fix(auto-reply): preserve sessions after compaction failures by fuller-stack-dev · Pull Request #70479 · openclaw/openclaw

fuller-stack-dev · 2026-04-23T04:45:14Z

Summary

preserve the active session mapping when overflow recovery or compaction failure cannot recover a turn
keep the existing dynamic overflow guidance while adding explicit preserved-session guidance to retry, /compact, or /new
mark preserved-session failure replies for delivery despite source-reply suppression
add regression coverage for embedded overflow payloads and thrown compaction-failure errors
update the session compaction docs for the preserved-session failure path

Real behavior proof

Behavior addressed: auto-reply no longer rotates/orphans the active session mapping when compaction or overflow recovery fails; the user gets a delivered failure reply that keeps dynamic overflow hints.

Real environment tested: local OpenClaw source checkout in an isolated PR worktree after rebasing onto current upstream/main b8e9ab9385 (fix(codex): surface native compaction failures (#85160)); pnpm openclaw --version built and ran the pushed head 193d3c0. No live WebChat/model run.

Exact steps or command run after this patch:

git fetch upstream main && git fetch origin fix/preserve-session-on-compaction-failure
git rebase upstream/main (clean rebase; no textual conflicts)
git grep -n "resetSessionAfterCompactionFailure" upstream/main -- src/auto-reply/reply src/agents confirmed current upstream/main still has the reset path, so this PR is still relevant after fix(codex): surface native compaction failures #85160.
pnpm docs:list
pnpm openclaw --version
node --import tsx --input-type=module production-helper probe importing buildContextOverflowRecoveryText, markReplyPayloadForSourceSuppressionDelivery, and getReplyPayloadMetadata from src/auto-reply.
pnpm exec oxfmt --check --threads=1 src/auto-reply/reply/agent-runner-execution.ts src/auto-reply/reply/agent-runner.ts src/auto-reply/reply/agent-runner-execution.test.ts CHANGELOG.md docs/reference/session-management-compaction.md
focused Vitest through scripts/run-vitest.mjs exports, because local Node 23 did not execute that script's import.meta.main branch: node --input-type=module with spawnWatchedVitestProcess({ pnpmArgs: ['exec', 'node', ...resolveVitestNodeArgs(process.env), resolveVitestCliEntry(), 'run', 'src/auto-reply/reply/agent-runner-execution.test.ts', '--reporter=dot'] })
compaction-adjacent Vitest selection through the same wrapper exports: src/agents/command/cli-compaction.test.ts, src/agents/command/session-store.test.ts, extensions/codex/src/app-server/compact.test.ts
git diff --check upstream/main...HEAD
autoreview branch pass with tests disabled, using the focused proof above: --mode branch --base upstream/main

Evidence after fix: terminal output from the local OpenClaw source setup and production-helper probe after rebasing on #85160:

$ pnpm openclaw --version
OpenClaw 2026.5.21 (193d3c0)

$ node --import tsx --input-type=module
runtime probe after upstream compaction rebase
active session before: session
active session after: session
deliverDespiteSourceReplySuppression: true
contains preserved-session guidance: true
contains dynamic reserveTokensFloor hint: true
reply preview: ⚠️ Auto-compaction could not recover this turn. I kept this conversation mapped to the current session. Please try again, use /compact, or use /new to start a fresh session.

Supplemental focused runtime coverage: src/auto-reply/reply/agent-runner-execution.test.ts passed 111 tests. The compaction-adjacent selection from #85160 passed 100 tests across src/agents/command/cli-compaction.test.ts, src/agents/command/session-store.test.ts, and extensions/codex/src/app-server/compact.test.ts. The new regression tests assert the preserved active session remains session, replyOperation.updateSessionId is not called, updateSessionStore is not called, the failure payload has deliverDespiteSourceReplySuppression: true, and the reply includes both the preserved-session guidance and the existing dynamic reserveTokensFloor hint.

Observed result after fix: the pushed head is rebased on b8e9ab9385, the local OpenClaw source build reports 193d3c0, the production-helper probe keeps the session id unchanged, marks the failure reply for suppressed-source delivery, and keeps the dynamic overflow hint. Autoreview reported no actionable regressions against the new upstream/main.
Maintainer supplemental before/after proof (May 22, 2026): I ran the same deterministic failure-path probe against clean archive copies of pre-fix origin/main and this PR head 193d3c0fdd5a42cc5291f4f279e80f872fdd1051. The probe exercises the actual runAgentTurnWithFallback branches for both embedded context-overflow error payloads and thrown compaction failures, with the real resetReplyRunSession helper wired on the pre-fix side.

Exact commands run for the before/after proof:

Pre-fix: env OPENCLAW_PR70479_EXPECT=broken OPENCLAW_VITEST_MAX_WORKERS=1 OPENCLAW_VITEST_NO_OUTPUT_TIMEOUT_MS=120000 node scripts/run-vitest.mjs src/auto-reply/reply/pr70479-live-proof.test.ts --reporter=verbose in a git archive origin/main checkout.
PR head: env OPENCLAW_PR70479_EXPECT=fixed OPENCLAW_VITEST_MAX_WORKERS=1 OPENCLAW_VITEST_NO_OUTPUT_TIMEOUT_MS=120000 node scripts/run-vitest.mjs src/auto-reply/reply/pr70479-live-proof.test.ts --reporter=verbose in a git archive 193d3c0fdd5a42cc5291f4f279e80f872fdd1051 checkout.
PR committed regression slice: env OPENCLAW_VITEST_MAX_WORKERS=1 OPENCLAW_VITEST_NO_OUTPUT_TIMEOUT_MS=120000 node scripts/run-vitest.mjs src/auto-reply/reply/agent-runner-execution.test.ts -t "preserves the active session" --reporter=verbose.

Before fix evidence from origin/main:

PR70479_PROOF {"scenario":"embedded context-overflow payload","expectedMode":"broken","activeSessionId":"session-after","followupSessionId":"session-after","updateSessionStoreCalls":1,"updateSessionIdCalls":1,"deliverDespiteSourceReplySuppression":true,"text":"⚠️ Context limit exceeded. I've reset our conversation to start fresh - please try again.\n\nTo prevent this, increase your compaction buffer by setting `agents.defaults.compaction.reserveTokensFloor` to 20000 or higher in your config."}
PR70479_PROOF {"scenario":"thrown compaction failure","expectedMode":"broken","activeSessionId":"session-after","followupSessionId":"session-after","updateSessionStoreCalls":1,"updateSessionIdCalls":1,"deliverDespiteSourceReplySuppression":true,"text":"⚠️ Context limit exceeded during compaction. I've reset our conversation to start fresh - please try again.\n\nTo prevent this, increase your compaction buffer by setting `agents.defaults.compaction.reserveTokensFloor` to 20000 or higher in your config."}

After fix evidence from PR head:

PR70479_PROOF {"scenario":"embedded context-overflow payload","expectedMode":"fixed","activeSessionId":"session-before","followupSessionId":"session-before","updateSessionStoreCalls":0,"updateSessionIdCalls":0,"deliverDespiteSourceReplySuppression":true,"text":"⚠️ Auto-compaction could not recover this turn. I kept this conversation mapped to the current session. Please try again, use /compact, or use /new to start a fresh session.\n\nTo prevent this, increase your compaction buffer by setting `agents.defaults.compaction.reserveTokensFloor` to 20000 or higher in your config."}
PR70479_PROOF {"scenario":"thrown compaction failure","expectedMode":"fixed","activeSessionId":"session-before","followupSessionId":"session-before","updateSessionStoreCalls":0,"updateSessionIdCalls":0,"deliverDespiteSourceReplySuppression":true,"text":"⚠️ Auto-compaction could not recover this turn. I kept this conversation mapped to the current session. Please try again, use /compact, or use /new to start a fresh session.\n\nTo prevent this, increase your compaction buffer by setting `agents.defaults.compaction.reserveTokensFloor` to 20000 or higher in your config."}

Observed result from the before/after proof: pre-fix origin/main rewrites the active agent:main:main session from session-before to session-after and calls both updateSessionStore and replyOperation.updateSessionId. The PR head keeps the active session at session-before in both failure modes, makes zero session-store/updateSessionId calls, and returns a failure payload marked deliverDespiteSourceReplySuppression: true with retry, /compact, and /new guidance. The committed focused regression slice passed: 2 passed | 109 skipped for the two preserves the active session tests.

What was not tested: live WebChat UI overflow with a real model; the compaction/overflow failure behavior was exercised with a local OpenClaw source runtime probe plus focused runtime-unit coverage.

Notes

This remains separate from fix(agents): derive overflow budgets from provider errors #70473. That PR improves overflow-budget extraction; this PR fixes the later auto-reply session remap/orphaning behavior after compaction recovery still fails.

greptile-apps · 2026-04-23T04:47:13Z

Greptile Summary

This PR fixes session orphaning after compaction failures in the auto-reply agent. Instead of silently rotating the session key to a new session ID when overflow recovery or compaction fails, both failure paths now preserve the active session mapping and surface an explicit user message with actionable guidance (retry, /compact, or /new). The resetSessionAfterCompactionFailure callback is fully removed from the runAgentTurnWithFallback interface and all call-sites, resolving the dead-parameter concern flagged in a prior review. Regression tests are added for both the embedded-overflow payload path and the thrown-error compaction failure path.

Confidence Score: 5/5

This PR is safe to merge — it correctly removes a session-rotation footgun and replaces it with session-preserving, user-visible guidance, with targeted regression coverage for both failure paths.

All changes are narrowly scoped to the compaction-failure paths. The dead resetSessionAfterCompactionFailure parameter raised in the previous review is fully removed. Both new code paths are covered by regression tests that assert the exact user-facing text and the fail() call. No P0/P1 issues identified.

No files require special attention.

_{Reviews (2): Last reviewed commit: "fix(auto-reply): drop dead compaction re..." | Re-trigger Greptile}

fuller-stack-dev · 2026-04-23T05:48:53Z

Addressed the dead resetSessionAfterCompactionFailure surface in 3befdda128 by removing the unused execution param and all remaining call-site/test stubs.

Current red CI on the previous run does not point at this branch’s diff:

checks-node-extensions-shard-6 failed in extensions/codex/src/app-server/{approval,elicitation}-bridge.test.ts; those same targeted tests pass locally on this branch.
check-lint flagged extensions/codex/src/app-server/compact.ts:192 (NaN vs Number.NaN), which is outside this PR’s changed files.

This branch only changes:

src/auto-reply/reply/agent-runner.ts
src/auto-reply/reply/agent-runner-execution.ts
src/auto-reply/reply/agent-runner-execution.test.ts
docs/reference/session-management-compaction.md

fuller-stack-dev · 2026-04-23T05:56:08Z

@greptileai review again puhlez

fuller-stack-dev · 2026-04-23T07:48:19Z

Rebased onto latest main (f8894d5) and retriggering CI. The docker-e2e-fast failure on the previous run was a timeout (exit 124) in the bundled-channel Docker E2E, unrelated to our changes.

martingarramon · 2026-04-23T18:40:42Z

Static review. LGTM-shape on removing the silent session rotation — surfacing "please /compact or /new" is the right UX.

Three checks:

Retry-loop regression. The embeddedError branch previously returned only after resetSessionAfterCompactionFailure returned true. Now it returns unconditionally on any isContextOverflowError without the didResetAfterCompactionFailure guard. Confirm the surrounding retry harness can't re-enter with the same error — the old guard tracked "already reset once"; if that state lived inside the removed function, the retry loop needs a replacement latch.
Factory-level did-reset state. agent-runner.ts:1179-1185's resetSession helper is still called from resetSessionAfterRoleOrderingConflict via the same factory. If the factory tracks "did-reset-once" that the removed call was flipping, the role-ordering path could inherit stale state.
Dead-reference check. grep -rn resetSessionAfterCompactionFailure src/ after merge should return zero hits — signature removed at ~L587 but the full tree needs confirmation.

Non-blocking: user-facing string change is a behavior regression for anyone grepping transcripts for "reset our conversation to start fresh". Worth a release-note mention.

vincentkoc · 2026-04-27T06:49:54Z

ProjectClownfish pushed a narrow repair to this branch so the original contributor path can stay canonical.

Source PR: #70479
Validation: pnpm -s vitest run src/auto-reply/reply/agent-runner-execution.test.ts; pnpm check:changed
Contributor credit is preserved in the branch history and PR context.

clawsweeper · 2026-04-27T07:44:23Z

Codex review: passed.

Workflow note: Future ClawSweeper reviews update this same comment in place.

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.

Summary
The PR removes the auto-reply compaction-failure session reset hook, adds preserved-session recovery guidance for overflow/compaction failure paths, and updates focused tests, docs, and the changelog.

Reproducibility: yes. at source level with high confidence. Current main routes both embedded overflow payloads and thrown compaction failures into resetSessionAfterCompactionFailure, and the PR body includes before/after terminal proof of those branches.

PR rating
Overall: 🐚 platinum hermit
Proof: 🐚 platinum hermit
Patch quality: 🐚 platinum hermit
Summary: The patch is focused and has deterministic before/after runtime proof; live WebChat proof would be stronger but is not required to understand this branch behavior.

Rank-up moves:

none

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.

Real behavior proof
Sufficient (terminal): The updated PR body includes maintainer before/after terminal proof showing the pre-fix reset and the PR-head preserved session mapping with delivered failure payload metadata.

Mantis proof suggestion
A live WebChat/runtime proof would be useful because the change affects visible failure guidance and active session mapping after compaction failure. A maintainer can ask Mantis to capture proof by posting a new PR comment that starts with the OpenClaw Mantis account mention, followed by:

visual task: verify WebChat compaction-failure recovery preserves agent:main:main mapping and shows retry /compact /new guidance.

Risk before merge

Merging changes the failure path from automatic session rotation to preserved-session guidance, so maintainers should be comfortable owning that session-state behavior change.
The proof now exercises the actual fallback branches in terminal runs, but it still is not a live WebChat UI overflow with a real model.

Maintainer options:

Proceed with the supplied proof (recommended)
Accept the deterministic before/after runAgentTurnWithFallback proof as enough evidence for this failure path and let exact-head merge gates run.
Ask for one live WebChat proof
Require a redacted live WebChat/model overflow run or runtime log showing the active mapping remains unchanged and the guidance is delivered.
Keep paused for maintainer validation
Hold the PR until a maintainer with the right setup validates the failure path directly.

Next step before merge
No repair PR is needed; the review found no actionable code defect and the remaining work is normal exact-head merge gating or maintainer proof preference.

Security
Cleared: The diff touches auto-reply control flow, focused tests, docs, and changelog only, with no dependency, CI, secret, permission, package-resolution, or external code-execution change.

Review details

Best possible solution:

Land the session-preserving recovery path if maintainers accept the supplied before/after runtime proof, with CI and exact-head automerge gates handling normal merge safety.

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

Yes, at source level with high confidence. Current main routes both embedded overflow payloads and thrown compaction failures into resetSessionAfterCompactionFailure, and the PR body includes before/after terminal proof of those branches.

Is this the best way to solve the issue?

Yes. Preserving the active mapping and returning explicit retry /compact /new guidance is a narrow fix for the existing orphaning path, without adding a new config mode or duplicating session policy.

Label changes:

add proof: sufficient: Contributor real behavior proof is sufficient. The updated PR body includes maintainer before/after terminal proof showing the pre-fix reset and the PR-head preserved session mapping with delivered failure payload metadata.
add rating: 🐚 platinum hermit: Current PR rating is 🐚 platinum hermit because proof is 🐚 platinum hermit, patch quality is 🐚 platinum hermit, and The patch is focused and has deterministic before/after runtime proof; live WebChat proof would be stronger but is not required to understand this branch behavior.
remove rating: 🦪 silver shellfish: Current PR rating is rating: 🐚 platinum hermit, so this older rating label is no longer current.

Label justifications:

P1: The PR targets a user-visible session orphaning path that can break long-running WebChat continuity after compaction failure.
merge-risk: 🚨 session-state: Merging intentionally changes failure-path session mapping behavior from rotation to preservation.
merge-risk: 🚨 message-delivery: The fix depends on delivered failure guidance despite source-reply suppression.
rating: 🐚 platinum hermit: Current PR rating is 🐚 platinum hermit because proof is 🐚 platinum hermit, patch quality is 🐚 platinum hermit, and The patch is focused and has deterministic before/after runtime proof; live WebChat proof would be stronger but is not required to understand this branch behavior.
status: 🚀 automerge armed: This PR is in ClawSweeper's automerge lane. Sufficient (terminal): The updated PR body includes maintainer before/after terminal proof showing the pre-fix reset and the PR-head preserved session mapping with delivered failure payload metadata.
proof: sufficient: Contributor real behavior proof is sufficient. The updated PR body includes maintainer before/after terminal proof showing the pre-fix reset and the PR-head preserved session mapping with delivered failure payload metadata.

What I checked:

Current main still has the reset callback: Current main wires resetSessionAfterCompactionFailure from runReplyAgent into runAgentTurnWithFallback, so the PR is not obsolete on main. (src/auto-reply/reply/agent-runner.ts:1440, 55cfe00a3a40)
Current main resets embedded overflow payloads: The embedded context-overflow payload branch on current main calls resetSessionAfterCompactionFailure before returning recovery text. (src/auto-reply/reply/agent-runner-execution.ts:2233, 55cfe00a3a40)
Current main resets thrown compaction failures: The thrown compaction-failure path on current main calls resetSessionAfterCompactionFailure before returning recovery text. (src/auto-reply/reply/agent-runner-execution.ts:2367, 55cfe00a3a40)
PR preserves the active session mapping: The PR removes the compaction reset callback and makes both failure branches return preserved-session guidance instead of rotating to a fresh session id. (src/auto-reply/reply/agent-runner-execution.ts:2234, 193d3c0fdd5a)
PR regression coverage targets both failure modes: The new tests assert the active session id remains unchanged, updateSessionId and updateSessionStore are not called, and the failure payload is marked for delivery despite source-reply suppression. (src/auto-reply/reply/agent-runner-execution.test.ts:4372, 193d3c0fdd5a)
Failure payloads bypass source-reply suppression: markAgentRunFailureReplyPayload wraps failure replies with markReplyPayloadForSourceSuppressionDelivery, and dispatch honors deliverDespiteSourceReplySuppression when automatic source delivery is suppressed. (src/auto-reply/reply/dispatch-from-config.ts:2140, 55cfe00a3a40)

Likely related people:

steipete: Git history ties the compaction overflow reset and subsequent recovery/refactor work to the current auto-reply failure path. (role: introduced behavior; confidence: high; commits: 67743325eeb5, 9faa95d55832, ea018a68ccb9; files: src/auto-reply/reply/agent-runner.ts, src/auto-reply/reply/agent-runner-execution.ts)
joshavant: Recently merged the native compaction-failure surfacing work that this PR rebased on and explicitly separates from. (role: adjacent compaction owner; confidence: medium; commits: b8e9ab9385c0; files: extensions/codex/src/app-server/compact.ts, src/agents/command/cli-compaction.ts, src/agents/pi-embedded-runner/run.ts)
Takhoffman: Raised the main review concerns in the PR discussion around retry-loop behavior, factory reset state, and dead reset references. (role: reviewer; confidence: medium; files: src/auto-reply/reply/agent-runner-execution.ts, src/auto-reply/reply/agent-runner.ts)

Codex review notes: model gpt-5.5, reasoning high; reviewed against 55cfe00a3a40.

clawsweeper · 2026-05-20T11:35:23Z

ClawSweeper PR egg

✨ Hatched: 🥚 common Gilded Lint Imp

Hatch command

Comment @clawsweeper hatch when this PR is hatchable.

Hatchability rules:

Merged PRs are hatchable.
Open PRs are hatchable when they are status: 👀 ready for maintainer look, status: 🚀 automerge armed, or labeled clawsweeper:automerge.
Closed unmerged PRs are hatchable only when one of those hatchable labels is still present in the durable record.

Rarity: 🥚 common.
Trait: sparkles near resolved comments.
Image traits: location green-check meadow; accessory review stamp; palette cobalt, lime, and pearl; mood patient; pose standing beside its cracked shell; shell brushed metal shell; lighting clean product lighting; background small green status lights.
Share on X: post this hatch
Copy: My PR egg hatched a 🥚 common Gilded Lint Imp in ClawSweeper.

What is this egg doing here?

Eggs appear after the PR passes real-behavior proof. It is here for vibes, not verdicts: it does not change labels, ratings, merge decisions, or automation.
The shell reacts to review momentum: open follow-up work warms it up, re-review makes it wobble, and a clean final review lets it hatch.
Hatchability usually comes from sufficient real-behavior proof, no blocking P0/P1/P2 findings, no security attention needed, and clean correctness. A merged PR is already final, so merge makes the egg hatchable independently.
The hatch is seeded from this repository and PR number, so the same PR keeps the same creature; the reviewed head SHA can only change safe visual details.
Rarity is just collectible sparkle: 🥚 common, 🌱 uncommon, 💎 rare, ✨ glimmer, and 🌈 legendary.

clawsweeper · 2026-05-22T03:06:38Z

🦞✅
ClawSweeper merged this PR after the passing review.

Source: clawsweeper[bot]
Feedback: structured ClawSweeper verdict: pass (sha=193d3c0fdd5a42cc5291f4f279e80f872fdd1051)
Merge status: merged by ClawSweeper automerge
Merged at: 2026-05-22T05:04:42Z
Merge commit: cd1cae5be9ba

What merged:

The PR removes the auto-reply compaction-failure session reset hook, adds preserved-session recovery guidance for overflow/compaction failure paths, and updates focused tests, docs, and the changelog.
Reproducibility: yes. at source level with high confidence. Current main routes both embedded overflow paylo ... resetSessionAfterCompactionFailure, and the PR body includes before/after terminal proof of those branches.

Automerge notes:

PR branch already contained follow-up commit before automerge: fix(auto-reply): drop dead compaction reset hook
PR branch already contained follow-up commit before automerge: fix(auto-reply): preserve sessions after compaction failures

The automerge loop is complete.

Automerge progress:

2026-05-22 03:06:37 UTC review queued 193d3c0fdd5a (queued)

2026-05-22 05:04:23 UTC review passed 193d3c0fdd5a (structured ClawSweeper verdict: pass (sha=193d3c0fdd5a42cc5291f4f279e80f872fdd1...)

2026-05-22 05:04:44 UTC merged 193d3c0fdd5a (merged by ClawSweeper automerge)

clawsweeper · 2026-05-22T03:14:49Z

🦞✅
ClawSweeper is pausing this repair loop for human review.

Source: clawsweeper[bot]
Reason: Automerge is armed, but the remaining blocker is a maintainer proof judgment that automation cannot supply for the contributor's runtime environment.; Cleared: Cleared: the diff touches auto-reply control flow, tests, docs, and changelog only, with no dependency, CI, secret, permission, package-resolution, or external code-execution change. (sha=193d3c0fdd5a42cc5291f4f279e80f872fdd1051)

Why human review is needed:
This item has security-sensitive risk. ClawSweeper is pausing instead of making an autonomous change that could affect trust, credentials, permissions, or exposure.

What the maintainer can do as a next step:
If the maintainer accepts the current risk and wants ClawSweeper to continue merge gates, comment @clawsweeper approve. If the security-sensitive detail still needs changes, describe the safe path or push the fix, then comment @clawsweeper automerge. If the risk should not be automated, keep the PR paused for manual review or comment @clawsweeper stop.

I added clawsweeper:human-review and left the final call with a maintainer.

joshavant · 2026-05-22T04:31:48Z

@clawsweeper re-review

clawsweeper · 2026-05-22T04:33:44Z

🦞🧹
ClawSweeper could not start a re-review for this item.

Reason: re-review requires an open issue or PR.

joshavant · 2026-05-22T04:45:49Z

Updated the PR body with supplemental maintainer before/after proof for the requested session-state and delivery-path evidence.

The added proof shows pre-fix origin/main rewrites agent:main:main from session-before to session-after and calls both updateSessionStore and replyOperation.updateSessionId; the PR head keeps session-before, makes zero session-store/updateSessionId calls, and returns a deliverDespiteSourceReplySuppression: true failure payload with retry, /compact, and /new guidance for both embedded overflow payloads and thrown compaction failures.

@clawsweeper re-review

openclaw-barnacle · 2026-05-22T04:49:16Z

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

…red production Restoration commit bbecf06 reverted PR openclaw#70479 (preserve-session-on-compaction) and re-introduced session-system-events as a {text, forceSenderIsOwnerFalse} block, but left three test/production sites stale. - get-reply-run.media-only.test.ts: mock drainFormattedSystemEventBlock (replaces the string-returning legacy helper that get-reply-run no longer imports) and return the new block shape from each mockResolvedValueOnce. - agent-runner-execution.test.ts: override resetSessionAfterCompactionFailure to true in the two compaction-failure tests so the restored recovery branch is actually taken, and update expected reply text to match the restored 'I've reset our conversation to start fresh' copy. - sessions/store-cache.ts: stamp storedAt on serialized-cache entries and drop them from getSerializedSessionStore once they outlive the session store TTL, restoring the parity-with-object-cache assertion added in the restoration commit.

…w#70479) Summary: - The PR removes the auto-reply compaction-failure session reset hook, adds preserved-session recovery guidance for overflow/compaction failure paths, and updates focused tests, docs, and the changelog. - Reproducibility: yes. at source level with high confidence. Current main routes both embedded overflow paylo ... resetSessionAfterCompactionFailure, and the PR body includes before/after terminal proof of those branches. Automerge notes: - PR branch already contained follow-up commit before automerge: fix(auto-reply): drop dead compaction reset hook - PR branch already contained follow-up commit before automerge: fix(auto-reply): preserve sessions after compaction failures Validation: - ClawSweeper review passed for head 193d3c0. - Required merge gates passed before the squash merge. Prepared head SHA: 193d3c0 Review: openclaw#70479 (comment) Co-authored-by: FullerStackDev <263060202+fuller-stack-dev@users.noreply.github.com> Co-authored-by: vincentkoc <25068+vincentkoc@users.noreply.github.com> Co-authored-by: clawsweeper[bot] <274271284+clawsweeper[bot]@users.noreply.github.com> Approved-by: takhoffman Co-authored-by: takhoffman <781889+takhoffman@users.noreply.github.com>

openclaw-barnacle Bot added docs Improvements or additions to documentation size: S labels Apr 23, 2026

fuller-stack-dev force-pushed the fix/preserve-session-on-compaction-failure branch from 3befdda to 3367448 Compare April 23, 2026 07:48

clawsweeper Bot mentioned this pull request Apr 27, 2026

[Bug]: Context overflow / compaction can orphan agent:main:main and silently rotate WebChat to a new session #70472

Closed

openclaw-barnacle Bot added size: M and removed size: S labels Apr 27, 2026

clawsweeper Bot mentioned this pull request Apr 29, 2026

fix(agents): derive overflow budgets from provider errors #70473

Merged

openclaw-barnacle Bot added the triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. label May 20, 2026

clawsweeper Bot added the merge-risk: 🚨 session-state 🚨 May lose, corrupt, stale, or mis-associate session, agent, or context state. label May 20, 2026

fuller-stack-dev force-pushed the fix/preserve-session-on-compaction-failure branch from 74a145f to 9152393 Compare May 22, 2026 02:37

joshavant self-assigned this May 22, 2026

openclaw-barnacle Bot added the stale Marked as stale due to inactivity label May 22, 2026

clawsweeper Bot merged commit cd1cae5 into openclaw:main May 22, 2026
162 of 168 checks passed

github-actions Bot mentioned this pull request May 22, 2026

📡 Upstream Digest — 2026-05-22 07:37 UTC curtismercier/openclaw-mods#914

Open

Uh oh!

Conversation

fuller-stack-dev commented Apr 23, 2026 • edited by joshavant Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Real behavior proof

Notes

Uh oh!

greptile-apps Bot commented Apr 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Uh oh!

fuller-stack-dev commented Apr 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

fuller-stack-dev commented Apr 23, 2026

Uh oh!

fuller-stack-dev commented Apr 23, 2026

Uh oh!

martingarramon commented Apr 23, 2026

Uh oh!

vincentkoc commented Apr 27, 2026

Uh oh!

clawsweeper Bot commented Apr 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented May 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Hatch command

Uh oh!

clawsweeper Bot commented May 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented May 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

joshavant commented May 22, 2026

Uh oh!

clawsweeper Bot commented May 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

joshavant commented May 22, 2026

Uh oh!

openclaw-barnacle Bot commented May 22, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

fuller-stack-dev commented Apr 23, 2026 •

edited by joshavant

Loading

greptile-apps Bot commented Apr 23, 2026 •

edited

Loading

fuller-stack-dev commented Apr 23, 2026 •

edited

Loading

clawsweeper Bot commented Apr 27, 2026 •

edited

Loading

clawsweeper Bot commented May 20, 2026 •

edited

Loading

clawsweeper Bot commented May 22, 2026 •

edited

Loading

clawsweeper Bot commented May 22, 2026 •

edited

Loading

clawsweeper Bot commented May 22, 2026 •

edited

Loading