Conversation
Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
…9184) Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
…iegosouzapw#8870) Validated in local merge-train T4 (HouMinXi+Zartharas+Andrian+artickc)
…egosouzapw#8877) Validated in local merge-train T5 (base49+contributors+pacocartones)
…iegosouzapw#9077) Validated in local merge-train T5 (base49+contributors+pacocartones)
…dflare Workers AI (diegosouzapw#8717) (diegosouzapw#8808) Validated in local merge-train T5 (base49+contributors+pacocartones)
Validated in local merge-train T5 (base49+contributors+pacocartones)
…zapw#8878) Validated in local merge-train T5 (base49+contributors+pacocartones)
…diegosouzapw#8958) (diegosouzapw#8961) Validated in local merge-train T5 (base49+contributors+pacocartones)
…holders (diegosouzapw#9001) Validated in local merge-train T5 (base49+contributors+pacocartones)
…trict BYOK providers (diegosouzapw#9005) Validated in local merge-train T5 (base49+contributors+pacocartones)
…souzapw#9020) Validated in local merge-train T5 (base49+contributors+pacocartones)
…apw#9065) Validated in local merge-train T5 (base49+contributors+pacocartones)
… /login (diegosouzapw#9097) Validated in local merge-train T5 (base49+contributors+pacocartones)
…uzapw#9063) Validated in local merge-train T5 (base49+contributors+pacocartones)
…ouzapw#9073) Validated in local merge-train T5 (base49+contributors+pacocartones)
…iegosouzapw#9083) Validated in local merge-train T5 (base49+contributors+pacocartones)
…ents (diegosouzapw#9088) Validated in local merge-train T5 (base49+contributors+pacocartones)
…ublish resolver (diegosouzapw#9553) * fix(build): resolve npm-cli.js on POSIX layouts in the shim-free prepublish resolver The diegosouzapw#8858 resolver only tried <dir(node)>/node_modules/npm/bin — the Windows layout. On POSIX (GitHub hosted runners, nvm, system installs) npm lives at <prefix>/lib/node_modules/npm while node is <prefix>/bin/ node, so resolveBundledNpmEntry returned null and npm run build:cli died installing @omniroute/opencode-plugin deps on every fresh checkout ('npm-cli.js not found next to the running Node binary') — redding Fast Production Build and dast-smoke for the whole PR queue. Extract the resolver to scripts/build/resolveNpmEntry.ts with injectable seams and try, in order: npm_execpath (exported by npm run itself), the Windows beside-the-binary layout, the POSIX <prefix>/lib layout. TDD: tests/unit/build/resolve-npm-entry.test.ts — the POSIX-layout and npm_execpath cases plus a live regression guard fail against the old single-candidate logic (2/5) and pass with the fix (5/5). * docs(env): register the 7 env vars orphaned by the 08-05 merge batch The Docs Gates env/docs contract went red on the release tip: diegosouzapw#9260 added OMNIROUTE_INTERNAL_SERVICE_TOKEN(_FILE) and diegosouzapw#9324 added OPENROUTER_PROVIDER_STATS_ENABLED/_TTL_MS without .env.example entries, and the diegosouzapw#9286 Redis sidecar vars (REDIS_BIND_HOST, REDIS_PORT, OMNIROUTE_REDIS_BIND_HOST) never reached ENVIRONMENT.md. Inherited base-red on every open PR. Defaults and descriptions taken from the consuming source files. --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>
…iegosouzapw#9554) * fix(quality): reconcile inherited file-size drift on the release tip 13 files sit above their frozen LOC on the clean tip 8180b49 (measured by the gate itself). The PR-mode base-relative check (diegosouzapw#8522) correctly lets innocent PRs pass, but per-PR rebaselines were lost across successive conflict resolutions of this hot file during the 08-05/06 merge batch — so the absolute mode (nightly, local runs) is permanently red and stops distinguishing real growth from inherited drift. Frozen values updated to the measured tip, each annotated with the merged PR that grew the file (diegosouzapw#9024 diegosouzapw#9324 diegosouzapw#9329 diegosouzapw#9193 diegosouzapw#9332 diegosouzapw#9228 diegosouzapw#9236 diegosouzapw#9314 diegosouzapw#9260 diegosouzapw#8934 diegosouzapw#9196 diegosouzapw#9163); executors default.ts and kiro.ts (above the 1000 cap with no frozen entry) join the frozen set. * fix(quality): prune orphaned ESLint suppressions and clear the 5 unsuppressed errors The 'No new ESLint warnings' job reds the whole queue with exit 2: 'There are suppressions left that do not occur anymore' — the 08-05 merge batch removed code whose violations were frozen in eslint-suppressions.json, leaving orphaned entries (673->670 files, 4338->4333 violations after eslint --prune-suppressions). The full-tree run also surfaced 5 real unsuppressed errors merged with the batch, fixed here instead of suppressed (new violations must be fixed, per policy): 4x no-explicit-any in tests/unit/catalog-order-contract.test.ts ((conn as any).id -> typed cast) and 1x react/no-unescaped-entities in the agent-bridge SetupWizard (diegosouzapw#9095). Also restores the _comment policy header the successive hot-file conflict resolutions had dropped (TS7 debt freeze provenance + prune policy). * fix(quality): absorb the two file-size growths merged while this PR was in CI The base kept moving during the reconcile cycle: diegosouzapw#9184 grew src/sse/handlers/chat.ts 1857->1877 and diegosouzapw#9005 grew open-sse/executors/default.ts 1027->1042. Re-measured on the merged tree; gate back to 0 violations. * fix(tests): move the orphaned RTL ratchet test to a collected path as node:test diegosouzapw#8828 added tests/unit/scripts/check-rtl-ratchet.test.ts — a path no runner collects (the node:test globs enumerate an explicit subdir list without scripts/, and vitest.config.ts never included it), so the file NEVER ran and the test-discovery orphan gate reds the queue. Moved to tests/unit/ (collected by node:test) and converted from vitest describe/it/expect to node:test+assert to match the runner and the sibling check-*.test.ts files. 5/5 green under the real runner. --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>
…e from (diegosouzapw#9559) * fix(mcp): give the audit tests a loader seam createRequire cannot hide from Since diegosouzapw#8959 the audit DB loads better-sqlite3 via createRequire() (so Electron/global-install resolution works) — which vi.doMock cannot intercept: it only patches Vitest's ESM module graph. The audit.test.ts better-sqlite3 mock therefore never engaged; the tests opened a REAL empty sqlite file in the temp DATA_DIR ('no such table: mcp_tool_audit' on stderr) and every mock assertion counted zero calls. The 3 failures are deterministic (reproduced 3/3 locally), redding Vitest (fast-path) for the entire PR queue — long misdiagnosed as a flake (diegosouzapw#9095 merge notes call it 'pre-existing audit.test.ts flake'). - Shutdown tests inject the mock through the audit connection cache (globalThis.__omnirouteMcpAuditDb) — the module's own seam. - The node:sqlite fallback test drives __setBetterSqliteLoaderForTests, a test-only loader override; the production createRequire path is untouched (node:sqlite itself is import()'d, so its doMock still works). 3/3 red -> 3/3 green; full open-sse/mcp-server vitest suite 88/88. * chore: align changelog slug with the PR number (9559) --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>
…moke base-red) (diegosouzapw#9558) * fix(build): exec native tool binaries directly in runBuildTool diegosouzapw#8858 routed every resolved local bin through process.execPath to avoid Windows .cmd shims — but esbuild >=0.25 ships bin/esbuild as the NATIVE platform executable (ELF on Linux), so Node parsed machine code as JS and build:cli died with 'SyntaxError: Invalid or unexpected token', turning dast-smoke red for every PR. runBuildTool now sniffs the entry's magic bytes (ELF / Mach-O / PE) and execs native binaries directly; JS entries keep going through this Node binary (the .cmd-shim avoidance diegosouzapw#8858 wanted). Validation (RED->GREEN on this box): - RED: node node_modules/esbuild/bin/esbuild --version -> SyntaxError (ELF) - GREEN: the exact failing CI step reproduced via the new logic bundles open-sse/mcp-server/server.ts successfully (4.2MB output, 1.3s). * fix(docs): add MDX frontmatter to the 20 remaining docs without it Same failure class as AGENTROUTER_WAF (diegosouzapw#9503) and DOCKER_RELEASE_CHANNELS (this run's dast-smoke red): any doc without frontmatter breaks the fumadocs MDX loader during next build, killing build:cli/dast-smoke for every PR. Swept ALL of docs/ (i18n mirrors excluded) in one pass so this class cannot recur one file at a time. * docs(env): document OMNIROUTE_INTERNAL_SERVICE_TOKEN(+_FILE), OPENROUTER_PROVIDER_STATS_* and embedded-Redis binding vars Pre-existing env/docs contract drift from recently merged features made check:env-doc-sync red for any docs-touching PR. Values and defaults read from the defining modules (internalServiceAuth.ts, openrouterProviderStats.ts). * fix(build): resolve bundled npm-cli.js in the standard Unix layout + safe npm fallback off-Windows The opencode-plugin step hard-failed on GitHub runners because resolveBundledNpmEntry only looked next to the node binary (Windows zip layout); hostedtoolcache Node keeps npm at <prefix>/lib/node_modules/npm. Added that candidate, and when neither exists on non-Windows the step now falls back to plain 'npm' — the .cmd-shim hazard diegosouzapw#8858 avoids is Windows-only. * test(mutation): register xai-agent-tools-passthrough.test.ts in stryker tap.testFiles The test landed on release/v3.8.50 covering open-sse/handlers/chatCore/passthroughHelpers.ts without the stryker registration, so Fast Quality Gates' drift detection reds any PR that carries it. Mechanical registration so its mutant kills count. --------- Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>
…8339) Validated in local merge-train T6 (ungrouped batch 1)
…quality-validation-benign-error fix(combo): ignore benign empty error fields in streaming quality validation (502 false-positive on opencode tool calls)
Owner
|
Thanks for the PR. Please address the mandatory items (tests and/or merge blockers) in this branch, then rerun checks before /merge-prs. |
2 similar comments
Owner
|
Thanks for the PR. Please address the mandatory items (tests and/or merge blockers) in this branch, then rerun checks before /merge-prs. |
Owner
|
Thanks for the PR. Please address the mandatory items (tests and/or merge blockers) in this branch, then rerun checks before /merge-prs. |
Co-authored-by: diegosouzapw <diegosouzapw@users.noreply.github.com>
…pr-9296 # Conflicts: # open-sse/config/imageRegistry.ts # open-sse/handlers/imageGeneration/providers/adobeFirefly.ts # open-sse/services/adobeFireflyClient.ts # tests/unit/adobe-firefly.test.ts
Owner
|
Obrigado pelo PR. Mantive a revisão de
|
…odel-capabilities fix(adobe-firefly): sync discovered models and capabilities
Agent clients (OpenCode, Claude Code, Cursor) fan out heavy sub-requests that land on the admission gate together. With the single heavyweight slot, concurrent heavy requests were rejected immediately with a retryable 503; clients burn their retry budget in seconds and the agent dies mid-task. Heavy requests now wait up to OMNIROUTE_CHAT_ADMISSION_QUEUE_MS (default 5000ms) for a slot before the 503, served FIFO; 0 restores the legacy immediate-reject behaviour. Applied to both the byte-based path (admitChatRequest) and the structure-based path (admitChatStructure, now async).
Keeps the env/docs contract (check-env-doc-sync, diegosouzapw#7793) in sync with the new admission queue budget introduced in this PR.
herjarsa
force-pushed
the
fix/chat-admission-queue
branch
from
August 9, 2026 09:20
c065c12 to
5441321
Compare
Contributor
Author
|
Rebased onto the current Fixed in this update:
Remaining CI failures are pre-existing base drift, not introduced by this PR (none of the affected modules/files are touched by the diff):
These look like the base branch's own accumulated ratchet drift (the release branch is mid-merge per the other open PRs). Happy to address any of them here if you'd prefer they be fixed in this PR rather than the base. |
The base's diegosouzapw#9697 (radar referral links) and the New-API aggregator work added keys to en.json and 42 locales but left vi.json behind: 23 missing radarPage/providers/miniPlayground keys plus 8 providers.* entries still carrying __MISSING__ markers. That broke i18n-vi-completeness and the radar page-tab/key-input tests for every PR against release/v3.8.50. Adds the missing translations and replaces the __MISSING__ placeholders, restoring exact en<->vi key parity (11391/11391).
Contributor
Author
|
Third push — fixes the remaining unit-test failures, all of which were Vietnamese-locale drift inherited from the base (not from the admission change):
Verified locally: all 18 i18n tests pass, prettier clean, ypecheck:core clean, full |
Contributor
Author
|
Superseded by #NEW — cherry-pick of the 3 commits onto current main. |
Contributor
Author
Re-opened with rebased commitsThis PR was closed and re-opened with a fresh rebased branch \herjarsa:rebase/9816-chat-admission-queue. The original 3 commits have been cherry-picked onto current \main:
Resolutions applied (all took the PR version via --theirs):
Diff stats: 7 files changed, 1014 insertions(+), 78 deletions(-) |
Contributor
Author
|
Closing to re-open as a clean PR with the rebased commits. The original PR carried 1812 accumulated commits (141k additions) against an old main; the rebased version has only the 3 intended commits. |
This was referenced Aug 12, 2026
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
Agent-style clients (OpenCode, Claude Code, Cursor) fan out heavy sub-requests that land on the chat admission gate together. With the single heavyweight slot (OMNIROUTE_CHAT_MAX_HEAVY_IN_FLIGHT=1), the second concurrent heavy request was rejected immediately with a retryable 503 chat_admission_busy.
Those clients retry on a ~1s cadence; a burst of 2-3 heavy sub-requests burns the whole retry budget in seconds and the agent dies mid-task with no user-visible error — the UI just goes blank/waiting while later requests log 200.
Fix
Heavy chat requests now wait for a slot instead of failing instantly:
ull on deadline.