HERMES-GIELIE/api
nesquena-hermes 74dcc4fe6f
Release v0.51.237 — Release HE (stage-q8): reconcile early-cancel against live worker state (#3492)
## Release v0.51.237 — Release HE (stage-q8)

Phase 3 MEDIUM-ring pick (3-factor framework): a real concurrency/state-consistency fix from a regular contributor (@franksong2702, ★★★ 38 merges). The hardest review of the sweep — the dual gate caught **two** silent data-loss bugs across two review rounds.

### Fixed
| PR | Author | Fix |
|----|--------|-----|
| #3475 | @franksong2702 | Cancelling a live turn immediately after send now reliably stops the worker and settles the session to a cancelled state (was: spinner over a blank page). `cancel_stream()` falls back to the live active-run registry (`ACTIVE_RUNS`) + session agent cache when `STREAMS` has already detached, so the worker still receives `interrupt("Cancelled by user")`. `/api/session` reports run-journal active state from the live registry instead of trusting a persisted `active_stream_id`. |

### Pre-release dual gate caught TWO silent data-loss bugs (both fixed + regression-tested)
The PR's refactor moved `agent.interrupt()` ahead of the partial-text/reasoning/tool-call snapshot, and that snapshot was no longer under `streams_lock`:
1. **Codex round 1** — the worker's `finally` (which pops `STREAM_PARTIAL_TEXT`/`STREAM_REASONING_TEXT`/`STREAM_LIVE_TOOL_CALLS` under `STREAMS_LOCK`) could clear those buffers the instant `interrupt()` wakes it, so a cancelled turn **silently lost its already-streamed text**. (Opus reviewed the original and said ship — it assumed the snapshot was still lock-protected; the stale comment claimed so but the code wasn't. Verified against the actual code → Codex was right.)
2. **Codex round 2** — my first fix only snapshotted on the `STREAMS`-present path; the detached `ACTIVE_RUNS`-only path (the case this PR adds) still lost text. Fixed by hoisting the snapshot above the `if stream_present` branch so it runs unconditionally under the lock.

Both fixes have regression tests **verified to fail against the buggy versions** (`test_cancel_preserves_partial_text_when_interrupt_pops_buffers` + `test_cancel_preserves_partial_text_on_detached_active_run_path`).

### Gate results
- **Full pytest suite**: 7481 passed, 9 skipped, 3 xpassed, **0 failed**
- **ruff forward gate**: CLEAN
- **browser-smoke gate**: CLEAN
- **Codex (regression)**: SHIP ONLY WITH FIXES (×2 rounds) → both MUST-FIXes applied → re-reviewed **SAFE TO SHIP**
- **Opus (correctness)**: reviewed the original (SAFE); the shipped version is strictly safer (adds the under-lock snapshot Opus deemed unnecessary)
- Deadlock concern cleared: no path takes `ACTIVE_RUNS_LOCK` then `STREAMS_LOCK`.

Rebased onto current master (streaming.py/routes.py merged clean — no overlap with the #3468 dedup change shipped in HC).

Closes #3475.

Co-authored-by: franksong2702 <franksong2702@users.noreply.github.com>
2026-06-03 11:26:12 -07:00
..
__init__.py
agent_health.py Release v0.51.225 — Release GS (stage-p7 — remote gateway health probe resolves gateway_state #3355) (#3458) 2026-06-02 18:56:25 -07:00
agent_sessions.py fix: keep webui mirrored sessions out of cli filter 2026-05-28 00:51:12 +02:00
auth.py fix: harden WebUI bugfix sweep 2026-05-28 13:38:50 -04:00
background.py feat(commands): /background, /btw slash commands + undo button + reasoning chip 2026-04-24 01:24:51 +00:00
clarify.py feat: add distinct attention request sound 2026-05-30 19:40:40 +00:00
commands.py fix(security): drop unsafe-eval + add jsdelivr to CSP, sanitize plugin error 2026-05-11 17:53:02 +00:00
compression_anchor.py fix(sidebar): keep newer continuation visible over older snapshot 2026-05-28 13:38:06 -06:00
config.py Release v0.51.228 — Release GV (stage-p12 — workspace tree-drop #3411 + large-markdown preview #3378) (#3463) 2026-06-02 20:58:22 -07:00
dashboard_probe.py fix: preserve IPv6 dashboard link brackets 2026-05-19 21:35:10 -04:00
extensions.py Opus pre-release follow-ups for PR #1445 2026-05-02 03:49:40 +00:00
gateway_chat.py fix(gateway): carry WebUI delivery context on gateway-backed path too (Codex MUST-FIX on #3324) 2026-06-02 19:42:14 +00:00
gateway_watcher.py release: v0.50.241 (#1293) 2026-04-29 19:54:07 -07:00
goals.py fix(i18n): localize /goal runtime status strings 2026-05-10 15:21:24 +08:00
helpers.py fix(todos): hydrate cold-load state from session snapshot 2026-06-02 17:33:49 +00:00
kanban_bridge.py perf(http): enable HTTP/1.1 keep-alive 2026-05-24 18:26:56 +00:00
metering.py fix: show TPS in assistant message headers 2026-05-04 21:26:43 +00:00
models.py Release v0.51.222 — Release GP (stage-p4 — backend bugfix batch: title language drift #3293 + orphaned CLI sidecar prune #3238 + pin-quota lineage #3288) (#3452) 2026-06-02 17:35:18 -07:00
oauth.py fix(oauth): serialize Anthropic env fallback reads 2026-05-07 02:47:19 +00:00
onboarding.py Classify onboarding DNS probe failures consistently 2026-05-15 08:51:30 +08:00
passkeys.py feat: support passkey-only auth 2026-05-25 00:14:38 +00:00
paths.py fix(profiles): avoid config import cycle 2026-06-01 05:05:16 +00:00
plugins.py fix(#2622): plugin card UX — legible Open button, single badge, visible toggle + reject protocol-relative tab.path 2026-06-02 03:58:02 +00:00
profiles.py fix(profiles): TOCTOU-safe _SKILLS_STATS_CACHE read via .get() (Opus pre-release nit on #3341) 2026-06-01 18:38:39 +00:00
providers.py fix: honor shared OpenCode key at runtime 2026-05-30 02:18:06 +00:00
request_diagnostics.py fix: add request wedge diagnostics 2026-05-08 15:37:08 +00:00
rollback.py v0.50.255: Opus follow-ups (4 fixes) + CHANGELOG 2026-05-01 17:19:53 +00:00
routes.py Release v0.51.237 — Release HE (stage-q8): reconcile early-cancel against live worker state (#3492) 2026-06-03 11:26:12 -07:00
run_journal.py fix: filter interrupted recovery control text from visible transcript (#3321) 2026-06-01 20:07:16 +00:00
runner_client.py harden: runner client rejects non-http(s) base_url + no redirect-follow (#3073) 2026-05-31 17:49:26 +00:00
runtime_adapter.py Stage 398: PR #2696 — feat(runtime): add runner-local adapter selection (RuntimeAdapter slice 4c, feature-flagged) 2026-05-21 17:43:54 +00:00
session_discoverability.py Merge PR #3048 2026-05-28 17:47:33 +00:00
session_events.py sync session list from server events 2026-05-20 08:18:56 -06:00
session_lifecycle.py fix: align WebUI memory commits with CLI boundaries 2026-05-17 05:04:57 -06:00
session_ops.py fix(session): preserve sidecar truncation boundary 2026-05-25 21:21:15 +08:00
session_recovery.py fix(session): rebuild missing startup index 2026-05-20 23:43:30 +08:00
skill_usage.py fix: remove webui-side writer to avoid conflict with agent 2026-05-27 20:57:32 +08:00
startup.py fix: P0 hotfixes — API regression, code block parser, chmod override 2026-05-01 12:10:48 +00:00
state_sync.py fix: use Optional[T] for default None parameters in state_sync.py 2026-06-01 17:20:17 +00:00
streaming.py Release v0.51.237 — Release HE (stage-q8): reconcile early-cancel against live worker state (#3492) 2026-06-03 11:26:12 -07:00
system_health.py feat: add VPS resource health panel 2026-05-05 17:30:56 +00:00
terminal.py Release v0.51.236 — Release HD (stage-q7) (#3491) 2026-06-03 10:46:06 -07:00
todo_state.py fix(todos): hydrate cold-load state from session snapshot 2026-06-02 17:33:49 +00:00
turn_journal.py fix: handle Windows media and turn journal edge cases 2026-05-30 11:15:35 +02:00
updates.py fix(updates): re-exec correctly in both source and frozen builds 2026-06-02 17:33:49 +00:00
upload.py Release v0.51.221 — Release GO (stage-p3e — block all workspace symlink escapes + portable TOCTOU hardening [security]) (#3398) (#3451) 2026-06-02 16:56:30 -07:00
usage.py fix: display canonical cache hit percentage 2026-05-19 02:27:12 -06:00
workspace.py Release v0.51.234 — Release HB (stage-q4) (#3488) 2026-06-03 09:53:32 -07:00
workspace_git.py fix(workspace): tighten git subprocess trust boundary 2026-05-20 11:02:45 +00:00
worktrees.py Harden worktree removal safeguards 2026-05-13 09:49:15 +08:00