Skip to content

fix(autopilot): reconcile stage timeouts - #13992

Closed
junjiequ wants to merge 2 commits into
kirodotdev:mainfrom
junjiequ:fix/go-all-timeout-reconcile
Closed

junjiequ wants to merge 2 commits into
kirodotdev:mainfrom
junjiequ:fix/go-all-timeout-reconcile

Conversation

@junjiequ

@junjiequ junjiequ commented Sep 25, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Four independent Autopilot Go-All sessions hit the same 30-minute card while actively editing, testing, or monitoring work. The implementation used orchestrator.stage_timeout_seconds as a hard deadline around the running agent turn, even though the bundled Autopilot contract says it gates whether another turn may start and an admitted stage may outlast it.

The timeout aftermath also split ownership: KiroCrew cancelled the dashboard coroutine without reliably cancelling the native ACP turn, preserved a stage boundary that trapped ordinary guidance, and exposed no clear recovery controls.

Why it matters

Normal productive stages were classified as runaway work and stopped at exactly 30 minutes. In two traced reproductions, the native ACP turn continued after the dashboard claimed it stopped: one completed a build and another began additional file writes. Late output became detached from the dashboard, temporary evidence could be reclaimed, and user guidance queued behind the preserved boundary.

What changed (motivation → approach → change)

The timeout model now has separate responsibilities:

  • orchestrator.stage_timeout_seconds is an unattended start gate. A productive stage may finish after 30 minutes; Go-All pauses before admitting the next stage and accurately names both the stage that spent the budget and the pending stage.
  • A human-approved Go is never rejected by the unattended budget.
  • Running stage turns use the ordinary agent.chat_turn_timeout_secs ceiling (four hours by default), while existing ACP liveness controls continue to handle model silence and in-flight tool stalls.
  • Cancelling an acquired dashboard turn sends native ACP session/cancel, waits up to agent.soft_stop_budget_secs for acknowledgement, and retains the cancellation task across a repeated outer cancellation.
  • A coroutine cancelled before acquiring the shared session lease does not cancel by key, preventing it from stopping another turn that owns the lease.
  • An unacknowledged native stop produces a visible warning instead of silently claiming the previous tool is gone.
  • The exact interrupted-stage boundary remains resumable, ordinary guidance is admitted while paused, destructive-history protections remain armed, and the timeout card offers Go / Cancel.

This restores the documented Autopilot contract instead of increasing the 30-minute value or removing runaway protection.

Backwards compatibility

Compatible. Existing configuration keys and defaults remain unchanged. Go-All still stops before starting more work once a stage spends its budget. Explicit Go, failure limits, round limits, whole-plan limits, session leases, and provider liveness controls remain in force.

Tests

Added or updated regressions that prove:

  • productive stage work may outlast the 30-minute start budget;
  • a spent budget pauses before the next automatic stage and is attributed to the correct stage;
  • human-approved Go bypasses the unattended start gate;
  • the independent chat-turn ceiling still stops a genuinely runaway stage;
  • pre-lease cancellation cannot stop another turn on the same folded key;
  • acquired cancellation waits for acknowledgement;
  • unacknowledged cancellation is visible;
  • repeated cancellation retains native cancellation and partial output;
  • timeout guidance, retry-boundary, clock-reset, held-note, and decision-strip behavior remains intact.

Validation on afbb21042:

  • 185 adjacent timeout, tracker, stage-loop, cancellation, note, decision-strip, and config-baseline tests passed;
  • Black, isort, flake8, documentation lint, git diff --check, and mypy across 1,725 source files passed;
  • the branch includes current upstream main without rewriting the published first commit.

The diff-scoped backend gate collected 34,962 tests in its first batch. It found the config-baseline drift introduced by the updated setting help text, which is fixed and passes in isolation. Its other failure is the unchanged Code Review Sage symlink-race test: the fixture requires a pytest temporary directory outside every managed root, while this environment sets TMPDIR under KiroCrew's managed scratch root. The same isolated node fails without this feature diff.

The first published CI round also found a synchronous cancel_current test double on Linux and Windows; that fixture now uses the async contract and passes locally. One unrelated fairness timing assertion missed its <0.1s threshold at 0.198s.

Manual verification

Observed the premature timeout signature in four independent local dashboard sessions. Two were traced through both dashboard and native ACP records, proving the native turn continued after the dashboard timeout. The two additional screenshots show productive file edits immediately before the same Stage 1 timeout card.

A gateway running this revision is still needed for the final visual capture of the corrected boundary card.

Screenshots / video

The reported screenshots show the old failure state. The fixed live card will be captured once a gateway runs this revision; the backend and transcript behavior are covered by the regressions above.

Related Issues

  • Required RFC: #14043
  • Status: this implementation remains draft until the RFC is accepted.
  • No public issue was filed; four live local reproductions led directly to the RFC.

Pattern harvest

Rule candidate: review-prompt
Pattern: a workflow budget that governs admission must be checked between units of work, not reused as the cancellation deadline for an already-running unit. Native cancellation must be scoped to the lease owner and its acknowledgement must be visible.

Checklist

  • At most two commits, with Conventional Commits titles
  • Existing adjacent tests pass and new regressions cover the behavior
  • Self-review completed; code follows project style guidelines
  • Documentation and generated config baseline updated
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

N/A — repository placeholder retained; no CLA text invented.

@dwu96

dwu96 commented Sep 25, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@junjiequ
junjiequ marked this pull request as ready for review September 25, 2026 22:04
@junjiequ
junjiequ requested review from a team and krishdhasmana September 25, 2026 22:04
@github-actions github-actions Bot added fork Pull request from a fork (external contributor) readiness: checking Automated validation is still running labels Sep 25, 2026
@github-actions

github-actions Bot commented Sep 25, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Fable 5, fork) — 🔴 BLOCK

Premise-level review of afbb210426318bb7f796964fc0c5ae3473eaf47c via the fork AI-review pipeline — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

First-Principles-Verdict: BLOCK

Retiming a Go-All stage from a 30-minute stop to a 4-hour ceiling is a product-shape change; base pins say "stops," and no RFC covers it.

Not justified as shipped

  • Item 1 — product-shape change without recorded decision: "This restores the documented Autopilot contract," yet the base contract table it rewrites says "Wall-clock budget per stage before auto-run stops."
    Clears when: an RFC covering the stage-timeout-as-start-gate shape is merged non-draft on the base branch and the author pushes or rebases, or a maintainer pushes this branch to this repository and records /ai-review override first-principles afbb210426318bb7f796964fc0c5ae3473eaf47c: <decision>.
  • Item 2 — rides along: the pause-with-Go/Cancel card exists only to serve Item 1's new semantics.
  • Item 4 — rides along: the attended-Go bypass is the plan-watchdog's auto_run-only rationale applied by symmetry to the stage budget.
  • Item 9 — rides along: awaiting_guidance exists only to admit guidance during Item 1's new pause.
  • Item 10 — rides along: help text, spec and docstring rewrites encode Item 1's new contract.

What this change ships

Inventory (10 items) — 5 justified

Intent: stop Autopilot Go-All from cutting productive stages at 30 minutes, and make a dashboard timeout actually stop the native agent turn — declared a FIX; the timeout-semantics half is a shape change.

  1. A running Go-All stage is no longer stopped at 30 minutes; it may run to the 4-hour chat ceiling — product-shape change without recorded decision (blocker c)
  2. A spent stage budget now pauses with a Go/Cancel card instead of stopping auto-run — rides along on Item 1
  3. The timeout card names the stage that spent the budget, not the pending stage — justified
  4. A human-approved Go is no longer rejected for a spent stage budget — rides along on Item 1
  5. Cancelling an acquired dashboard turn now sends native ACP session/cancel and waits for acknowledgement — justified
  6. An unacknowledged native stop shows a visible warning — justified
  7. A coroutine cancelled before acquiring the session lease no longer cancels another turn by key — justified
  8. A repeated cancellation retains the native cancel task and the partial reply — justified
  9. Guidance can be sent during a timeout pause without consuming the stage boundary (awaiting_guidance) — rides along on Item 1
  10. Config help, spec and docstrings reworded to the start-gate contract — rides along on Item 1

Blockers

product-shape change without accepted RFC — the Autopilot loop's default stage stop is replaced. The description claims "the bundled Autopilot contract says it gates whether another turn may start and an admitted stage may outlast it," but the base carries the opposite as pinned decisions this diff deletes: the config table row "Wall-clock budget per stage before auto-run stops" (autopilot.md:501), the tracker docstring naming "the orchestrator's asyncio.wait_for around a stage turn" as a consumer of the budget, and the test assertion "a stage cut at the ceiling must stop auto-run." No evidence (git history, pin message, linked issue — "N/A … no public issue has been filed") shows those pins wrong, and the base RFC list names no autopilot entry (rfc-orchestrator-chat-sessions is superseded). The cancellation-handoff items (5–8) and the card attribution fix (3) stand on their own traced defects and need no shape decision.
Clears when: an RFC covering this shape is merged non-draft on the base branch and the author pushes or rebases, or a maintainer pushes the branch to this repository and records /ai-review override first-principles afbb210426318bb7f796964fc0c5ae3473eaf47c: <decision>.

[FIRST-PRINCIPLES-REVIEWED] afbb210

@github-actions

github-actions Bot commented Sep 25, 2026 •

Copy link
Copy Markdown
Contributor

GPT 5.6 Review (fork) — 🔴 changes requested (blocking)

Reviewed afbb210426318bb7f796964fc0c5ae3473eaf47c via the fork AI-review pipeline; updated in place on each push.

1 of 1 blocking finding(s) are security-class and were withheld from adjudication, so the blocking verdict stands.

BLOCKING -- src/kiro_crew/dashboard/chat_runner.py:17326 -- Native cancellation occurs after the ACP stream marks the turn inactive
state.sessions.cancel_current(session_key, wait_ack_timeout=cancel_ack_timeout)
Chat ceiling during a silent tool -> ACP generator finalizer clears active state -> cancel returns no_turn -> native tool continues writing after the timeout.
Anchor: residual/crash-data-loss-corruption
Fix: Send session/cancel before unwinding the active ACP stream, and do not treat no_turn as acknowledgement.

[BLOCK-MERGE] afbb210
[GPT-REVIEWED] afbb210

Adjudication (Opus 5) — is blocking on each finding proportionate?

F1 — harm: UNBOUNDED-class residual per the fence anchor (silent loss of a turn's work while a native tool keeps running).
Conditions confirmed: _bounded_turn cancels the _run_chat task at turn_dispatch.py:331, the ACP stream generator's unwind finally unmarks the turn and sets _turn_done at session_handle.py:1685-1689, so is_turn_active is False at session_handle.py:1136 and AcpSessionProvider.cancel returns "no_turn" at session_provider.py:508-509 without ever sending METHOD_CANCEL (session_handle.py:1789); the new callback returns silently on "no_turn" (patch, chat_runner.py +373).
Recovery path: none — drain_active_turns only rescues "no_turn" when has_unfinished_turn is True (session_lifecycle.py:2015-2019), and the same _turn_done.set() makes that False too (session_handle.py:1149). The condition combination is the ordinary one for this new code path (ceiling fires while a tool is in flight), not an extreme one, and it is the exact invariant the PR's own spec text asserts, so no rarity argument is available.

[ADJUDICATION] afbb210 total=0 uphold=0 downgrade=0
[GPT-ADJUDICATED] afbb210
[ADJUDICATION-FENCED] afbb210 fenced=1 flagged=0
UPHOLD-FENCED F1 src/kiro_crew/dashboard/chat_runner.py:17326 -- The stream generator's unwind clears turn-active state before this handler runs, so the cancel resolves to no_turn and no native session/cancel is sent on the very path this code exists to cover; no drain or later gate recovers it.
[GPT-ADJUDICATED-FENCED] afbb210

@github-actions

github-actions Bot commented Sep 25, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Fable 5, fork) — 🔴 BLOCK (blocking)

Design-level review of afbb210426318bb7f796964fc0c5ae3473eaf47c via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: BLOCK

Redefines stage_timeout_seconds against the repo's own pins, citing an Autopilot "start-gate contract" this repo nowhere states.

Blockers

Rewrites the pins for the opposite behavior and treats them as a gap, not a decision
The description claims the timeout was misused "even though the bundled Autopilot contract says it gates whether another turn may start" — but the base spec says the reverse ("the running stage has its own ceiling", autopilot.md:287; config help "Max seconds per stage before auto-run stops"; the tracker docstring naming "the orchestrator's asyncio.wait_for around a stage turn"), and the rewritten test pinned it in its own assertion ("a stage cut at the ceiling must stop auto-run"). Cause → mechanism → consequence: an existing operator key silently changes meaning → a running Go-All stage turn is now bounded only by agent.chat_turn_timeout_secs (4 h default) → unattended runtime can exceed both the operator's 30-min stage cap and the 2 h whole-plan watchdog (boundary-checked only) by up to 4 h, with no config migration or warning — the PR's own edit deletes the sections.py line calling the plan budget "the only bound on total unattended runtime." The premature-cut pain is real, but the pause-with-Go/Cancel aftermath, accurate stage attribution, and native ACP cancel handoff in this PR already fix it without redefining the key: keep the mid-turn stage ceiling and route its firing into the new pause instead.
Clears when: a maintainer explicitly ratifies the admission-only semantics (decision entry or review approval naming the redefinition), or the PR retains the stage budget as the running-turn ceiling while keeping the new pause/resume aftermath.

Suggestions

  • Split the native ACP session/cancel handoff (lease-scoped cancel, ack warning, repeated-cancel retention) into its own PR — it fixes a traced real bug on every dashboard cancellation path and stands regardless of the budget-semantics question.

[DESIGN-REVIEWED] afbb210

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Sep 25, 2026
@github-actions

github-actions Bot commented Sep 25, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5 Review (fork) — ✅ no blocking findings

Reviewed afbb210426318bb7f796964fc0c5ae3473eaf47c via the fork AI-review pipeline; updated in place on each push.

Review details

Nothing blocks; the new native-ACP cancel handoff is inert on the exact path it was added for.

FINDING — src/kiro_crew/dashboard/chat_runner.py:17321 — when _bounded_turn cancels a stage turn parked awaiting the next ACP frame (the case where a tool is still running in the child), the CancelledError unwinds through _prompt_loop's / AcpSessionHandle's finally, which runs _turn_done.set() before this handler executes, so if _acquired: → cancel_current → provider.cancel hits if not self._client.has_active_turn(): return "no_turn" (providers/acp.py:2156, acp/session_provider.py:508), no session/cancel is ever written, and if outcome in ("acked", "no_turn"): return reports it as success — the native turn and its tool keep running while the card invites the user to Go → Fix: on this path force the cancel past the liveness early-return (a force=/has_unfinished_turn-based predicate on the provider seam), and stop counting "no_turn" as success here; note the added tests replace the seam with AsyncMock(return_value="acked"), so none of them can observe the short-circuit.

FINDING — src/kiro_crew/dashboard/chat_orchestrator.py:1476 — the start-gate's SEL record still writes resources=f"slot={slot.key},stage={stage_num}" while the card two lines above deliberately attributes the spent budget to budget_stage, so the audit trail logs stage_timeout against the pending stage that never ran → Fix: record stage={budget_stage} (optionally ,pending={stage_num}) and name the operation as a refused admission.

[OPUS-REVIEWED] afbb210

Treat the 30-minute stage budget as an admission gate between automatic turns instead of cancelling productive work mid-stage. Keep the independent chat-turn ceiling, cancel only a native turn whose lease this runner owns, surface unacknowledged cancellation, and merge current upstream main.
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Sep 25, 2026
@junjiequ
junjiequ marked this pull request as draft September 25, 2026 23:58
@junjiequ

Copy link
Copy Markdown
Contributor Author

Status: returned to draft pending RFC #14043. The RFC records six distinct 30-minute timeout events and two confirmed dashboard/native split-brain continuations. Current-head review also proved the prototype's native cancel runs too late: ACP stream unwind can clear active-turn state before cancel_current, causing no_turn without sending session/cancel. Next revision will split the urgent pre-unwind, lease-owned cancellation fix from the admission-only stage-budget change. The latter remains blocked until the RFC is accepted on main.

@dwu96

dwu96 commented Sep 26, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

3 similar comments
@dwu96

dwu96 commented Sep 26, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 27, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 27, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 28, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

2 similar comments
@dwu96

dwu96 commented Sep 28, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 28, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 28, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

4 similar comments
@dwu96

dwu96 commented Sep 28, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 28, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 28, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 28, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Sep 28, 2026
@dwu96

dwu96 commented Sep 29, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

5 similar comments
@dwu96

dwu96 commented Sep 29, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Sep 30, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Oct 1, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Oct 1, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Oct 1, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

@dwu96

dwu96 commented Oct 2, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

1 similar comment
@dwu96

dwu96 commented Oct 2, 2026

Copy link
Copy Markdown
Contributor

👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review.

When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically.

Copy link
Copy Markdown
Contributor

PR triage (automated, 2026-10-05) — stale against main + overlaps #16301

Checked against origin/main @ 5edf2feecb (merge base 1a2cafae, 1,582 commits behind):

  • The Autopilot half has no target on main any more. chore: remove Autopilot plan/stage loop #15362 (backend stage loop removal), chore(frontend): remove Autopilot mode #15359 (frontend Autopilot removal) and chore: remove the stage-boundary plumbing chat Autopilot left behind #15682 (stage-boundary plumbing removal) are merged. On main, src/kiro_crew/dashboard/chat_orchestrator.py and docs/system-specs/modules/autopilot.md are deleted, and stage_timeout_seconds no longer appears in config/sections.py. The merged RFC docs/request-for-change/rfc-autopilot-stage-budgets.md (docs(rfc): Preserve Autopilot incident evidence #16146) is status: superseded, and it says this draft implementation "no longer applies".
  • The native-cancel half is not on main yet. On main, _run_chat's except asyncio.CancelledError (dashboard/chat_runner.py ~L16987) still only persists the partial reply, and turn_dispatch._bounded_turn still cancels the task without sending session/cancel first. So this part is not covered.
  • Overlap: open PR fix(dashboard): cancel native turns before timeout unwind #16301 (same author, "cancel native turns before timeout unwind") is the split this PR's 2026-09-26 status comment promised. It moves the cancel ahead of the unwind in turn_dispatch.py and fences it by session owner. That fixes the no_turn problem the GPT/Opus reviews flagged in this PR's chat_runner.py hunk, where the cancel runs after the unwind.
  • git merge-tree shows conflicts in config-baseline.json, session.md, and a modify/delete on autopilot.md.

Recommendation: this PR is not being closed, because its cancel goal is still open on main. Please carry the native-cancel work only in #16301, then close this draft. If you want to keep it, rebase it onto main and drop the Autopilot/stage-budget changes.

@iamwhatever

Copy link
Copy Markdown
Contributor

Thanks for this, @junjiequ. We're not building a replacement here: the Autopilot stage loop this PR fixes was removed from main (#15362, #15359, #15682), so the files and the stage_timeout_seconds setting it touches no longer exist. The still-relevant native-turn cancel work carries on in your own #16301. Leaving this open for your call.

@iamwhatever

Copy link
Copy Markdown
Contributor

Thanks for this fix, @junjiequ. Closing because the Autopilot plan/stage loop it adjusts has been removed from main: #15362 (backend, merged 2026-09-30) and #15359 (frontend), per rfc-retire-chat-autopilot. On main, docs/system-specs/modules/autopilot.md is gone and the Stage Timeout setting no longer exists in src/kiro_crew/config/sections.py. The remaining generic chat-turn cancellation defect is tracked in #14058. Happy to reopen if part of this still applies.

@github-actions github-actions Bot removed the readiness: action required A blocking check or review needs attention label Oct 26, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

fork Pull request from a fork (external contributor) merge conflict Branch has merge conflicts with its base — author must resolve before merge

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants