.detail);
+    return originalMark(name, options);
+  });
+
+  const body = { messages: [{ role: "user", content: "hello world" }] };
+  const expectedBytes = Buffer.byteLength(JSON.stringify(body), "utf8");
+
+  await drainSSEStream({
+    mode: "passthrough",
+    sourceFormat: FORMATS.OPENAI,
+    provider: "openai",
+    model: "gpt-4.1-mini",
+    body,
+  });
+
+  assert.equal(observedMarks.length, 1, "mark() should have been called exactly once");
+  assert.equal(observedMarks[0], expectedBytes, "mark detail should be the JSON byte length");
+});
+
+test("createSSEStream clears the request-body-size mark immediately (no unbounded growth)", async () => {
+  performance.clearMarks("omni-request-body-size");
+
+  const body = { messages: [{ role: "user", content: "hello" }] };
+  for (let i = 0; i < 5; i++) {
+    await drainSSEStream({
+      mode: "passthrough",
+      sourceFormat: FORMATS.OPENAI,
+      provider: "openai",
+      model: "gpt-4.1-mini",
+      body,
+    });
+  }
+
+  assert.equal(
+    performance.getEntriesByName("omni-request-body-size").length,
+    0,
+    "the mark must not accumulate across repeated calls"
+  );
+});
+
+test("createSSEStream skips the mark when body is absent (bodySize stays 0)", async () => {
+  performance.clearMarks("omni-request-body-size");
+
+  await drainSSEStream({
+    mode: "passthrough",
+    sourceFormat: FORMATS.OPENAI,
+    provider: "openai",
+    model: "gpt-4.1-mini",
+  });
+
+  assert.equal(
+    performance.getEntriesByName("omni-request-body-size").length,
+    0,
+    "no mark expected when there is no request body"
+  );
+});
diff --git a/tests/unit/telemetry-auto-cleanup-6848.test.ts b/tests/unit/telemetry-auto-cleanup-6848.test.ts
new file mode 100644
index 00000000000..3f214e022ea
--- /dev/null
+++ b/tests/unit/telemetry-auto-cleanup-6848.test.ts
@@ -0,0 +1,214 @@
+/**
+ * Issue #6848 — Auto-cleanup for telemetry tables that grow without bound.
+ *
+ * domain_cost_history, compression_cache_stats, xp_audit_log, and
+ * compression_run_telemetry had no retention cleanup, causing unbounded
+ * DB growth and OOM crashes on relays with heavy traffic.
+ *
+ * These tests call the REAL cleanup functions against a real SQLite adapter
+ * (seeded with test data). If the production DELETE logic breaks, these
+ * tests WILL fail.
+ *
+ * DATA_DIR isolation is self-contained (mkdtempSync below), not dependent on
+ * the test:unit harness's `--import ./tests/_setup/isolateDataDir.ts`. This
+ * file calls real DELETE-based cleanup functions against getDbInstance(),
+ * which resolves to the developer's real ~/.omniroute/storage.sqlite when
+ * DATA_DIR is unset — running this file directly (as documented under
+ * "Running Tests": `node --import tsx/esm --test tests/unit/<file>.test.ts`)
+ * would otherwise delete real rows from that database, not test rows. Do NOT
+ * remove the DATA_DIR override below.
+ */
+
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-6848-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+// Import the real functions — they use getDbInstance(), which resolves to the
+// isolated temp DB created above (set before this import so the module's
+// first getDbInstance() call already sees the isolated DATA_DIR).
+const {
+  cleanupDomainCostHistory,
+  cleanupCompressionCacheStats,
+  cleanupXpAuditLog,
+  cleanupCompressionRunTelemetry,
+} = await import("../../src/lib/db/cleanup.ts");
+
+const { getDbInstance, resetDbInstance } = await import("../../src/lib/db/core.ts");
+
+// Repo test rule: DB-touching tests must close the handle in test.after(),
+// or the native test runner can hang indefinitely on a dangling connection.
+test.after(() => {
+  resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+const DAY = 86_400; // seconds
+
+/** Ensure compression_run_telemetry table exists (created lazily in production). */
+function ensureTelemetryTable(): void {
+  const db = getDbInstance()!;
+  db.exec(`
+    CREATE TABLE IF NOT EXISTS compression_run_telemetry (
+      id INTEGER PRIMARY KEY AUTOINCREMENT,
+      timestamp INTEGER NOT NULL,
+      request_id TEXT,
+      model TEXT,
+      provider TEXT,
+      source TEXT,
+      tokens_before INTEGER NOT NULL,
+      tokens_after INTEGER NOT NULL,
+      ratio REAL,
+      cost_delta REAL,
+      output_styles TEXT,
+      output_style_bypass TEXT,
+      output_tokens INTEGER
+    )
+  `);
+}
+
+// ─── Tests ───────────────────────────────────────────────────────────────
+
+test("#6848 cleanupDomainCostHistory: deletes rows older than retention window", async () => {
+  const db = getDbInstance()!;
+  const now = Math.floor(Date.now() / 1000);
+  const insert = db.prepare(
+    "INSERT INTO domain_cost_history (api_key_id, cost, timestamp) VALUES (?, ?, ?)"
+  );
+
+  // 3 old (40 days ago), 2 recent (5 days ago)
+  insert.run("key1", 1.0, now - 40 * DAY);
+  insert.run("key1", 2.0, now - 40 * DAY);
+  insert.run("key1", 3.0, now - 40 * DAY);
+  insert.run("key1", 4.0, now - 5 * DAY);
+  insert.run("key1", 5.0, now - 5 * DAY);
+
+  const result = await cleanupDomainCostHistory();
+
+  assert.strictEqual(result.deleted, 3);
+  assert.strictEqual(result.errors, 0);
+
+  const remaining = db.prepare("SELECT COUNT(*) as cnt FROM domain_cost_history").get() as {
+    cnt: number;
+  };
+  assert.strictEqual(remaining.cnt, 2);
+});
+
+test("#6848 cleanupCompressionCacheStats: deletes rows older than retention window", async () => {
+  const db = getDbInstance()!;
+  const oldDate = new Date(Date.now() - 40 * DAY * 1000).toISOString();
+  const recentDate = new Date(Date.now() - 5 * DAY * 1000).toISOString();
+  const insert = db.prepare(
+    "INSERT INTO compression_cache_stats (provider, compression_mode, created_at) VALUES (?, ?, ?)"
+  );
+
+  insert.run("openai", "auto", oldDate);
+  insert.run("openai", "auto", oldDate);
+  insert.run("anthropic", "auto", recentDate);
+
+  const result = await cleanupCompressionCacheStats();
+
+  assert.strictEqual(result.deleted, 2);
+  assert.strictEqual(result.errors, 0);
+
+  const remaining = db.prepare("SELECT COUNT(*) as cnt FROM compression_cache_stats").get() as {
+    cnt: number;
+  };
+  assert.strictEqual(remaining.cnt, 1);
+});
+
+test("#6848 cleanupXpAuditLog: deletes rows older than retention window", async () => {
+  const db = getDbInstance()!;
+  const oldDate = new Date(Date.now() - 40 * DAY * 1000).toISOString();
+  const recentDate = new Date(Date.now() - 5 * DAY * 1000).toISOString();
+  const insert = db.prepare(
+    "INSERT INTO xp_audit_log (api_key_id, action, xp_earned, created_at) VALUES (?, ?, ?, ?)"
+  );
+
+  insert.run("key1", "login", 10, oldDate);
+  insert.run("key1", "login", 10, oldDate);
+  insert.run("key1", "login", 10, oldDate);
+  insert.run("key1", "login", 10, recentDate);
+
+  const result = await cleanupXpAuditLog();
+
+  assert.strictEqual(result.deleted, 3);
+  assert.strictEqual(result.errors, 0);
+
+  const remaining = db.prepare("SELECT COUNT(*) as cnt FROM xp_audit_log").get() as { cnt: number };
+  assert.strictEqual(remaining.cnt, 1);
+});
+
+test("#6848 cleanupCompressionRunTelemetry: deletes rows older than retention window", async () => {
+  ensureTelemetryTable();
+  const db = getDbInstance()!;
+  const now = Math.floor(Date.now() / 1000);
+  const insert = db.prepare(
+    "INSERT INTO compression_run_telemetry (timestamp, tokens_before, tokens_after) VALUES (?, ?, ?)"
+  );
+
+  insert.run(now - 40 * DAY, 1000, 500);
+  insert.run(now - 40 * DAY, 2000, 800);
+  insert.run(now - 5 * DAY, 1500, 600);
+
+  const result = await cleanupCompressionRunTelemetry();
+
+  assert.strictEqual(result.deleted, 2);
+  assert.strictEqual(result.errors, 0);
+
+  const remaining = db.prepare("SELECT COUNT(*) as cnt FROM compression_run_telemetry").get() as {
+    cnt: number;
+  };
+  assert.strictEqual(remaining.cnt, 1);
+});
+
+test("#6848 no rows deleted when all data is within retention window (calls all 4 real functions)", async () => {
+  ensureTelemetryTable();
+  const db = getDbInstance()!;
+  const now = Math.floor(Date.now() / 1000);
+  const recentISO = new Date().toISOString();
+
+  db.prepare("INSERT INTO domain_cost_history (api_key_id, cost, timestamp) VALUES (?, ?, ?)").run(
+    "k",
+    1,
+    now - DAY
+  );
+  db.prepare(
+    "INSERT INTO compression_cache_stats (provider, compression_mode, created_at) VALUES (?, ?, ?)"
+  ).run("p", "auto", recentISO);
+  db.prepare(
+    "INSERT INTO xp_audit_log (api_key_id, action, xp_earned, created_at) VALUES (?, ?, ?, ?)"
+  ).run("k", "a", 5, recentISO);
+  db.prepare(
+    "INSERT INTO compression_run_telemetry (timestamp, tokens_before, tokens_after) VALUES (?, ?, ?)"
+  ).run(now - DAY, 100, 50);
+
+  const r1 = await cleanupDomainCostHistory();
+  const r2 = await cleanupCompressionCacheStats();
+  const r3 = await cleanupXpAuditLog();
+  const r4 = await cleanupCompressionRunTelemetry();
+
+  assert.strictEqual(r1.deleted, 0);
+  assert.strictEqual(r2.deleted, 0);
+  assert.strictEqual(r3.deleted, 0);
+  assert.strictEqual(r4.deleted, 0);
+});
+
+test("#6848 DEFAULT_DATABASE_SETTINGS has new retention keys", async () => {
+  const mod = await import("../../src/types/databaseSettings.ts");
+  const defaults = mod.DEFAULT_DATABASE_SETTINGS.retention;
+
+  assert.ok(typeof defaults.domainCostHistory === "number");
+  assert.ok(typeof defaults.compressionCacheStats === "number");
+  assert.ok(typeof defaults.xpAuditLog === "number");
+  assert.ok(typeof defaults.compressionRunTelemetry === "number");
+
+  assert.strictEqual(defaults.domainCostHistory, 30);
+  assert.strictEqual(defaults.compressionCacheStats, 30);
+  assert.strictEqual(defaults.xpAuditLog, 30);
+  assert.strictEqual(defaults.compressionRunTelemetry, 30);
+});
diff --git a/tests/unit/test-process-detection.test.ts b/tests/unit/test-process-detection.test.ts
new file mode 100644
index 00000000000..bf427e2830f
--- /dev/null
+++ b/tests/unit/test-process-detection.test.ts
@@ -0,0 +1,52 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { isAutomatedTestProcess } from "@/shared/utils/testProcess";
+
+// Regression guard for the substring bug: eleven subsystems decided "am I under a test runner?" with
+// `process.argv.some((a) => a.includes("test"))`. JavaScript agrees that 'latest'.includes('test') is
+// true, so ANY argv carrying a `latest` path segment — a release symlink like /opt/app/latest/server.js,
+// an npm cache path, a `--model=latest` flag — silently put the process into "test mode". For
+// src/lib/db/backup.ts that means isSqliteAutoBackupDisabled() returns true: no SQLite auto-backup, no
+// warning. Same class for migrationRunner, cloud sync, health checks and quota recovery.
+//
+// argv/env are parameters (not read off the global) precisely so this file can assert the negative
+// cases — under the test runner the globals always say "test", which is what made the bug invisible.
+
+const NO_ENV = {} as NodeJS.ProcessEnv;
+
+test("a 'latest' path segment is NOT a test run (the bug that disabled backups)", () => {
+  assert.equal(isAutomatedTestProcess(["node", "/opt/omniroute/latest/server.js"], NO_ENV), false);
+  assert.equal(isAutomatedTestProcess(["node", "C:\\apps\\latest\\bin\\omniroute.mjs", "serve"], NO_ENV), false);
+});
+
+test("other words merely containing 'test' are not test runs either", () => {
+  for (const argv of [
+    ["node", "/srv/protest/app.js"],
+    ["node", "/home/u/contest-bot/index.js"],
+    ["node", "server.js", "--model=latest"],
+    ["node", "/opt/attestation/run.js"],
+  ]) {
+    assert.equal(isAutomatedTestProcess(argv, NO_ENV), false, `argv should not read as a test run: ${argv.join(" ")}`);
+  }
+});
+
+test("the real runners are still detected", () => {
+  assert.equal(isAutomatedTestProcess(["node", "--test", "tests/unit/foo.test.ts"], NO_ENV), true, "node --test");
+  assert.equal(isAutomatedTestProcess(["node", "/repo/tests/unit/foo.test.ts"], NO_ENV), true, "a tests/ path");
+  assert.equal(isAutomatedTestProcess(["node", "/repo/src/foo.test.ts"], NO_ENV), true, "a *.test.ts file");
+  assert.equal(isAutomatedTestProcess(["node", "/repo/node_modules/.bin/vitest", "run"], NO_ENV), true, "vitest binary");
+  assert.equal(isAutomatedTestProcess(["node", "/repo/node_modules/.bin/jest"], NO_ENV), true, "jest binary");
+});
+
+test("env still wins, as before", () => {
+  assert.equal(isAutomatedTestProcess(["node", "server.js"], { NODE_ENV: "test" } as NodeJS.ProcessEnv), true);
+  assert.equal(isAutomatedTestProcess(["node", "server.js"], { VITEST: "1" } as NodeJS.ProcessEnv), true);
+  assert.equal(isAutomatedTestProcess(["node", "server.js"], { NODE_ENV: "production" } as NodeJS.ProcessEnv), false);
+});
+
+test("a production serve is never a test run", () => {
+  assert.equal(
+    isAutomatedTestProcess(["C:\\Program Files\\nodejs\\node.exe", "bin/omniroute.mjs", "serve"], { NODE_ENV: "production" } as NodeJS.ProcessEnv),
+    false
+  );
+});
diff --git a/tests/unit/theoldllm-provider-proxy.test.ts b/tests/unit/theoldllm-provider-proxy.test.ts
new file mode 100644
index 00000000000..f9092268b1b
--- /dev/null
+++ b/tests/unit/theoldllm-provider-proxy.test.ts
@@ -0,0 +1,60 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { fetchTheOldLlmWithProviderProxy } from "../../open-sse/executors/theoldllm.ts";
+
+test("theoldllm dispatches through its provider proxy assignment", async () => {
+  const assignedProxy = {
+    type: "http",
+    host: "residential.example",
+    port: 8080,
+    username: "user",
+    password: "secret",
+  };
+  let observedProxy: unknown = null;
+  let fetchCalls = 0;
+
+  const response = await fetchTheOldLlmWithProviderProxy(
+    { model: "GPT_5_4", messages: [], stream: true },
+    new AbortController().signal,
+    {
+      resolveProxy: async () => assignedProxy,
+      runWithProxy: async (proxy, request) => {
+        observedProxy = proxy;
+        return request();
+      },
+      fetch: (async () => {
+        fetchCalls++;
+        return new Response("ok", { status: 200 });
+      }) as typeof fetch,
+    }
+  );
+
+  assert.equal(response.status, 200);
+  assert.equal(fetchCalls, 1);
+  assert.deepEqual(observedProxy, assignedProxy);
+});
+
+test("theoldllm fails closed when an assigned proxy pool has no active proxy", async () => {
+  let fetchCalls = 0;
+
+  await assert.rejects(
+    () =>
+      fetchTheOldLlmWithProviderProxy(
+        { model: "GPT_5_4", messages: [], stream: true },
+        new AbortController().signal,
+        {
+          resolveProxy: async () => null,
+          hasBlockingProxyAssignment: () => true,
+          runWithProxy: async (_proxy, request) => request(),
+          fetch: (async () => {
+            fetchCalls++;
+            return new Response("unexpected", { status: 200 });
+          }) as typeof fetch,
+        }
+      ),
+    /No active proxy is available/
+  );
+
+  assert.equal(fetchCalls, 0, "a dead assigned proxy pool must never fall back to direct egress");
+});
diff --git a/tests/unit/token-health-check-sweep.test.ts b/tests/unit/token-health-check-sweep.test.ts
new file mode 100644
index 00000000000..841d6d250b2
--- /dev/null
+++ b/tests/unit/token-health-check-sweep.test.ts
@@ -0,0 +1,146 @@
+/**
+ * Regression test: sweep() re-entrancy guard.
+ *
+ * Bug: initTokenHealthCheck() schedules sweep() on both a one-shot startup
+ * timer and a recurring setInterval(sweep, TICK_MS), both fire-and-forget
+ * (unawaited). If a sweep is still processing connections (slowed down by
+ * many OAuth connections plus the inter-connection stagger delay) when the
+ * next tick fires, two sweeps could run concurrently — doubling upstream
+ * refresh traffic and racing DB writes for the same connections.
+ *
+ * Fix: a `sweeping` boolean guard on the shared health-check global state is
+ * set synchronously before the first `await` in sweep() and cleared in a
+ * `finally` block, so a sweep() call that starts while another is already in
+ * flight returns immediately instead of re-entering the loop.
+ *
+ * This test drives the REAL (unmocked) sweep() against a temp SQLite DB —
+ * matching the tests/unit/apikey-connection-health-check.test.ts and
+ * tests/unit/token-health-check.test.ts convention — rather than mocking
+ * @/lib/localDb, since mock.module() is unavailable in this tsx/ESM + Node
+ * native test-runner setup (see tests/unit/proxyfetch-undici-retry.test.ts
+ * and tests/unit/rule12-error-sanitization-sweep.test.ts). Connections are
+ * created with healthCheckInterval: 0 so checkConnection() is a fast no-op
+ * (returns before any network/DB-write work), letting the test control
+ * timing purely via the inter-connection HEALTHCHECK_STAGGER_MS delay.
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+process.env.NODE_ENV = "test";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-sweep-reentrancy-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const core = await import("../../src/lib/db/core.ts");
+const providersDb = await import("../../src/lib/db/providers.ts");
+const { sweep } = await import("../../src/lib/tokenHealthCheck.ts");
+
+async function resetStorage() {
+  core.resetDbInstance();
+  for (let attempt = 0; attempt < 10; attempt++) {
+    try {
+      if (fs.existsSync(TEST_DATA_DIR)) {
+        fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+      }
+      break;
+    } catch (error) {
+      const code = (error as { code?: string } | undefined)?.code;
+      if ((code === "EBUSY" || code === "EPERM") && attempt < 9) {
+        await new Promise((resolve) => setTimeout(resolve, 50 * (attempt + 1)));
+      } else {
+        throw error;
+      }
+    }
+  }
+  fs.mkdirSync(TEST_DATA_DIR, { recursive: true });
+}
+
+test.after(async () => {
+  core.resetDbInstance();
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+/** Read the shared health-check global state's `sweeping` flag. */
+function isSweeping(): boolean {
+  const hc = (globalThis as Record<string, unknown>).__omnirouteTokenHC as
+    | { sweeping?: boolean }
+    | undefined;
+  return hc?.sweeping ?? false;
+}
+
+/** Create `count` distinct OAuth connections whose checkConnection() is a fast no-op. */
+async function createNoOpOauthConnections(count: number, namePrefix: string) {
+  for (let i = 0; i < count; i++) {
+    await providersDb.createProviderConnection({
+      provider: "anthropic",
+      name: `${namePrefix}-${i}`,
+      authType: "oauth",
+      isActive: true,
+      healthCheckInterval: 0, // disabled — checkConnection() returns immediately (no network/db-write)
+    });
+  }
+}
+
+test("sweep() skips re-entrant calls while a previous sweep is still in flight", async () => {
+  await resetStorage();
+  process.env.HEALTHCHECK_STAGGER_MS = "300";
+  await createNoOpOauthConnections(4, "reentrancy");
+
+  const seeded = await providersDb.getProviderConnections({ authType: "oauth" });
+  assert.equal(seeded.length, 4, "precondition: 4 oauth connections exist");
+  assert.equal(isSweeping(), false, "precondition: no sweep in flight yet");
+
+  const first = sweep();
+  // sweep() sets state.sweeping = true synchronously before its first
+  // `await`, so this is already true the instant sweep() returns control to
+  // us — no microtask boundary needed to observe it.
+  assert.equal(isSweeping(), true, "first sweep should mark sweeping=true immediately");
+
+  const start = Date.now();
+  await sweep(); // second, concurrent call — must be skipped by the guard
+  const elapsedMs = Date.now() - start;
+
+  // With 4 connections and a 300ms inter-connection stagger, a REAL second
+  // sweep would take >= 300ms just to clear a single stagger gap (and up to
+  // ~900ms for all three gaps). The guard must make this call return near-
+  // instantly instead of re-entering the loop.
+  assert.ok(
+    elapsedMs < 250,
+    `re-entrant sweep() call should short-circuit almost instantly, took ${elapsedMs}ms`
+  );
+
+  // The first sweep is still running (its stagger delays haven't elapsed) —
+  // proves the second call didn't reset or otherwise interfere with the
+  // first sweep's in-flight state.
+  assert.equal(isSweeping(), true, "first sweep should still be in flight after the skipped call");
+
+  await first;
+  assert.equal(isSweeping(), false, "sweeping flag clears once the in-flight sweep finishes");
+});
+
+test("sweep() resets the sweeping flag after a normal completion, allowing the next sweep to run", async () => {
+  await resetStorage();
+  process.env.HEALTHCHECK_STAGGER_MS = "0";
+  await createNoOpOauthConnections(2, "sequential");
+
+  await sweep();
+  assert.equal(isSweeping(), false, "flag resets after first sweep completes");
+
+  // A second, fully sequential call must not be skipped by leftover state —
+  // it should run and complete normally rather than hang or short-circuit.
+  await sweep();
+  assert.equal(isSweeping(), false, "flag resets after second sweep completes too");
+});
+
+test("sweep() resets the sweeping flag even when there are no connections to process", async () => {
+  await resetStorage();
+  process.env.HEALTHCHECK_STAGGER_MS = "0";
+  // No connections created — sweep() takes the early `connections.length === 0` return.
+
+  await sweep();
+
+  assert.equal(isSweeping(), false, "sweeping flag must be false after an empty sweep");
+});
diff --git a/tests/unit/ui/combos-page-smoke.test.tsx b/tests/unit/ui/combos-page-smoke.test.tsx
new file mode 100644
index 00000000000..9f35b91a2d0
--- /dev/null
+++ b/tests/unit/ui/combos-page-smoke.test.tsx
@@ -0,0 +1,12 @@
+// @vitest-environment jsdom
+import { describe, it, expect } from "vitest";
+
+describe("combos page memoization", () => {
+  it("combos page module exports a default component", async () => {
+    const mod = await import(
+      "@/app/(dashboard)/dashboard/combos/page"
+    );
+    expect(mod.default).toBeDefined();
+    expect(typeof mod.default).toBe("function");
+  });
+});
diff --git a/tests/unit/ui/edgeStyles.test.ts b/tests/unit/ui/edgeStyles.test.ts
index 6d331ba2590..d709fe2c428 100644
--- a/tests/unit/ui/edgeStyles.test.ts
+++ b/tests/unit/ui/edgeStyles.test.ts
@@ -8,7 +8,7 @@ describe("flow edgeStyles (U0 — extracted from ProviderTopology)", () => {
     assert.equal(FLOW_EDGE_COLORS.active, "#22c55e");
     assert.equal(FLOW_EDGE_COLORS.error, "#ef4444");
     assert.equal(FLOW_EDGE_COLORS.last, "#f59e0b");
-    assert.equal(FLOW_EDGE_COLORS.idle, "var(--color-border)");
+    assert.equal(FLOW_EDGE_COLORS.idle, "var(--color-text-muted)");
   });
 
   it("styles an error edge", () => {
@@ -37,9 +37,9 @@ describe("flow edgeStyles (U0 — extracted from ProviderTopology)", () => {
 
   it("styles an idle edge", () => {
     assert.deepEqual(edgeStyle(false, false, false), {
-      stroke: "var(--color-border)",
+      stroke: "var(--color-text-muted)",
       strokeWidth: 1,
-      opacity: 0.2,
+      opacity: 0.3,
     });
   });
 
diff --git a/tests/unit/ui/evals-tab-smoke.test.tsx b/tests/unit/ui/evals-tab-smoke.test.tsx
new file mode 100644
index 00000000000..e9e4dd05e17
--- /dev/null
+++ b/tests/unit/ui/evals-tab-smoke.test.tsx
@@ -0,0 +1,12 @@
+// @vitest-environment jsdom
+import { describe, it, expect } from "vitest";
+
+describe("EvalsTab memoization", () => {
+  it("EvalsTab page module exports a default component", async () => {
+    const mod = await import(
+      "@/app/(dashboard)/dashboard/usage/components/EvalsTab"
+    );
+    expect(mod.default).toBeDefined();
+    expect(typeof mod.default).toBe("function");
+  });
+});
diff --git a/tests/unit/ui/flowCanvas.test.tsx b/tests/unit/ui/flowCanvas.test.tsx
index 58b6d02fb05..362583f997b 100644
--- a/tests/unit/ui/flowCanvas.test.tsx
+++ b/tests/unit/ui/flowCanvas.test.tsx
@@ -1,4 +1,6 @@
 // @vitest-environment jsdom
+import { readFileSync } from "node:fs";
+import { resolve } from "node:path";
 import React, { act } from "react";
 import { createRoot } from "react-dom/client";
 import { describe, it, expect, beforeAll, beforeEach, afterEach } from "vitest";
@@ -48,7 +50,7 @@ const edges = [{ id: "a-b", source: "a", target: "b" }];
 describe("FlowCanvas (U0 — shared ReactFlow wrapper)", () => {
   it("renders the canvas with Controls and hides the attribution", () => {
     const container = mount(<FlowCanvas nodes={nodes} edges={edges} fitKey="x" />);
-    expect(container.querySelector(".react-flow")).toBeTruthy();
+    expect(container.querySelector(".react-flow.omniroute-flow")).toBeTruthy();
     expect(container.querySelector(".react-flow__controls")).toBeTruthy();
     // proOptions.hideAttribution => the attribution element must not render.
     expect(container.querySelector(".react-flow__attribution")).toBeNull();
@@ -60,4 +62,13 @@ describe("FlowCanvas (U0 — shared ReactFlow wrapper)", () => {
     );
     expect(container.querySelector(".omni-test-canvas")).toBeTruthy();
   });
+
+  it("themes controls through React Flow variables so library shorthands cannot override them", () => {
+    const css = readFileSync(resolve("src/app/globals.css"), "utf8");
+
+    expect(css).toContain(".react-flow.omniroute-flow");
+    expect(css).toContain("--xy-controls-button-background-color: var(--color-surface)");
+    expect(css).toContain("--xy-controls-button-color: var(--color-text-main)");
+    expect(css).toContain("--xy-controls-button-border-color: var(--color-border)");
+  });
 });
diff --git a/tests/unit/free-pool-tab.test.tsx b/tests/unit/ui/free-pool-tab.test.tsx
similarity index 90%
rename from tests/unit/free-pool-tab.test.tsx
rename to tests/unit/ui/free-pool-tab.test.tsx
index 91ee2497762..74b90ba8953 100644
--- a/tests/unit/free-pool-tab.test.tsx
+++ b/tests/unit/ui/free-pool-tab.test.tsx
@@ -2,11 +2,12 @@
 import React, { act } from "react";
 import { createRoot } from "react-dom/client";
 import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
+import deMessages from "../../../src/i18n/messages/de.json";
 
 // ── Mocks ─────────────────────────────────────────────────────────────────────
 
 vi.mock("next-intl", () => ({
-  useTranslations: () => (key: string) => key,
+  useTranslations: () => Object.assign((key: string) => key, { has: () => false }),
 }));
 
 // Stub localStorage before importing the component
@@ -32,7 +33,7 @@ Object.defineProperty(globalThis, "localStorage", {
 // ── Import component after mocks ─────────────────────────────────────────────
 
 const { default: FreePoolTab } =
-  await import("../../src/app/(dashboard)/dashboard/settings/components/proxy/FreePoolTab");
+  await import("../../../src/app/(dashboard)/dashboard/settings/components/proxy/FreePoolTab");
 
 // ── Helpers ───────────────────────────────────────────────────────────────────
 
@@ -91,6 +92,36 @@ afterEach(() => {
 // ── Tests ─────────────────────────────────────────────────────────────────────
 
 describe("FreePoolTab source toggles", () => {
+  it("has every render-critical German translation", () => {
+    const requiredKeys = [
+      "freePoolTab",
+      "proxyFreePoolFilterProtocol",
+      "proxyFreePoolProtocol",
+      "proxyFreePoolCountryPlaceholder",
+      "proxyFreePoolFilterCountry",
+      "proxyFreePoolMinQualityPlaceholder",
+      "proxyFreePoolMinQualityLabel",
+      "proxyFreePoolSyncAll",
+      "proxyFreePoolTotal",
+      "proxyFreePoolInPool",
+      "proxyFreePoolAvgQuality",
+      "proxyFreePoolSelected",
+      "proxyFreePoolAddSelected",
+      "proxyFreePoolAddVisible",
+      "proxyFreePoolSource",
+      "proxyFreePoolHostPort",
+      "proxyFreePoolType",
+      "proxyFreePoolCountry",
+      "proxyFreePoolQuality",
+      "proxyFreePoolLatency",
+      "proxyFreePoolEmpty",
+    ] as const;
+
+    for (const key of requiredKeys) {
+      expect(deMessages.settings[key], `missing settings.${key}`).toBeTruthy();
+    }
+  });
+
   it("renders a toggle group with exactly 4 buttons", async () => {
     const el = renderTab();
     await waitForCondition(() => el.querySelector("[role='group']") !== null);
@@ -262,7 +293,7 @@ describe("FreePoolTab sync error surfacing (#5595)", () => {
     await waitForCondition(() => el.querySelector("[role='group']") !== null);
 
     const syncBtn = Array.from(el.querySelectorAll("button")).find((b) =>
-      b.textContent?.includes("proxyFreePoolSyncAll")
+      b.textContent?.includes("Sync all")
     )!;
     expect(syncBtn).toBeTruthy();
     act(() => {
diff --git a/tests/unit/ui/home-provider-topology-section-4606.test.tsx b/tests/unit/ui/home-provider-topology-section-4606.test.tsx
index 2cd2d3ccf24..cd9885a26d5 100644
--- a/tests/unit/ui/home-provider-topology-section-4606.test.tsx
+++ b/tests/unit/ui/home-provider-topology-section-4606.test.tsx
@@ -15,7 +15,10 @@ vi.mock("next-intl", () => ({
 }));
 vi.mock("next/dynamic", () => ({
   default: () => (props: Record<string, unknown>) => (
-    <div data-testid="provider-topology" data-providers={String((props.providers as unknown[])?.length ?? 0)} />
+    <div
+      data-testid="provider-topology"
+      data-providers={String((props.providers as unknown[])?.length ?? 0)}
+    />
   ),
 }));
 vi.mock("@/shared/components", () => ({
@@ -26,9 +29,8 @@ vi.mock("@/hooks/useLiveDashboard", () => ({
   useLiveRequests: () => liveRequestsMock(),
 }));
 
-const { HomeProviderTopologySection } = await import(
-  "../../../src/app/(dashboard)/dashboard/HomeProviderTopologySection"
-);
+const { HomeProviderTopologySection } =
+  await import("../../../src/app/(dashboard)/dashboard/HomeProviderTopologySection");
 
 let container: HTMLDivElement;
 let root: ReturnType<typeof createRoot>;
@@ -63,4 +65,8 @@ it("renders the topology card and forwards providers to ProviderTopology", () =>
   const topology = container.querySelector("[data-testid='provider-topology']");
   expect(topology).not.toBeNull();
   expect(topology?.getAttribute("data-providers")).toBe("2");
+  expect(container.textContent).toContain("activeError");
+  expect(container.textContent).toContain("active");
+  expect(container.textContent).toContain("recent");
+  expect(container.textContent).toContain("modelStatusError");
 });
diff --git a/tests/unit/ui/onboarding-public-endpoint.test.tsx b/tests/unit/ui/onboarding-public-endpoint.test.tsx
new file mode 100644
index 00000000000..b6513af79bc
--- /dev/null
+++ b/tests/unit/ui/onboarding-public-endpoint.test.tsx
@@ -0,0 +1,107 @@
+// @vitest-environment jsdom
+import React from "react";
+import { act } from "react";
+import { createRoot } from "react-dom/client";
+import { afterEach, beforeEach, expect, it, vi } from "vitest";
+
+const { pushMock, replaceMock } = vi.hoisted(() => ({
+  pushMock: vi.fn(),
+  replaceMock: vi.fn(),
+}));
+
+vi.mock("next/navigation", () => ({
+  useRouter: () => ({
+    push: pushMock,
+    replace: replaceMock,
+  }),
+}));
+
+vi.mock("next-intl", () => ({
+  useTranslations: () => (key: string) => key,
+}));
+
+vi.mock("@/shared/hooks", () => ({
+  useDisplayBaseUrl: () => "https://api.example.com",
+}));
+
+const { default: OnboardingWizard } =
+  await import("../../../src/app/(dashboard)/dashboard/onboarding/page");
+
+function jsonResponse(body: unknown): Response {
+  return {
+    ok: true,
+    json: async () => body,
+  } as Response;
+}
+
+let container: HTMLDivElement;
+let root: ReturnType<typeof createRoot>;
+
+async function waitForText(text: string): Promise<void> {
+  for (let attempt = 0; attempt < 20; attempt += 1) {
+    if (container.textContent?.includes(text)) return;
+    await act(async () => {
+      await new Promise((resolve) => setTimeout(resolve, 0));
+    });
+  }
+  throw new Error(`Timed out waiting for text: ${text}`);
+}
+
+function findButton(label: string): HTMLButtonElement {
+  const button = Array.from(container.querySelectorAll("button")).find(
+    (candidate) => candidate.textContent?.trim() === label
+  );
+  if (!button) throw new Error(`Button not found: ${label}`);
+  return button;
+}
+
+async function clickButton(label: string): Promise<void> {
+  await act(async () => {
+    findButton(label).click();
+  });
+}
+
+beforeEach(() => {
+  (
+    globalThis as typeof globalThis & { IS_REACT_ACT_ENVIRONMENT?: boolean }
+  ).IS_REACT_ACT_ENVIRONMENT = true;
+  pushMock.mockReset();
+  replaceMock.mockReset();
+  vi.stubGlobal(
+    "fetch",
+    vi.fn(async (input: RequestInfo | URL) => {
+      if (String(input) === "/api/settings") {
+        return jsonResponse({ setupComplete: false, apiPort: 20131 });
+      }
+      throw new Error(`Unexpected request: ${String(input)}`);
+    })
+  );
+  container = document.createElement("div");
+  document.body.appendChild(container);
+  root = createRoot(container);
+});
+
+afterEach(() => {
+  act(() => root.unmount());
+  container.remove();
+  vi.unstubAllGlobals();
+  vi.clearAllMocks();
+});
+
+it("renders the public API URL instead of rebuilding an internal apiPort URL", async () => {
+  await act(async () => {
+    root.render(<OnboardingWizard />);
+  });
+
+  await waitForText("getStarted");
+  await clickButton("getStarted");
+  await clickButton("skip");
+  await clickButton("skip");
+  await clickButton("skip");
+  await clickButton("skip");
+
+  await waitForText("https://api.example.com/api/v1");
+  expect(container.textContent).toContain("https://api.example.com/api/v1");
+  expect(container.textContent).not.toContain(":20131/api/v1");
+  expect(replaceMock).not.toHaveBeenCalled();
+});
diff --git a/tests/unit/ui/rtkTomlImportCard.test.tsx b/tests/unit/ui/rtkTomlImportCard.test.tsx
new file mode 100644
index 00000000000..660f75fbad7
--- /dev/null
+++ b/tests/unit/ui/rtkTomlImportCard.test.tsx
@@ -0,0 +1,179 @@
+// @vitest-environment jsdom
+import React, { act } from "react";
+import { createRoot } from "react-dom/client";
+import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
+
+vi.mock("next-intl", () => ({
+  useTranslations: () => (key: string, params?: Record<string, unknown>) =>
+    params ? `${key}:${JSON.stringify(params)}` : key,
+}));
+
+const { default: RtkTomlImportCard } =
+  await import("@/app/(dashboard)/dashboard/context/rtk/RtkTomlImportCard");
+
+const containers: HTMLElement[] = [];
+
+function mount(ui: React.ReactElement): HTMLElement {
+  const container = document.createElement("div");
+  document.body.appendChild(container);
+  containers.push(container);
+  act(() => createRoot(container).render(ui));
+  return container;
+}
+
+async function setTextarea(container: HTMLElement, value: string) {
+  const input = container.querySelector("[data-testid='rtk-toml-content']") as HTMLTextAreaElement;
+  await act(async () => {
+    const setter = Object.getOwnPropertyDescriptor(
+      window.HTMLTextAreaElement.prototype,
+      "value"
+    )!.set!;
+    setter.call(input, value);
+    input.dispatchEvent(new Event("input", { bubbles: true }));
+  });
+}
+
+async function click(element: Element | null) {
+  await act(async () => (element as HTMLElement).click());
+}
+
+async function selectFile(container: HTMLElement, file: File) {
+  const input = container.querySelector("[data-testid='rtk-toml-file']") as HTMLInputElement;
+  Object.defineProperty(input, "files", { configurable: true, value: [file] });
+  await act(async () => input.dispatchEvent(new Event("change", { bubbles: true })));
+}
+
+beforeEach(() => {
+  (
+    globalThis as typeof globalThis & { IS_REACT_ACT_ENVIRONMENT?: boolean }
+  ).IS_REACT_ACT_ENVIRONMENT = true;
+});
+
+afterEach(() => {
+  while (containers.length) containers.pop()?.remove();
+  document.body.innerHTML = "";
+  vi.restoreAllMocks();
+  vi.unstubAllGlobals();
+});
+
+describe("RtkTomlImportCard", () => {
+  it("validates TOML without requesting installation", async () => {
+    const fetchMock = vi.fn(async () =>
+      Response.json({
+        sha256: "abc",
+        passed: true,
+        filters: [
+          {
+            id: "my-tool",
+            description: "",
+            category: "generic",
+            commandPatterns: ["^my-tool"],
+            testCount: 1,
+          },
+        ],
+        outcomes: [{ filterId: "my-tool", testName: "works", passed: true }],
+        warnings: [],
+      })
+    );
+    vi.stubGlobal("fetch", fetchMock);
+    const container = mount(<RtkTomlImportCard />);
+
+    await setTextarea(container, "schema_version = 1");
+    await click(container.querySelector("[data-testid='rtk-toml-validate']"));
+
+    const [, options] = fetchMock.mock.calls[0];
+    expect(JSON.parse(String(options?.body))).toMatchObject({
+      action: "validate",
+      overwrite: false,
+    });
+    expect(container.querySelector("[data-testid='rtk-toml-result']")?.textContent).toContain(
+      "my-tool"
+    );
+  });
+
+  it("installs with explicit overwrite and refreshes the filter catalog", async () => {
+    const onInstalled = vi.fn(async () => {});
+    const fetchMock = vi.fn(async () =>
+      Response.json({
+        sha256: "abc",
+        passed: true,
+        filters: [],
+        outcomes: [],
+        warnings: [],
+        installedPath: "rtk/filters.toml",
+        backupCreated: true,
+      })
+    );
+    vi.stubGlobal("fetch", fetchMock);
+    const container = mount(<RtkTomlImportCard onInstalled={onInstalled} />);
+
+    await setTextarea(container, "schema_version = 1");
+    await click(container.querySelector("[data-testid='rtk-toml-overwrite']"));
+    await click(container.querySelector("[data-testid='rtk-toml-install']"));
+
+    const [, options] = fetchMock.mock.calls[0];
+    expect(JSON.parse(String(options?.body))).toMatchObject({ action: "install", overwrite: true });
+    expect(onInstalled).toHaveBeenCalledOnce();
+    expect(container.querySelector("[data-testid='rtk-toml-result']")?.textContent).toContain(
+      "tomlBackupCreated"
+    );
+  });
+
+  it("shows the sanitized API error message", async () => {
+    vi.stubGlobal(
+      "fetch",
+      vi.fn(async () =>
+        Response.json({ error: { message: "filter 'bad' has an unsafe regex" } }, { status: 400 })
+      )
+    );
+    const container = mount(<RtkTomlImportCard />);
+
+    await setTextarea(container, "schema_version = 1");
+    await click(container.querySelector("[data-testid='rtk-toml-validate']"));
+
+    expect(container.querySelector("[data-testid='rtk-toml-error']")?.textContent).toContain(
+      "unsafe regex"
+    );
+  });
+
+  it("shows failing inline-test outcomes without reporting a successful validation", async () => {
+    vi.stubGlobal(
+      "fetch",
+      vi.fn(async () =>
+        Response.json({
+          sha256: "abc",
+          passed: false,
+          filters: [],
+          outcomes: [{ filterId: "bad", testName: "expected output", passed: false }],
+          warnings: [],
+        })
+      )
+    );
+    const container = mount(<RtkTomlImportCard />);
+
+    await setTextarea(container, "schema_version = 1");
+    await click(container.querySelector("[data-testid='rtk-toml-validate']"));
+
+    const result = container.querySelector("[data-testid='rtk-toml-result']")?.textContent ?? "";
+    expect(result).toContain("tomlValidationFailed");
+    expect(result).toContain("tomlTestFailed");
+    expect(result).not.toContain("tomlValidationPassed");
+  });
+
+  it("rejects oversized files before reading them", async () => {
+    const container = mount(<RtkTomlImportCard />);
+    await setTextarea(container, "existing content");
+    const file = new File([new Uint8Array(1024 * 1024 + 1)], "filters.toml");
+    const text = vi.spyOn(file, "text");
+
+    await selectFile(container, file);
+
+    expect(text).not.toHaveBeenCalled();
+    expect(container.querySelector("[data-testid='rtk-toml-error']")?.textContent).toContain(
+      "tomlFileReadError"
+    );
+    expect(
+      (container.querySelector("[data-testid='rtk-toml-content']") as HTMLTextAreaElement).value
+    ).toBe("existing content");
+  });
+});
diff --git a/tests/unit/ui/sidebar-proxy-expansion.test.tsx b/tests/unit/ui/sidebar-proxy-expansion.test.tsx
new file mode 100644
index 00000000000..3dac6c95ec3
--- /dev/null
+++ b/tests/unit/ui/sidebar-proxy-expansion.test.tsx
@@ -0,0 +1,46 @@
+import { describe, expect, it } from "vitest";
+
+import {
+  HIDEABLE_SIDEBAR_ITEM_IDS,
+  SIDEBAR_SECTIONS,
+  getSectionItems,
+  normalizeHiddenSidebarItems,
+  type SidebarSectionId,
+} from "../../../src/shared/constants/sidebarVisibility";
+import {
+  expandActiveSection,
+  hydrateExpandedSections,
+  toggleExpandedSection,
+} from "../../../src/shared/utils/sidebarExpansionState";
+
+describe("sidebar proxy expansion", () => {
+  it("proxy navigation is always present and cannot be hidden by legacy settings", () => {
+    const omniProxy = SIDEBAR_SECTIONS.find((section) => section.id === "omni-proxy");
+    expect(omniProxy).toBeDefined();
+    expect(getSectionItems(omniProxy!).some((item) => item.id === "proxy")).toBe(true);
+    expect((HIDEABLE_SIDEBAR_ITEM_IDS as readonly string[]).includes("proxy")).toBe(false);
+    expect(normalizeHiddenSidebarItems(["proxy", "logs"])).toEqual(["logs"]);
+  });
+
+  it("opening another section closes all unpinned siblings", () => {
+    const expanded = new Set<SidebarSectionId>(["omni-proxy", "analytics"]);
+    const next = toggleExpandedSection(expanded, new Set(), "configuration");
+    expect([...next]).toEqual(["configuration"]);
+  });
+
+  it("hydration preserves a stored all-collapsed state", () => {
+    const expanded = hydrateExpandedSections([], new Set());
+    expect([...expanded]).toEqual([]);
+  });
+
+  it("hydration expands only sections explicitly pinned by the user", () => {
+    const expanded = hydrateExpandedSections([], new Set<SidebarSectionId>(["monitoring"]));
+    expect([...expanded]).toEqual(["monitoring"]);
+  });
+
+  it("route changes replace stale unpinned sections but retain explicit pins", () => {
+    const pinned = new Set<SidebarSectionId>(["configuration"]);
+    const next = expandActiveSection(pinned, "monitoring");
+    expect([...next]).toEqual(["configuration", "monitoring"]);
+  });
+});
diff --git a/tests/unit/ui/useToolBatchStatuses.test.tsx b/tests/unit/ui/useToolBatchStatuses.test.tsx
index adf23f88d17..55ab4c15ddf 100644
--- a/tests/unit/ui/useToolBatchStatuses.test.tsx
+++ b/tests/unit/ui/useToolBatchStatuses.test.tsx
@@ -87,8 +87,9 @@ async function mountHook(): Promise<{
 // ── Lifecycle ─────────────────────────────────────────────────────────────────
 
 beforeEach(() => {
-  (globalThis as typeof globalThis & { IS_REACT_ACT_ENVIRONMENT?: boolean }).IS_REACT_ACT_ENVIRONMENT =
-    true;
+  (
+    globalThis as typeof globalThis & { IS_REACT_ACT_ENVIRONMENT?: boolean }
+  ).IS_REACT_ACT_ENVIRONMENT = true;
 });
 
 afterEach(() => {
@@ -165,7 +166,11 @@ describe("useToolBatchStatuses", () => {
       await new Promise((r) => setTimeout(r, 50));
     });
 
-    expect((mockFetch as ReturnType<typeof vi.fn>).mock.calls.length).toBeGreaterThan(callsAfterMount);
+    expect((mockFetch as ReturnType<typeof vi.fn>).mock.calls.length).toBeGreaterThan(
+      callsAfterMount
+    );
+    const calls = (mockFetch as ReturnType<typeof vi.fn>).mock.calls;
+    expect(calls.at(-1)?.[0]).toBe("/api/cli-tools/all-statuses?refresh=true");
   });
 
   it("registers focus event listener on mount", async () => {

From 75df3e642f017305c265802787c08b936eea2c78 Mon Sep 17 00:00:00 2001
From: Minxi Hou <houminxi@gmail.com>
Date: Sat, 18 Jul 2026 16:07:39 -0300
Subject: [PATCH 14/14] fix(antigravity): keep buffered JSON contract for
 non-streaming callers
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

#3786's Pro-family fallback-chain retry loop (execute()) calls executeOnce()
per candidate and inspects result.response directly, expecting a
synthesized chat.completion JSON body on success. The streaming-passthrough
migration made ALL non-streaming (stream: false) responses a raw SSE
pass-through instead, so a successful retry candidate's response.json()
threw ("data: {...}" is not valid JSON) — breaking the fallback chain
(tests/unit/agy-pro-fallback-chain-3786.test.ts, 3 of 13 red).

Route non-streaming (stream: false) responses back through
collectStreamToResponse (buffered collect-to-JSON), which already uses
FETCH_TIMEOUT_MS with no hardcoded 120s ceiling, so long-thinking models
are not penalized. Passthrough is reserved for actual streaming clients
(stream: true), which was the PR's real target scenario.

Extracted the branch into buildAntigravityAttemptResult() to keep
runAntigravityAttempt under the 80-line ratchet cap.

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
---
 open-sse/executors/antigravity.ts | 53 +++++++++++++++++++++++++++++--
 1 file changed, 51 insertions(+), 2 deletions(-)

diff --git a/open-sse/executors/antigravity.ts b/open-sse/executors/antigravity.ts
index c3d4e4fbe8b..e56ea921640 100644
--- a/open-sse/executors/antigravity.ts
+++ b/open-sse/executors/antigravity.ts
@@ -1258,7 +1258,8 @@ export class AntigravityExecutor extends BaseExecutor {
     );
     if (embedded) return { action: "return", result: embedded };
 
-    const result = await buildFinalAntigravityResult(
+    const result = await this.buildAntigravityAttemptResult(
+      model,
       stream,
       response,
       url,
@@ -1267,11 +1268,59 @@ export class AntigravityExecutor extends BaseExecutor {
       requestToolNameMap,
       accountId,
       signal,
-      updateAntigravityRemainingCredits
+      log
     );
     return { action: "return", result };
   }
 
+  /**
+   * #3786 — Non-streaming callers (stream: false) keep the buffered
+   * collect-to-JSON contract: `execute()` (including the Pro-family
+   * fallback-chain retry loop) inspects `result.response` directly and
+   * expects a synthesized `chat.completion` JSON body, not a raw SSE
+   * pass-through. Passthrough is reserved for actual streaming clients
+   * (buildFinalAntigravityResult's stream:true branch), where the client
+   * itself drains the SSE bytes — collectStreamToResponse already uses
+   * FETCH_TIMEOUT_MS (no hardcoded 120s ceiling), so long-thinking models
+   * are not penalized by buffering here.
+   */
+  private async buildAntigravityAttemptResult(
+    model: string,
+    stream: boolean,
+    response: Response,
+    url: string,
+    finalHeaders: Record<string, string>,
+    transformedBody: Record<string, unknown>,
+    requestToolNameMap: Map<string, string> | null,
+    accountId: string,
+    signal: AbortSignal | null | undefined,
+    log: SafeAntigravityLog
+  ): Promise<SsePassthroughResult> {
+    if (!stream && response.ok && response.body) {
+      return this.collectStreamToResponse(
+        response,
+        model,
+        url,
+        finalHeaders,
+        transformedBody,
+        log,
+        signal
+      );
+    }
+
+    return buildFinalAntigravityResult(
+      stream,
+      response,
+      url,
+      finalHeaders,
+      transformedBody,
+      requestToolNameMap,
+      accountId,
+      signal,
+      updateAntigravityRemainingCredits
+    );
+  }
+
   /**
    * Handle a 429/503 response for one URL-index attempt: resolve the retry-after
    * time (headers, then error-body classification + Google-One-AI credits retry),