[merge after #11] Profiling by hjc-puro · Pull Request #12 · NousResearch/hermes-agent

hjc-puro · 2025-11-18T12:12:35Z

Aggregates logged times for openai calls and tool calls

…urface Clean CSS polish, no conflicts.

…arg mismatch' (NousResearch#12) from fix/delegate-workspace-kwarg into main

Address adversarial review findings: 1. Race condition (#1): Job-level concurrency with cancel-in-progress collapses back-to-back pushes; ref: main checkout always gets latest branch state; explicit push target (origin HEAD:main). 2. Loop prevention (#2): File-whitelist check before commit aborts if any file outside nix/{tui,web}.nix was modified, preventing accidental self-triggering. 3. Silent infra failures (#8): nix-lockfile-check now fails explicitly when fix-lockfiles exits without reporting stale status (catches nix setup failures, network errors, script bugs that bypass continue-on-error). 4. Commit traceability (#11): Auto-fix commits include source SHA and workflow run URL in the commit body. 5. Explicit push target (#12): git push origin HEAD:main instead of bare git push.

* ci(nix): auto-fix stale npm hashes on push to main When a PR merges to main with updated package-lock.json or package.json in ui-tui/ or web/, the new auto-fix-main job detects stale npmDepsHash values and pushes a fix commit directly to main. This eliminates the recurring manual hash-bump PRs (#15420, #15314, #15272, #15244) by reusing the existing fix-lockfiles --apply pipeline. The fix commit only touches nix/*.nix files, which are outside the push path filter (package-lock.json / package.json), so it cannot re-trigger itself. Closes #15314 * fix(ci): use GitHub App token for auto-fix-main push GITHUB_TOKEN commits are invisible to workflow triggers (GitHub's infinite-loop prevention). The auto-fix-main job pushes directly to main, so the fix commit never triggered downstream nix.yml verification. Mint a short-lived token via the repo's GitHub App (daimon-nous, APP_ID + APP_PRIVATE_KEY secrets) so the push is treated as a real event and nix.yml fires to verify the corrected hashes. Tested via workflow_dispatch dry-run: app token minted successfully, checkout with app token succeeded, fix job correctly gated. Resolves review feedback from Bugbot (r3144569551). * ci(nix): rename lockfile check job for required status check Rename 'check' → 'nix-lockfile-check' so the status check name is unambiguous when added as a required check on main. * fix(ci): harden auto-fix-main against races, loops, and silent failures Address adversarial review findings: 1. Race condition (#1): Job-level concurrency with cancel-in-progress collapses back-to-back pushes; ref: main checkout always gets latest branch state; explicit push target (origin HEAD:main). 2. Loop prevention (#2): File-whitelist check before commit aborts if any file outside nix/{tui,web}.nix was modified, preventing accidental self-triggering. 3. Silent infra failures (#8): nix-lockfile-check now fails explicitly when fix-lockfiles exits without reporting stale status (catches nix setup failures, network errors, script bugs that bypass continue-on-error). 4. Commit traceability (#11): Auto-fix commits include source SHA and workflow run URL in the commit body. 5. Explicit push target (#12): git push origin HEAD:main instead of bare git push. --------- Co-authored-by: alt-glitch <alt-glitch@users.noreply.github.com>

* ci(nix): auto-fix stale npm hashes on push to main When a PR merges to main with updated package-lock.json or package.json in ui-tui/ or web/, the new auto-fix-main job detects stale npmDepsHash values and pushes a fix commit directly to main. This eliminates the recurring manual hash-bump PRs (NousResearch#15420, NousResearch#15314, NousResearch#15272, NousResearch#15244) by reusing the existing fix-lockfiles --apply pipeline. The fix commit only touches nix/*.nix files, which are outside the push path filter (package-lock.json / package.json), so it cannot re-trigger itself. Closes NousResearch#15314 * fix(ci): use GitHub App token for auto-fix-main push GITHUB_TOKEN commits are invisible to workflow triggers (GitHub's infinite-loop prevention). The auto-fix-main job pushes directly to main, so the fix commit never triggered downstream nix.yml verification. Mint a short-lived token via the repo's GitHub App (daimon-nous, APP_ID + APP_PRIVATE_KEY secrets) so the push is treated as a real event and nix.yml fires to verify the corrected hashes. Tested via workflow_dispatch dry-run: app token minted successfully, checkout with app token succeeded, fix job correctly gated. Resolves review feedback from Bugbot (r3144569551). * ci(nix): rename lockfile check job for required status check Rename 'check' → 'nix-lockfile-check' so the status check name is unambiguous when added as a required check on main. * fix(ci): harden auto-fix-main against races, loops, and silent failures Address adversarial review findings: 1. Race condition (NousResearch#1): Job-level concurrency with cancel-in-progress collapses back-to-back pushes; ref: main checkout always gets latest branch state; explicit push target (origin HEAD:main). 2. Loop prevention (NousResearch#2): File-whitelist check before commit aborts if any file outside nix/{tui,web}.nix was modified, preventing accidental self-triggering. 3. Silent infra failures (NousResearch#8): nix-lockfile-check now fails explicitly when fix-lockfiles exits without reporting stale status (catches nix setup failures, network errors, script bugs that bypass continue-on-error). 4. Commit traceability (NousResearch#11): Auto-fix commits include source SHA and workflow run URL in the commit body. 5. Explicit push target (NousResearch#12): git push origin HEAD:main instead of bare git push. --------- Co-authored-by: alt-glitch <alt-glitch@users.noreply.github.com>

12 輪 iteration 累積核心可交付物： 1. skill-helper-template/ — Missive helper（已 pilot 部署到 runtime ✅） - query.py：純 stdlib，繞 hermes 7 道閘 - install.sh / adopt.sh：runtime 部署 + 業務 repo 採納自動化 - missive-skill-patch.md：CK_Missive SKILL.md 直接複製貼上 - README.md：5/5 採納就緒度表 + 4 步完整解阻塞最小路徑 2. ck-{lvrland,pile,showcase}-bridge-stub/scripts/query.py + skill-patch.md 3 skill 預製，等 CF Tunnel NousResearch#12-13 上線後 5–10 min 採納 3. ck-observability-bridge-skeleton/scripts/query.py + skill-patch.md multi-backend 設計（Loki/Prom/Grafana/Alertmanager），9 actions 4. hermes-integration-final-report.md：12 輪 wrap-up 一頁總覽 5. hermes-acceptance-sop.md：採納後 e2e 驗收（hermes-web :9119 為主入口） 6. commit-organization.md：33 untracked → 6 batches commit 指引驗證：missive 端到端通過 hermes runtime 75s 取真實業務 query (rag_search 5 公文)。採納路徑：業務 repo session: bash docs/plans/skill-helper-template/adopt.sh <skill> . Refs: hermes-runtime-blockers-postmortem.md (7 層真因 — 下一 commit) hermes-integration-final-report.md (整體狀態)

per ADR-0024 (proposed)： - 3 tools functional: lvrland_health / lvrland_query_sync / lvrland_price_trends - ACTION_HANDLERS 對應 LvrLand /api/v1/ai/query / /api/v1/analytics/price-volume-trends - tests: 12/12 ✅ (含 stub_http fixture / fallback / error paths / register_all) - + scripts/query.py helper (Phase 1.5 採納用，待 CF Tunnel NousResearch#12) - + skill-patch.md：CK_lvrland session 採納步驟接續 ck-missive-bridge / ck-observability-bridge / ck-showcase-bridge / ck-pilemgmt-bridge 完成 ADR-0020 Phase 1 4 bridge skill 矩陣。 Refs: hermes-agent docs/plans/ck-lvrland-bridge-stub/SKILL.md

* ci(nix): auto-fix stale npm hashes on push to main When a PR merges to main with updated package-lock.json or package.json in ui-tui/ or web/, the new auto-fix-main job detects stale npmDepsHash values and pushes a fix commit directly to main. This eliminates the recurring manual hash-bump PRs (NousResearch#15420, NousResearch#15314, NousResearch#15272, NousResearch#15244) by reusing the existing fix-lockfiles --apply pipeline. The fix commit only touches nix/*.nix files, which are outside the push path filter (package-lock.json / package.json), so it cannot re-trigger itself. Closes NousResearch#15314 * fix(ci): use GitHub App token for auto-fix-main push GITHUB_TOKEN commits are invisible to workflow triggers (GitHub's infinite-loop prevention). The auto-fix-main job pushes directly to main, so the fix commit never triggered downstream nix.yml verification. Mint a short-lived token via the repo's GitHub App (daimon-nous, APP_ID + APP_PRIVATE_KEY secrets) so the push is treated as a real event and nix.yml fires to verify the corrected hashes. Tested via workflow_dispatch dry-run: app token minted successfully, checkout with app token succeeded, fix job correctly gated. Resolves review feedback from Bugbot (r3144569551). * ci(nix): rename lockfile check job for required status check Rename 'check' → 'nix-lockfile-check' so the status check name is unambiguous when added as a required check on main. * fix(ci): harden auto-fix-main against races, loops, and silent failures Address adversarial review findings: 1. Race condition (#1): Job-level concurrency with cancel-in-progress collapses back-to-back pushes; ref: main checkout always gets latest branch state; explicit push target (origin HEAD:main). 2. Loop prevention (#2): File-whitelist check before commit aborts if any file outside nix/{tui,web}.nix was modified, preventing accidental self-triggering. 3. Silent infra failures (NousResearch#8): nix-lockfile-check now fails explicitly when fix-lockfiles exits without reporting stale status (catches nix setup failures, network errors, script bugs that bypass continue-on-error). 4. Commit traceability (NousResearch#11): Auto-fix commits include source SHA and workflow run URL in the commit body. 5. Explicit push target (NousResearch#12): git push origin HEAD:main instead of bare git push. --------- Co-authored-by: alt-glitch <alt-glitch@users.noreply.github.com>

- NousResearch#1 客户端路径硬编码: 改为自动搜索常见位置+GEO_CLIENT_EXE环境变量覆盖 - NousResearch#9 cmd_status误用POST /api/stop探测: 改为无副作用的GET，不再杀任务 - NousResearch#11 push/ai_push的uid为空不拦截: 加前置检查，uid/udid为空时报错退出 - NousResearch#12 表格数据翻倍: querySelectorAll不再匹配.fixed-table-body的克隆tbody

…s, full-width chat, context meter, perf work Squash-merge of wip/desktop-categories-20260718 (72fe84f) onto main plus the purple/hardening-only items, reworked where upstream superseded the original implementation. Renderer + electron only; the context meter needed no code (current main is field-aligned with the backend — it was blank due to zeroed in-process counters + provider 400s, not drift). Restored from wip/desktop-categories-20260718: - Session categories: $sessionCategories persistent store + sanitizer + create/rename/delete/collapse/move (store/layout.ts), sidebar Categories section (session-categories-section.tsx), category membership retention in sessionsToKeep, sidebar mount + shared store subscription (index.tsx), sidebar.categories.* i18n keys (en/ja/zh/zh-hant/types), and the sanitize/move unit tests (session-categories.test.ts). - Category drag-and-drop REWORKED for upstream's pointer drag session: new category-drop-zones.ts registry (live-rect hit test), CategoryDropZone stripped of the dead native-HTML5 half, and session-drag.ts now targets registered category zones (highlight + file-on-release). The pre-merge session-pointer-drag.ts fallback and inline-refs effectAllowed patch (CRITICAL NousResearch#16) are superseded by upstream's non-native drag and were not ported; upstream's suppressDragClick covers the phantom-click concern. - System tray (Win/Linux): Tray with Open/Restart/Quit, double-click restore, close-to-tray, --hidden boot-to-tray for the login autostart, tray-aware window-all-closed, teardown on real quit (electron/main.ts). Tray Restart rewired to backendConnectionState.invalidate(). - CV-011 audio timeouts: /api/audio/* 120s ceiling in the hermes:api IPC and transcribeAudio/speakText timeoutMs (2026-07-18 discarded-transcription incident). - Sidebar shared-scroll wheel-dead-zone fix: one shared scroll container (SCROLL_Y without overscroll-contain + scrollbar-gutter:stable), getScrollElement plumbing, virtual list scrollMargin/initialOffset/gap handling with VirtualRow memo boundary (showProfileTags preserved), cron section loses overscroll-contain. - Timeline perf: PROMPT_HARVEST_MAX cap, MIN_ENTRIES effect guard, single querySelectorAll byId map per compute. - styles.css compositor perf: transform-only arc-border sweep on a 300% ::before, paused arcs behind the closed sidebar pane, opacity-only code-card streaming pulse on an inset ::after (+ reduced-motion hide). - getSessionMessages 60s timeout + hermes.test.ts expectation (RCA 2026-07-09). - 50-WHYS-SESSION-LATENCY.md (docs). Restored from purple/hardening: - Full-width chat: --composer-width 100% + tight aui_thread-content gutters (6610204; the 7/14 rebuild regressed to the narrow column). - session.create/session.resume 120s ACK budgets (63.1s GIL-stall incident), threaded through requestGateway timeoutMs at both create sites + resume. - toChatMessagesAsync cooperative transcript conversion (yield per 150 messages) + O(1) tool-result backfill index (chat-messages.ts builder refactor); cold-resume prefetch now converts cooperatively and hands the result to reconcileAuthoritativeMessages. - Trailing-edge debounce (2.5s) for the per-turn cron-jobs refresh fan-out (the cron/messaging session halves are superseded by upstream's batched listSidebarSessions call). - Immediate-first gateway reconnect (delay 0 attempt 0) with 2s open-then-die flap guard (primary) and the same backoff shape for secondary gateways. - update-marker onStale breadcrumb (CRITICAL NousResearch#2 / audit S-4): stale .hermes-update-in-progress self-heal is now logged. - stableRepoPaths content-stable identity so project-tree refreshes stop respawning git worktree list per repo (perf finding 2026-07-09). - sameCronJobs signature guard so the 30s jobs poll stops re-rendering the sidebar on identical data. - Markdown EXACT_CACHE_MAX 64 -> 512 for multi-session switch-back (applied to upstream's two-tier exact+append design). - backend-ready READY-token WHY comment (CRITICAL NousResearch#6/NousResearch#9) — the regex itself was already upstreamed. - Codicon ellipsis -> kebab-vertical in env-var-actions-menu and assistant-message (consistency with session rows). Deliberately dropped (superseded by main): session-pointer-drag.ts, inline-refs writeSessionDrag/effectAllowed, wip session-row drag wiring, per-id tool diff accessor + fallback subscription, use-session-state-cache bleed guard, cheap part comparator, tail-first resume, --composer-width JSX changes (none needed), restart-counter.cjs. Deliberately dropped (out of scope here): Clear Idle Sessions (FEATURE NousResearch#8 / CRITICAL NousResearch#12) — its backend half (tui_gateway session.release_idle) is a Python change; port both halves together in a follow-up. Non-desktop wip work (hermes_cli/web_server.py conversoar provider, discord adapter thread gates, root package.json, test_web_server.py) left on the wip branch. Gates: tsc renderer+electron clean; eslint 0 errors on all 31 changed TS files; vitest 72/72 targeted (categories/hermes/chat-messages/markdown- blocks/sidebar) + 20/20 gateway-boot/list-actions/state-cache + electron 455 passed / UI 1619 passed — remaining failures (2 electron POSIX-path, 5 messaging/skills) reproduce identically on pristine main.

* ci(nix): auto-fix stale npm hashes on push to main When a PR merges to main with updated package-lock.json or package.json in ui-tui/ or web/, the new auto-fix-main job detects stale npmDepsHash values and pushes a fix commit directly to main. This eliminates the recurring manual hash-bump PRs (NousResearch#15420, NousResearch#15314, NousResearch#15272, NousResearch#15244) by reusing the existing fix-lockfiles --apply pipeline. The fix commit only touches nix/*.nix files, which are outside the push path filter (package-lock.json / package.json), so it cannot re-trigger itself. Closes NousResearch#15314 * fix(ci): use GitHub App token for auto-fix-main push GITHUB_TOKEN commits are invisible to workflow triggers (GitHub's infinite-loop prevention). The auto-fix-main job pushes directly to main, so the fix commit never triggered downstream nix.yml verification. Mint a short-lived token via the repo's GitHub App (daimon-nous, APP_ID + APP_PRIVATE_KEY secrets) so the push is treated as a real event and nix.yml fires to verify the corrected hashes. Tested via workflow_dispatch dry-run: app token minted successfully, checkout with app token succeeded, fix job correctly gated. Resolves review feedback from Bugbot (r3144569551). * ci(nix): rename lockfile check job for required status check Rename 'check' → 'nix-lockfile-check' so the status check name is unambiguous when added as a required check on main. * fix(ci): harden auto-fix-main against races, loops, and silent failures Address adversarial review findings: 1. Race condition (NousResearch#1): Job-level concurrency with cancel-in-progress collapses back-to-back pushes; ref: main checkout always gets latest branch state; explicit push target (origin HEAD:main). 2. Loop prevention (NousResearch#2): File-whitelist check before commit aborts if any file outside nix/{tui,web}.nix was modified, preventing accidental self-triggering. 3. Silent infra failures (NousResearch#8): nix-lockfile-check now fails explicitly when fix-lockfiles exits without reporting stale status (catches nix setup failures, network errors, script bugs that bypass continue-on-error). 4. Commit traceability (NousResearch#11): Auto-fix commits include source SHA and workflow run URL in the commit body. 5. Explicit push target (NousResearch#12): git push origin HEAD:main instead of bare git push. --------- Co-authored-by: alt-glitch <alt-glitch@users.noreply.github.com>

Findings #1 and NousResearch#2, together, because NousResearch#2 makes #1 bypassable: a gate in the model-tools path is worth nothing while `registry.dispatch()` reaches handlers without consulting it. `resolve()` now has a production caller. A non-AUTO tool cannot execute without an execution capability, and a capability cannot exist without a decision — not "approved" as a boolean somebody might set early, but a token minted at consent, bound to the exact call, and destroyed by being used. Four properties, each with a test that fails without it. Bound to the tool and an argument fingerprint, so approving one call cannot execute a different one and mutating arguments between consent and execution invalidates the token. One-use, so a retry loop cannot turn one approval into several executions. Short-lived, because consent goes stale. And `consume()` raises rather than returning a boolean — a caller that forgets to check a boolean executes anyway, and this is the one check where forgetting must not be survivable. Fail-closed throughout. A broken permission lookup refuses; `requires_capability` returns True on any exception; an unreadable trust list trusts nothing; minting without a `tool_call_id` is refused outright, because a capability that could match any call is not a capability. Every refusal test asserts the handler ran **zero** times — a gate that refuses after the side effect is not a gate. **The default is `observe`, and that is a gap rather than a preference.** `get_tier()` returns ALWAYS_APPROVAL for unregistered tools — deliberately, so an unknown tool is never assumed safe — and most tools here were never registered. Switching straight to `enforce` would refuse nearly every call in the product, which is an outage, not a safety improvement. `observe` audits every call that *would* be refused, with its tool name, so the registration backlog is measurable against real traffic and the switch can be flipped once the audit goes quiet. This mirrors HERMES_APPROVAL_INTEGRITY_MODE, which exists for the same reason; both must reach `enforce` for the tier system to mean anything, and neither is there yet. Enforce mode is fully tested. Also corrected, per the review: the baseline-failure classification, which I had called environmental wholesale. Three of them are real defects at baseline — `atomic_config_write` NameErrors, a SQLite journal-mode assertion, and a jobs asset-contract failure — and the evidence now says so per failure. And the rollback procedure, which claimed one merge revert covered a many-commit range; it now gives a non-destructive runtime pin plus three repository options, with the history-rewriting one marked as needing explicit approval. Regression check: tests/tools 8532 passed / 19 failed — the failing set byte-identical to baseline, so the gate broke nothing. Still not merged to main. Remaining findings: NousResearch#6 undo wiring, NousResearch#10 Now composition, NousResearch#11 native resume, NousResearch#12 New Chat, NousResearch#13 free-form clarify, NousResearch#14 approval acknowledgement, NousResearch#15 reconnect, NousResearch#16 readiness coupling, NousResearch#17 sensor delivery. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Nu2Qaq5Y7EScuooGz8co34

…, coupling **NousResearch#14 — a card could resolve on a decision the gateway never recorded.** `approval.respond` answers with how many approvals it resolved, and zero is a *successful RPC that decided nothing* — the request expired, or was answered elsewhere. The client read the envelope and closed the card, so the owner believed they had answered while the agent stayed blocked. Now only a `resolved >= 1` resolves it, and only an explicit `status: "ok"` resolves a clarify. **NousResearch#13 — free-form clarify answers went out as new prompts.** The native path had a steering step that was a no-op returning true, after which the text was sent via `prompt.submit`: the question stayed open and the answer arrived as an unrelated message. Replaced by one `answerClarifyFreeText` that addresses the request by id natively and walks the menu on the terminal — one method, because splitting it into "steer, then send" is what let the two halves disagree. **NousResearch#11 — resume threw the transcript away.** `session.resume` returns `messages`; `open()` read only the identity fields. A refresh reattached to a running session and showed an empty feed, so execution continued against a conversation the owner could no longer see. History is now delivered through an `onHistory` hook that fires *before* `subscribe()` — ordering is the other half of the fix, since applying the past after live deltas have landed interleaves them — and merges through the existing hydration path rather than a second one. **NousResearch#12 — New Chat did nothing natively.** Both controls rotated a PTY identity that native mode does not use. The session stayed open, its durable id stayed in storage, and the next prompt continued the old conversation under a heading that said it was new. `startNew()` closes the session, clears the profile-scoped durable id before the rebuild so the effect cannot read a stale one, drops the live id immediately, and opens fresh without resuming. **NousResearch#16 — a missing TUI disabled the transport that exists to avoid the TUI.** The chat-readiness probe asks whether the *TUI* can start, and its verdict also gated native session creation. Each transport is now gated on its own prerequisite, and the probe does not run at all under native chat — the banner it raises describes a dependency native chat does not have. Gates: typecheck 0, eslint 0, 74 files / 706 frontend tests, build green. Remaining from the review: NousResearch#6 undo not wired to production, NousResearch#10 Now not consuming the ranking modules, NousResearch#15 reconnect, NousResearch#17 sensor delivery. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Nu2Qaq5Y7EScuooGz8co34

Phase 4 实施期 K-10 (per sprint plan §5). Sprint 1 (调研期 + Step 0 + Step 1) 已 push (commit 9a38e9a + 613ce0f + b0c61dc + 943c1f7 + 50f8f22, tag v0.18.0+cn.5). verify 后端跟 plan 假设 differ 1-2 层 (跟 CAND-084 8-03 22:10 lesson 1:1): - plan 假设 "加 context_compression.threshold config" → 实际 cli.py:339 已有 `"threshold": 0.50` + agent_init.py:1231 读 + :1489 传 → 0 work (1:1 集成已 mature) - plan 假设 "3 cherry-pick 7f670a0 + 5f84eac + 2d0e96a" → 3 commit 都已在 cn (5 branch contains) → 0 cherry-pick work - plan 假设 "smoke test 长 session 跑通 max_turns 500" → 实际 1 line CN change + 4 静态 source check test K-10 actual scope (跟 plan spirit 1:1 配对, scope adjust 1d → 0.25d): 1. cli.py:342 "max_turns": 90 → 500 (1 line change, 跟 compression.threshold 0.50 1:1 配对) 2. tests/test_k10_max_turns_and_compression.py 4 静态 source check (跟 test_split_bugs_prevent_regression 12 个 split bug 同 pattern: AST-style 静态 check, 0 yaml 依赖, 防改回归): - test_cli_max_turns_default_500 (K-10 main change) - test_cli_compression_threshold_050 (config 集成 regression guard) - test_agent_init_reads_compression_threshold_from_config (集成点 1:1 配对) - test_max_tail_message_floor_still_8 (cn-specific fix aaa3ee6 regression guard, 跟 test_split_bugs_prevent_regression NousResearch#12 1:1 配对) 验证: - 4/4 K-10 test pass (本地) - 12/12 split bugs regression test pass (本地 baseline 0 降) - grep 旧名 0 命中 (无 split bug) - additive 0 改旧 (cli.py 1 line change, agent_init.py / context_compressor.py 0 改) - K-9 跟 AIMC 冲突风险 high 推 next sprint (跟 plan §13 风险登记 1:1 配对) 跟 mavis 4 件套 1:1 配对: - 后端先调查再设计: verify 5 候选后端 mature, K-10 scope 0.25d adjust - Cherry-pick split bug class: 1 commit 0 split, _MAX_TAIL_MESSAGE_FLOOR regression guard - UX 倒退审计: 现有 happy path 不变, 仅 default 90→500 - 估时前必 verify 引擎能力: K-10 verify 后端发现 plan 假设 1d → 0.25d 差 4x, 跟 8-03 22:10 lesson 1:1 跟 AIMC 4 铁律 1:1: - 不反向调整 (CN 改 default 不动 upstream config pattern) - 不写回 upstream (0 push 跟 8-05 阶段批推 1:1, 4 commit 一起 batch) - 0 corrupt (1 line change, regression guard 防回退) - fail-fast (跟 batch push + CI 1:1 配对, K-10 commit 后 CI 3 job 必绿) NO push default (per plan §5.2 step 10 "STOP: 不 push origin, 留 Step 5 阶段批推"). Push 跟 K-6/K-7/K-8 3 commit 一起 batch 推 origin/cn, 收尾 tag v0.18.0+cn.6.

…cycle Audit lần 2 (15 findings): - NousResearch#2 P0: write_entry reset attempts=0 when re-emitting a FAILED entry, so MAX_ATTEMPTS dead-letter never fired. Merge now KEEPS attempts for any non-terminal existing state. - NousResearch#3 P1: TOCTOU idempotency — find_by_idempotency_key (unlocked scan) and write_entry were separate steps; two concurrent emits with the same key both dispatched (reproduced 2x adapter.send). New write_entry_idempotent() checks + writes under one global .dedup.lock, skipping when an entry with the same key exists under a DIFFERENT envelope_id (terminal or in-flight). - NousResearch#4 P1: stop() on a non-running watcher clobbered watcher_state.json to active=false while the real holder still polled. stop() now only persists when actually running. - NousResearch#9 P2: re-emit hardcoded kind=background_notify, dropping original kind. - NousResearch#12 P2: install_outbox_watcher's 'will start when loop runs' never re-armed — now re-arms on the next loop via call_soon.

…lStorage persistence, gateway crash pitfalls Added pitfalls NousResearch#11-13 from real debugging sessions: - NousResearch#11: cards omitting session_id in tools/call → host auto-injects - NousResearch#12: HMR reload wipes in-memory cache → localStorage persistence - NousResearch#13: gateway crash leaves Electron alive → silent message loss Also updated file map (chat-messages.ts, contrib/wiring.tsx), added ui/initialize handshake docs, and expanded verification checklist.

hjc-puro added 6 commits November 15, 2025 00:01

add tracking for cluster failurse

31c7333

add simple terminal

f813959

rate limits

1614c15

don't log exit code !=0 as terminal failure

ab7293b

add linewise profiling

349e37d

add profiling

e06a15b

hjc-puro changed the title ~~[merge simple terminal first] Profiling~~ [merge after #11] Profiling Nov 20, 2025

teknium1 closed this Feb 2, 2026

SHL0MS mentioned this pull request Mar 30, 2026

[UX] Argument-level tab completion for /resume, /skin, /personality, etc. #4059

Open

h4x3rotab pushed a commit to Clawdi-AI/hermes-agent that referenced this pull request Apr 10, 2026

feat: 6-column Kanban board with drag-and-drop (NousResearch#12)

e5c11a5

h4x3rotab pushed a commit to Clawdi-AI/hermes-agent that referenced this pull request Apr 10, 2026

Merge pull request NousResearch#12 from kcgcom/polish/chat-composer-s…

368e5e2

…urface Clean CSS polish, no conflicts.

kshitijk4poor mentioned this pull request Apr 10, 2026

feat(browser): expose remaining agent-browser commands #7336

Open

malaiwah pushed a commit to malaiwah/hermes-agent that referenced this pull request Apr 11, 2026

Merge pull request 'fix(delegate): workspace_path → workspace_note kw…

2d8fe90

…arg mismatch' (NousResearch#12) from fix/delegate-workspace-kwarg into main

kshitijk4poor mentioned this pull request Apr 15, 2026

feat: add TouchDesigner integration skill (twozero MCP) #10081

Closed

kskotetsu mentioned this pull request Apr 18, 2026

MMC3 scanline IRQ実装バグ (#12テストがFAIL) #12147

Closed

alaamohanad169-ship-it mentioned this pull request May 8, 2026

Add profiling module (fix for closed PR #12) #22040

Closed

Julientalbot mentioned this pull request May 8, 2026

feat(run_agent): opt-in prune of historical reasoning_content beyond N turns #22068

Draft

Fede654 mentioned this pull request May 10, 2026

feat(embodied_plan): in-intent narrative-replan composer + daemoncraft-base profile rules aligned with Autonomía Corporal nicoechaniz/hermes-agent#6

Open

5 tasks

cristianmgm7 mentioned this pull request Jun 10, 2026

feat(platforms): add Carbon Voice as a platform plugin #43226

Open

zenc-cp mentioned this pull request Jun 11, 2026

feat(adr-032): observability sink — consumer hop zenc-cp/hermes-agent#6

Merged

5 tasks

This was referenced Jun 27, 2026

fix: clear ended_at metadata when messages are appended to an ended session nicha16/hermes-agent#1

Closed

fix: clear ended_at metadata when messages are appended to an ended session #53761

Closed

jamesraddock mentioned this pull request Jul 12, 2026

[Bug]: Kanban check_respawn_guard mis-fires blocker_auth on workspace/spawn errors containing an auth/quota slug word → task wedged in 'ready' forever #63248

Open

zapabob mentioned this pull request Jul 21, 2026

main #68353

Closed

Sensenkawa mentioned this pull request Jul 30, 2026

Bug: providers vs custom_providers dual storage causes CLI/GUI mismatch + model version stuck in profile #71298

Open

sycamoregroupltd mentioned this pull request Aug 4, 2026

fix(kanban): gate auto-decomposer past block_loop_detected cards (t_15b7ebc4 vector 2) #78895

Open

andrexibiza mentioned this pull request Aug 6, 2026

[COMPLETE] Large-file decomposition: 20/20 done #78647

Closed

andrexibiza mentioned this pull request Aug 10, 2026

Shard tools/mcp_tool.py (god-file decomposition) #78642

Open

zapabob mentioned this pull request Aug 12, 2026

main #84448

Closed

victor-kyriazakos mentioned this pull request Aug 16, 2026

feat(relay): live-card ops — native draft streaming + task cards over the relay (gateway half) #85796

Open

Navigation Menu

Uh oh!

[merge after #11] Profiling - #12

[merge after #11] Profiling#12
hjc-puro wants to merge 6 commits into
mainfrom
profiling

hjc-puro commented Nov 18, 2025 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

hjc-puro commented Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

hjc-puro commented Nov 18, 2025 •

edited

Loading