fix(compaction): later provider timeouts discard completed chunks by mushuiyu886 · Pull Request #92899 · openclaw/openclaw

mushuiyu886 · 2026-06-14T06:32:06Z

Related: #92043

What Problem This Solves

Fixes an issue where multi-chunk compaction would discard summaries from completed chunks when a later provider request timed out, replacing useful progress with the generic "Summary unavailable due to size limits" fallback.

Why This Change Was Made

Non-caller provider timeouts now use the existing partial-summary recovery path after at least one chunk has completed. Caller cancellation and wrapper safety-timeout aborts remain terminal, so this no longer changes post-timeout AgentSession persistence, lifecycle, or stored-state policy.

This update intentionally narrows the previous 20-file proposal to the primary contributor-owned invariant. The bounded post-timeout commit policy, session lifecycle drain, retained-entry remapping, timeout-helper cleanup, and failure-copy changes are no longer part of this PR.

User Impact

When an early compaction chunk succeeds and a later provider request times out, OpenClaw retains the completed summary instead of losing all progress. Explicit cancellation behavior and the outer compaction safety timeout are unchanged.

Evidence

The narrowed behavior is covered by a red/green regression test, adjacent compaction tests, and an actual loopback OpenAI-compatible HTTP/SSE before/after run.

Real behavior proof

A loopback OpenAI-compatible HTTP/SSE provider exercised the production summary transport with two chunks: the first request returned a streamed summary and the second returned HTTP 504 with a provider timeout error.

Before on origin/main (277f009cc902aecf09ee0cdd8345223186c16fa9): 2 provider requests; first chunk completed; later request timed out; retainedCompletedChunk=false; partialMarkerPresent=false; generic fallback returned.
After on 1dc4de07426a218efe07d436d5db92bd1c7c37b8: 2 provider requests; first chunk completed; later request timed out; retainedCompletedChunk=true; partialMarkerPresent=true; generic fallback absent.
The proof used temporary loopback-only credentials and did not call a third-party provider.

Tests and checks

RED: node scripts/run-vitest.mjs src/agents/compaction-partial-summary.test.ts failed because a later timeout returned the generic fallback.
GREEN: the same focused file passed, 9/9 tests.
Adjacent compaction tests: compaction-partial-summary.test.ts plus compaction.summarize-fallback.test.ts, 13/13 tests passed.
Formatting, conflict-marker, max-lines, dependency-pin, Plugin SDK API baseline, deprecated API, plugin-boundary, package-patch, and temp-creation checks passed.
Local core typecheck was blocked by three unchanged packages/net-policy/src/ip.ts errors; the identical command reproduced the same errors on a clean origin/main control worktree with the same dependency tree.
Structured autoreview built the final two-file bundle, but the Codex Responses WebSocket returned HTTP 403 before producing findings. The final diff was manually reviewed against summarizeWithFallbackResult, caller-abort behavior, timeout classification, and adjacent partial-summary tests.

clawsweeper · 2026-06-14T06:34:38Z

Codex review: needs maintainer review before merge. Reviewed July 17, 2026, 10:03 PM ET / July 18, 2026, 02:03 UTC.

Summary
The branch changes summarizeChunks so a non-caller provider timeout after an earlier completed chunk returns the existing partial-summary result instead of discarding that progress for the generic fallback.

PR surface: Source -4, Tests +1. Total -3 across 2 files.

Reproducibility: yes. from the supplied production-path loopback HTTP/SSE proof and focused regression fixture: complete one summary chunk, then return a provider timeout for the next chunk while the caller signal remains unaborted. I did not independently execute that path in this read-only review.

Review metrics: none identified.

Merge readiness
Overall: 🐚 platinum hermit
Proof: 🦞 diamond lobster
Patch quality: 🐚 platinum hermit
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

Refresh the exact current-main merge result and preserve the timeout-plus-caller-abort assertions before approval.

Risk before merge

[P2] Returning a partial summary changes the session context that can be committed after a provider timeout; before merge, confirm the current-main three-way result still preserves terminal caller cancellation and the outer compaction safety-timeout behavior.

Maintainer options:

Refresh the current-main compaction proof (recommended)
Rebase or otherwise review the clean current-main merge result, then retain the focused timeout and caller-abort coverage before approval.

Next step before merge

[P2] No repair lane is indicated because the current narrowed branch has no actionable source-level finding; it needs normal owner approval after a current-main review refresh.

Security
Cleared: The two-file TypeScript change adds no dependency, workflow, permission, secret, package-resolution, or execution-surface change.

Review details

Best possible solution:

Land the narrow partial-summary recovery only if the rebased current-main result retains terminal caller cancellation and safety-timeout handling while preserving completed chunks for non-caller provider failures.

Do we have a high-confidence way to reproduce the issue?

Yes, from the supplied production-path loopback HTTP/SSE proof and focused regression fixture: complete one summary chunk, then return a provider timeout for the next chunk while the caller signal remains unaborted. I did not independently execute that path in this read-only review.

Is this the best way to solve the issue?

Yes for this bounded failure mode: it reuses the existing partial-summary path and preserves the established terminal boundary for an actually aborted caller signal. Cross-attempt persistence and timeout-budget policy remain separate work in #92043.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 3d1b77103511.

Label changes

Label changes:

add rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
add status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (live_output): The PR supplies a before/after loopback OpenAI-compatible HTTP/SSE run through the production summary transport, and the current-head real-behavior-proof check succeeded; the evidence directly demonstrates retention of a completed first chunk after a second-request HTTP 504.
remove rating: 🦐 gold shrimp: Current PR rating is rating: 🐚 platinum hermit, so this older rating label is no longer current.
remove merge-risk: 🚨 availability: Current PR review merge-risk labels are merge-risk: 🚨 session-state.
remove status: ⏳ waiting on author: Current PR status label is status: 👀 ready for maintainer look.

Label justifications:

P1: A later provider timeout can discard completed compaction progress and leave affected agent sessions with the generic context-recovery fallback.
merge-risk: 🚨 session-state: The patch intentionally changes which partial compaction summary may reach session persistence after a provider timeout.
rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (live_output): The PR supplies a before/after loopback OpenAI-compatible HTTP/SSE run through the production summary transport, and the current-head real-behavior-proof check succeeded; the evidence directly demonstrates retention of a completed first chunk after a second-request HTTP 504.
proof: sufficient: Contributor real behavior proof is sufficient. The PR supplies a before/after loopback OpenAI-compatible HTTP/SSE run through the production summary transport, and the current-head real-behavior-proof check succeeded; the evidence directly demonstrates retention of a completed first chunk after a second-request HTTP 504.

Evidence reviewed

PR surface:

Source -4, Tests +1. Total -3 across 2 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	1	2	6	-4
Tests	1	6	5	+1
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	2	8	11	-3

What I checked:

Narrow production behavior: The changed summarizeChunks catch block now rethrows only when the supplied caller signal is aborted, allowing later non-caller provider failures to flow into the existing partial-summary recovery when a chunk has already completed. (src/agents/compaction.ts:194, 1dc4de07426a)
Regression coverage: The changed regression test asserts that a later TimeoutError retains the first chunk summary, includes the partial marker and chunk range, and emits the existing partial-summary warning with one completed chunk. (src/agents/compaction-partial-summary.test.ts:170, 1dc4de07426a)
Real behavior proof: The PR body records a before/after loopback OpenAI-compatible HTTP/SSE run with two chunks: the first completed, the second returned HTTP 504, and the patched head retained the completed chunk and partial marker. The current-head Real behavior proof check also completed successfully. (1dc4de07426a)
Merged adjacent precedent: Merged PR fix(agents): retry compaction on provider-side AbortErrors #97504 established the adjacent compaction rule that provider-side abort-like errors are not caller cancellation when the signal is not aborted; this branch applies the same caller-signal boundary to later timeout recovery. (src/agents/compaction.ts:194, 245c18da5a6c)

Likely related people:

hugenshen: Authored the merged provider-side compaction abort-classification work that established the nearby caller-signal invariant used by this patch. (role: adjacent behavior introducer; confidence: high; commits: 245c18da5a6c; files: src/agents/compaction.ts, src/agents/compaction.summarize-fallback.test.ts)
vincentkoc: Was assigned this PR and is identified in the related preflight-compaction history as carrying the connection between reply lifecycle and compaction behavior. (role: recent compaction review owner; confidence: medium; commits: 280d1cb97; files: src/agents/compaction.ts)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

Review history (27 earlier review cycles; latest 8 shown)

reviewed 2026-07-14T23:22:06.270Z sha 67b8ab1 :: needs changes before merge. :: [P1] Restore the timeout helper used by the Pi runtime
reviewed 2026-07-14T23:40:01.725Z sha 67b8ab1 :: needs changes before merge. :: [P1] Restore the timeout helper used by the Pi runtime
reviewed 2026-07-14T23:57:07.279Z sha 67b8ab1 :: needs changes before merge. :: [P1] Restore the timeout helper required by Pi compaction
reviewed 2026-07-15T00:16:11.688Z sha 67b8ab1 :: needs changes before merge. :: [P1] Keep the shared timeout helper required by Pi
reviewed 2026-07-15T00:44:49.412Z sha 67b8ab1 :: needs changes before merge. :: [P1] Preserve the shared timeout helper required by Pi
reviewed 2026-07-16T02:28:26.309Z sha c5cdaef :: needs maintainer review before merge. :: none
reviewed 2026-07-16T04:06:30.443Z sha c5cdaef :: needs maintainer review before merge. :: none
reviewed 2026-07-18T01:21:44.726Z sha 1dc4de0 :: needs maintainer review before merge. :: none

openclaw-barnacle · 2026-07-15T04:29:59Z

This assigned pull request has been automatically marked as stale after being open for 27 days.
Please add updates or it will be closed.

github-actions · 2026-07-15T19:31:01Z

Security-sensitive guard cleared

This PR no longer has blocked security-sensitive file changes. A future security-sensitive change requires a fresh /allow-security-sensitive-change comment after the guard blocks that new head SHA.

Current SHA: 1dc4de07426a218efe07d436d5db92bd1c7c37b8

github-actions · 2026-07-15T19:31:42Z

Dependency graph guard cleared

This PR no longer has blocked dependency graph changes. A future dependency graph change requires a fresh /allow-dependencies-change comment after the guard blocks that new head SHA.

Current SHA: 1dc4de07426a218efe07d436d5db92bd1c7c37b8

openclaw-barnacle · 2026-07-16T04:36:31Z

This assigned pull request has been automatically marked as stale after being open for 27 days.
Please add updates or it will be closed.

mushuiyu886 · 2026-07-16T08:05:41Z

@clawsweeper re-review

clawsweeper · 2026-07-16T08:05:44Z

🦞👀
ClawSweeper picked this up.

Command router queued. I will update this comment with the next step.

openclaw-barnacle · 2026-07-17T04:35:03Z

This assigned pull request has been automatically marked as stale after being open for 27 days.
Please add updates or it will be closed.

mushuiyu886 · 2026-07-18T01:18:17Z

Updated the branch to resolve the current main conflicts and narrowed the PR to the primary contributor-owned invariant from the review.

Final surface: 2 files, 8 additions / 11 deletions.
Removed from this PR: post-timeout AgentSession commits, stored-state/boundary changes, lifecycle drain/invalidation, timeout-helper deletion, and failure-copy changes.
Preserved behavior: caller cancellation and wrapper safety-timeout aborts remain terminal.
Focused tests: 9/9 passed; adjacent compaction tests: 13/13 passed.
Actual loopback OpenAI-compatible HTTP/SSE proof: the first chunk completed and the second request returned HTTP 504. origin/main discarded the completed chunk; this head retained it with the existing partial-summary marker.
Formatting and repository guards passed. Core typecheck reached an unchanged packages/net-policy/src/ip.ts dependency/type mismatch that reproduced identically on a clean origin/main control worktree.
Codex structured autoreview could not produce findings because the Responses WebSocket returned HTTP 403; the final two-file diff was manually reviewed against timeout classification, caller-abort handling, and adjacent partial-summary paths.

@clawsweeper re-review

clawsweeper · 2026-07-18T01:18:19Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Re-review progress:

State: Complete
Detail: The durable review comment was published and its verdict was queued for target-serialized routing.
Run: https://github.com/openclaw/clawsweeper/actions/runs/29626379286
Updated: 2026-07-18T02:08:22.455Z

openclaw-barnacle · 2026-07-18T04:29:26Z

This assigned pull request has been automatically marked as stale after being open for 27 days.
Please add updates or it will be closed.

openclaw-barnacle Bot added agents Agent runtime and tooling size: XS proof: supplied External PR includes structured after-fix real behavior proof. labels Jun 14, 2026

openclaw-barnacle Bot added size: S size: M and removed size: XS size: S labels Jun 14, 2026

clawsweeper Bot mentioned this pull request Jun 14, 2026

Bug: Preflight compaction shares abortSignal with main turn, causing cascading turn abortion #79390

Closed

clawsweeper Bot added rating: 🦞 diamond lobster Very strong PR readiness with only minor maintainer review expected. and removed rating: 🐚 platinum hermit Good normal PR readiness with ordinary maintainer review expected. labels Jun 15, 2026

vincentkoc self-assigned this Jun 16, 2026

clawsweeper Bot mentioned this pull request Jun 16, 2026

Session hangs indefinitely when compaction times out, causing repeated duplicate message sends #43661

Closed

itanyplus mentioned this pull request Jun 17, 2026

Bug: 180s compaction timeout is a single wall clock over the whole chunk pipeline with no partial-progress reuse, so a legitimately-long compaction fails identically every turn #92043

Open

mushuiyu886 force-pushed the feat/issue-92043 branch from 5013dda to cdacf1e Compare June 18, 2026 01:33

openclaw-barnacle Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label Jun 18, 2026

clawsweeper Bot added the proof: sufficient ClawSweeper judged the real behavior proof convincing. label Jun 19, 2026

Merge branch 'main' into feat/issue-92043

353e18d

mushuiyu886 added 10 commits July 16, 2026 01:08

fix(compaction): prevent repeated restarts after safety timeouts

9f6fcf0

fix(compaction): resolve rebased CI failures

0d66267

fix(compaction): preserve timeout progress across locks and stages

850f24b

fix(compaction): preserve partial timeout ownership

7cf7428

fix(compaction): await lifecycle before disposal

10711ea

fix(compaction): bound post-timeout lifecycle cleanup

9c408a1

fix(compaction): retain completed phase checkpoints

9ff3092

fix(compaction): preserve repaired boundaries and lifecycle

39b9adb

test(compaction): align partial progress test support

7a54310

chore: reconcile hosted PR history

0fe148a

Merge branch 'main' into feat/issue-92043

c5cdaef

fix(compaction): retain completed chunks after provider timeouts

1dc4de0

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(compaction): later provider timeouts discard completed chunks#92899

fix(compaction): later provider timeouts discard completed chunks#92899
mushuiyu886 wants to merge 14 commits into
openclaw:mainfrom
mushuiyu886:feat/issue-92043

mushuiyu886 commented Jun 14, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 14, 2026 •

edited

Loading

Uh oh!

openclaw-barnacle Bot commented Jul 15, 2026

Uh oh!

github-actions Bot commented Jul 15, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Jul 15, 2026 •

edited

Loading

Uh oh!

openclaw-barnacle Bot commented Jul 16, 2026

Uh oh!

mushuiyu886 commented Jul 16, 2026

Uh oh!

clawsweeper Bot commented Jul 16, 2026

Uh oh!

openclaw-barnacle Bot commented Jul 17, 2026

Uh oh!

mushuiyu886 commented Jul 18, 2026

Uh oh!

clawsweeper Bot commented Jul 18, 2026 •

edited

Loading

Uh oh!

openclaw-barnacle Bot commented Jul 18, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Uh oh!

Conversation

mushuiyu886 commented Jun 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What Problem This Solves

Why This Change Was Made

User Impact

Evidence

Real behavior proof

Tests and checks

Uh oh!

clawsweeper Bot commented Jun 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openclaw-barnacle Bot commented Jul 15, 2026

Uh oh!

github-actions Bot commented Jul 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Security-sensitive guard cleared

Uh oh!

github-actions Bot commented Jul 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Dependency graph guard cleared

Uh oh!

openclaw-barnacle Bot commented Jul 16, 2026

Uh oh!

mushuiyu886 commented Jul 16, 2026

Uh oh!

clawsweeper Bot commented Jul 16, 2026

Uh oh!

openclaw-barnacle Bot commented Jul 17, 2026

Uh oh!

mushuiyu886 commented Jul 18, 2026

Uh oh!

clawsweeper Bot commented Jul 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openclaw-barnacle Bot commented Jul 18, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

mushuiyu886 commented Jun 14, 2026 •

edited

Loading

clawsweeper Bot commented Jun 14, 2026 •

edited

Loading

github-actions Bot commented Jul 15, 2026 •

edited

Loading

github-actions Bot commented Jul 15, 2026 •

edited

Loading

clawsweeper Bot commented Jul 18, 2026 •

edited

Loading