fix(compaction): add circuit breaker to stop token burn when summarizer unavailable by tanshanshan · Pull Request #86900 · openclaw/openclaw

tanshanshan · 2026-05-26T12:59:30Z

Summary

Fixes #58838. Staged compaction no longer burns requests across every split when summarization repeatedly falls back generically.

The maintainer rewrite:

classifies real summaries and generic fallbacks with a discriminated internal result;
opens the circuit after two consecutive generic split fallbacks;
cancels the incomplete compaction so unprocessed transcript messages remain intact;
resets the consecutive counter after a successful split while retaining degraded state through merge;
preserves a known prior summary when an isolated generic fallback prevents redistillation.

What Problem This Solves

When the summarizer is persistently unavailable, staged compaction can exhaust the retry/fallback path for every split. That multiplies token and latency cost. Stopping the loop by returning the summaries collected so far is unsafe: later transcript chunks were never processed, yet compaction would commit and drop them. This fix bounds the storm and cancels the incomplete compaction so the transcript survives.

Evidence

node scripts/run-vitest.mjs src/agents/compaction.circuit-breaker.test.ts src/agents/compaction.identifier-preservation.test.ts src/agents/compaction.summarize-fallback.test.ts src/agents/compaction-partial-summary.test.ts src/agents/agent-hooks/compaction-safeguard.test.ts passed 5 files and 133 tests. The breaker test proves exactly two calls before circuit-open and rejection; the intermittent test proves counter reset plus degraded merge state; safeguard coverage proves prior-summary retention. Fresh full-diff autoreview is clean.

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

Fixes fix(compaction): retry loop can burn tokens when summarizer model is unavailable #58838
Related current-main lock ownership: fix(agents): scope compaction lock reentry to logical writer #88919 (inspected; orthogonal)
This PR fixes a bug or regression

Motivation

A persistently unavailable summarizer could make every staged split exhaust its retry/fallback path, multiplying token and latency cost. Returning an incomplete partial summary would also risk committing away splits that were never processed. The breaker must stop the request storm and preserve the source transcript.

Real Behavior Proof

Focused regression command:

node scripts/run-vitest.mjs src/agents/compaction.circuit-breaker.test.ts src/agents/compaction.identifier-preservation.test.ts src/agents/compaction.summarize-fallback.test.ts src/agents/compaction-partial-summary.test.ts src/agents/agent-hooks/compaction-safeguard.test.ts

Result: 5 test files passed, 133 tests passed.

The circuit-breaker regression proves two consecutive generic fallbacks make only two summarization calls and reject the incomplete compaction. The intermittent sequence proves a successful split resets the consecutive counter, the merge still runs, and the typed result remains degraded. Safeguard coverage proves a prior summary survives degraded redistillation.

Scoped changed-gate command:

node scripts/check-changed.mjs -- src/agents/compaction.ts src/agents/compaction.circuit-breaker.test.ts src/agents/agent-hooks/compaction-safeguard.ts src/agents/agent-hooks/compaction-safeguard.test.ts

Conflict markers, max-lines ratchet, changelog attribution, wildcard export guards, duplicate coverage, dependency pins, and formatting passed. The delegated run then encountered pre-existing Plugin SDK API baseline drift on current main; this PR changes no Plugin SDK path or export. Exact-head hosted CI remains the landing gate.

Fresh full-diff autoreview result: clean, no accepted/actionable findings.

Root Cause

The staged loop treated a generic fallback string like a usable summary and had no typed signal for repeated failure. That made it impossible to bound repeated calls without brittle string matching, and an early partial return could silently discard unprocessed transcript chunks.

Regression Test Plan

Two consecutive generic fallbacks open the circuit after two calls and reject the compaction.
A successful split resets the consecutive counter and allows the merge.
Any generic split keeps the merged result typed as degraded.
A degraded result preserves the previous summary at the safeguard owner.
Existing fallback, partial-summary, and identifier-preservation tests remain green.

User-visible / Behavior Changes

During a sustained summarizer outage, OpenClaw stops staged summarization after two consecutive generic fallbacks and cancels that compaction attempt. The session transcript remains available for a later retry instead of being replaced by an incomplete placeholder summary.

Release-note context: compaction now stops repeated staged fallback storms while preserving session history when the summarizer is unavailable. Thanks @tanshanshan.

Security Impact (required)

New permissions/capabilities? No
Secrets/tokens handling changed? No
New/changed network calls? No
Command/tool execution surface changed? No
Data access scope changed? No

Human Verification (required)

Verified scenarios: consecutive fallback circuit-open; intermittent fallback recovery; degraded merge classification; prior-summary retention.
Edge cases checked: partial model output remains a real summary; aborts still propagate; single-pass and custom-provider paths remain unchanged.
Not verified: a deliberate live production-provider outage. Deterministic regression tests exercise the same internal failure states without spending external tokens.

Review Conversations

I replied to or resolved every bot review conversation I addressed in this PR.
I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

Backward compatible? Yes
Config/env changes? No
Migration needed? No

Risks and Mitigations

Risk: stopping early could lose current transcript context if an incomplete summary were committed. Mitigation: circuit-open throws into the safeguard's existing cancel path, which retains the original transcript. No config, storage schema, or public API changes.

clawsweeper · 2026-05-26T13:03:57Z

Codex review: needs real behavior proof before merge. Reviewed July 17, 2026, 3:14 AM ET / 07:14 UTC.

Summary
The PR adds typed compaction summary outcomes, aborts staged compaction after two consecutive generic fallbacks, and preserves an existing summary when degraded redistillation fails.

PR surface: Source +52, Tests +142. Total +194 across 4 files.

Reproducibility: yes. at source level: persistent summarizer failures traverse each staged split's retry and generic-fallback path, and the focused tests deterministically exercise that condition. A failing live current-main run was not performed during this review.

Review metrics: 1 noteworthy metric.

Fixed Fallback Cutoff: 1 threshold added: 2 consecutive generic fallbacks. This fixed cutoff determines when staged compaction abandons later transcript splits and enters the session-preserving cancellation path.

Root-cause cluster
Relationship: fixed_by_candidate
Canonical: #58838
Summary: This PR is the active candidate fix for the canonical staged-compaction token-burn issue; earlier breaker PRs closed unmerged, while the merged partial-summary and lock fixes address adjacent failure modes.

Members:

canonical: fix(compaction): retry loop can burn tokens when summarizer model is unavailable #58838 - The issue directly reports repeated compaction retry token burn and remains open with this PR as its closing candidate.
superseded: fix(compaction): add circuit breaker for retry loops #58846 - This earlier time-based circuit-breaker proposal targeted the same root cause but closed unmerged.
superseded: fix(compaction): add circuit breaker for retry loops #58857 - This second earlier breaker proposal also targeted the same root cause and closed unmerged.
partial_overlap: fix(compaction): preserve partial summary on mid-chain chunk failure #82952 - The merged PR preserves partial progress after a mid-chain failure but does not bound repeated staged generic fallbacks.
adjacent_distinct: fix(agents): scope compaction lock reentry to logical writer #88919 - The merged PR fixes compaction lock ownership and self-deadlock rather than summarizer retry storms.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🦪 silver shellfish
Proof: 🦪 silver shellfish
Patch quality: 🐚 platinum hermit
Result: blocked until real behavior proof from a real setup is added.

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

Rank-up moves:

[P1] Add current-head terminal or log proof showing the breaker stops later calls and cancellation preserves the source transcript; redact keys, endpoints, identifiers, and other private data.
[P1] Refresh against current main if needed and let exact-head CI finish.

Proof guidance:

[P1] Needs real behavior proof before merge: The rewritten head is supported by focused mocked regression tests but no current-head runtime or compiled fault-injection proof; add redacted terminal or log evidence to the PR body, which should trigger a fresh review, or ask a maintainer to comment @clawsweeper re-review. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

[P1] The changed head has not demonstrated in a real runtime or compiled fault-injection setup that two generic fallbacks stop subsequent requests and reach the transcript-preserving cancellation owner.
[P1] Merging changes session-state failure behavior: sustained summarizer outages now abort compaction after a fixed cutoff instead of continuing later splits and merge summarization.
[P1] The PR is behind current main and exact-head CI remains incomplete, so final integration behavior is not yet established.

Maintainer options:

Prove cancellation on the rewritten head (recommended)
Run a real or compiled fault-injection compaction that triggers two generic fallbacks and shows both bounded calls and unchanged source transcript before merging.
Apply an explicit proof override
A maintainer may accept the deterministic regression suite as sufficient for this internal path while explicitly owning the unobserved runtime cancellation risk.

Next step before merge

[P1] The existing branch already contains the focused repair; the remaining work is current-head real-behavior proof, exact-head CI, and normal maintainer merge review rather than another fix PR.

Security
Cleared: The diff changes internal compaction control flow and tests without new dependencies, permissions, secrets, downloads, workflow privileges, or package execution surfaces.

Review details

Best possible solution:

Keep the typed internal result and incomplete-compaction cancellation design, add current-head runtime or compiled fault-injection output proving that later calls stop and the original transcript remains intact, then merge after exact-head checks pass.

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

Yes at source level: persistent summarizer failures traverse each staged split's retry and generic-fallback path, and the focused tests deterministically exercise that condition. A failing live current-main run was not performed during this review.

Is this the best way to solve the issue?

Yes. Typed internal outcomes plus cancellation of an incomplete compaction is narrower and safer than string matching, committing partial coverage, or adding a process-wide time-based breaker.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against b64242a4a99a.

Label changes

Label changes:

add rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🐚 platinum hermit.
add status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The rewritten head is supported by focused mocked regression tests but no current-head runtime or compiled fault-injection proof; add redacted terminal or log evidence to the PR body, which should trigger a fresh review, or ask a maintainer to comment @clawsweeper re-review. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.
remove proof: sufficient: Current real behavior proof status is mock_only, not sufficient.
remove rating: 🐚 platinum hermit: Current PR rating is rating: 🦪 silver shellfish, so this older rating label is no longer current.
remove status: 👀 ready for maintainer look: Current PR status label is status: 📣 needs proof.

Label justifications:

P2: The PR addresses bounded but meaningful compaction token burn and session continuity during sustained summarizer failures.
merge-risk: 🚨 session-state: The patch changes when compaction aborts and which summary state survives, so an incorrect caller transition could lose, duplicate, or stale session context.
rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🐚 platinum hermit.
status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The rewritten head is supported by focused mocked regression tests but no current-head runtime or compiled fault-injection proof; add redacted terminal or log evidence to the PR body, which should trigger a fresh review, or ask a maintainer to comment @clawsweeper re-review. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence reviewed

PR surface:

Source +52, Tests +142. Total +194 across 4 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	2	76	24	+52
Tests	2	205	63	+142
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	4	281	87	+194

What I checked:

Typed fallback state: The PR head distinguishes usable summaries from generic fallbacks and defines a fixed two-fallback cutoff without adding public API or configuration surface. (src/agents/compaction.ts:42, 77024951dc43)
Transcript-preserving cancellation: The rewritten branch rejects an incomplete circuit-open compaction rather than committing summaries that omit unprocessed transcript splits. (src/agents/compaction.ts:265, 77024951dc43)
Prior-summary retention: The safeguard unwraps successful results and retains the known previous summary when redistillation returns a generic fallback. (src/agents/agent-hooks/compaction-safeguard.ts:397, 77024951dc43)
Focused regression coverage: The added test injects summarizer failures and checks the staged circuit-breaker path; the PR reports five focused files and 133 passing tests. (src/agents/compaction.circuit-breaker.test.ts:1, 77024951dc43)
Current-head proof gap: The rewritten PR body explicitly says a deliberate runtime provider-outage scenario was not verified; the supplied evidence for this head consists of deterministic regression tests and mocks. (77024951dc43)
Maintainer adoption: steipete assigned the PR and rewrote the branch into four focused compaction commits on July 17, 2026, providing current maintainer sponsorship for the implementation direction. (src/agents/compaction.ts:42, 1581286021bd)

Likely related people:

steipete: Introduced the staged-compaction helper in earlier history and authored the current maintainer rewrite across the breaker, degraded-state, and cancellation paths. (role: feature owner and recent rewriter; confidence: high; commits: ae0741a3466c, 1581286021bd, c59249f9c246; files: src/agents/compaction.ts, src/agents/agent-hooks/compaction-safeguard.ts, src/agents/compaction.circuit-breaker.test.ts)
amabito: Added the nearby per-chunk summary retry behavior central to the reported token-burn path and its regression coverage. (role: retry behavior contributor; confidence: high; commits: 068b9c97498d; files: src/agents/compaction.ts, src/agents/compaction.retry.test.ts)
SebTardif: Authored the merged partial-summary recovery work that established nearby fallback and incomplete-summary behavior used by this fix. (role: recent adjacent fallback contributor; confidence: high; commits: f0061ddc54c9; files: src/agents/compaction.ts, src/agents/compaction-partial-summary.test.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 (1 earlier review cycle)

reviewed 2026-06-24T18:19:27.405Z sha 0188074 :: needs maintainer review before merge. :: none

clawsweeper · 2026-05-26T13:16:57Z

ClawSweeper PR egg: 🎁 locked until real behavior proof passes.

Details

No creature or rarity is rolled until proof passes.
Eggs are collectible flavor only; they do not affect labels, ratings, merge decisions, or automation.

martingarramon

Resetting consecutiveChunkFailures on each successful chunk is correct — the breaker targets a persistently unavailable summarizer, not a single hard-to-summarize chunk.

One accuracy issue in the circuit-breaker catch block. When the breaker trips at the second consecutive failure (chunk at zero-based index N), completedChunks = chunks.indexOf(chunk) equals N. But the preceding chunk (index N-1) also failed and was not summarized, so the annotation:

chunks 1-${completedChunks} of ${chunks.length} were summarized

overcounts by one — it should read chunks 1-${completedChunks - 1}. The accumulated summary content is correct (it only updates on success); the annotation text is what's off.

The previousSummary ?? DEFAULT_SUMMARY_FALLBACK path when no chunk has succeeded handles the cold-start unavailable case cleanly.

clawsweeper · 2026-05-27T01:58:19Z

🦞🧹
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 targeted re-review finished, the durable review comment was updated, and the synced verdict was routed.
Run: https://github.com/openclaw/clawsweeper/actions/runs/26486161824
Updated: 2026-05-27T02:04:45.967Z

RomneyDa · 2026-05-29T06:22:55Z

Heads up: this PR needs to be updated against current main before the new required Dependency Guard check can pass.

clawsweeper · 2026-06-03T02:29:29Z

🦞🧹
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.

clawsweeper · 2026-06-03T02:45:42Z

🦞🧹
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 targeted re-review finished, the durable review comment was updated, and the synced verdict was routed.
Run: https://github.com/openclaw/clawsweeper/actions/runs/26928044524
Updated: 2026-06-04T03:20:42.609Z

clawsweeper · 2026-06-03T07:22:51Z

🦞👀
ClawSweeper picked this up.

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

Re-review progress:

State: Complete
Detail: The targeted re-review finished, the durable review comment was updated, and the synced verdict was routed.
Run: https://github.com/openclaw/clawsweeper/actions/runs/26869993909
Updated: 2026-06-03T07:29:58.247Z

clawsweeper · 2026-06-05T09:34:00Z

🦞🧹
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 targeted re-review finished, the durable review comment was updated, and the synced verdict was routed.
Run: https://github.com/openclaw/clawsweeper/actions/runs/27103022955
Updated: 2026-06-07T20:08:06.190Z

openclaw-barnacle · 2026-07-15T04:22:53Z

This pull request has been automatically marked as stale due to inactivity.
Please add updates or it will be closed.

Co-authored-by: tanshanshan <[email protected]>

steipete · 2026-07-17T07:18:38Z

Merged via squash.

Prepared head SHA: 77024951dc43b52bab847301b5a7167e187976e9
Landed commit: 473dfc2e51b89ccd3907a356b7e5f5801d1818d5

openclaw-barnacle Bot added agents Agent runtime and tooling size: S triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 26, 2026

openclaw-barnacle Bot added proof: supplied External PR includes structured after-fix real behavior proof. and removed triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 26, 2026

martingarramon reviewed May 26, 2026

View reviewed changes

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

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

This comment was marked as low quality.

Sign in to view

openclaw-barnacle Bot added size: L size: M and removed size: M size: L labels May 27, 2026

clawsweeper Bot added rating: 🦪 silver shellfish Thin PR readiness signal; proof, validation, or implementation needs work. and removed rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. labels May 27, 2026

openclaw-barnacle Bot added proof: supplied External PR includes structured after-fix real behavior proof. and removed triage: mock-only-proof Candidate: PR proof only shows tests, mocks, snapshots, lint, typecheck, or CI. labels May 28, 2026

This comment was marked as low quality.

Sign in to view

RomneyDa added dependency-guard-backfill and removed dependency-guard-backfill labels May 29, 2026

This comment was marked as low quality.

Sign in to view

clawsweeper Bot mentioned this pull request Jun 10, 2026

fix(compaction): retry loop can burn tokens when summarizer model is unavailable #58838

Closed

steipete and others added 5 commits July 17, 2026 08:09

fix(compaction): stop repeated staged fallbacks

1581286

Co-authored-by: tanshanshan <[email protected]>

fix(compaction): preserve degraded summary state

c59249f

refactor(compaction): keep result state internal

c396618

fix(compaction): cancel incomplete circuit-open runs

3920077

chore: refresh PR checks

7702495

github-actions Bot mentioned this pull request Jul 17, 2026

📡 Upstream Digest — 2026-07-17 07:54 UTC curtismercier/openclaw-mods#1320

Open

Yigtwxx mentioned this pull request Jul 17, 2026

fix(compaction): previous summary is duplicated when a later split degrades #109828

Merged

Uh oh!

Uh oh!

Conversation

tanshanshan commented May 26, 2026 • edited by steipete Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

What Problem This Solves

Evidence

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

Motivation

Real Behavior Proof

Root Cause

Regression Test Plan

User-visible / Behavior Changes

Security Impact (required)

Human Verification (required)

Review Conversations

Compatibility / Migration

Risks and Mitigations

Uh oh!

clawsweeper Bot commented May 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented May 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

martingarramon left a comment

Choose a reason for hiding this comment

Uh oh!

This comment was marked as low quality.

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

Uh oh!

Uh oh!

This comment was marked as low quality.

RomneyDa commented May 29, 2026

Uh oh!

This comment was marked as low quality.

This comment was marked as low quality.

clawsweeper Bot commented Jun 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as low quality.

clawsweeper Bot commented Jun 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as low quality.

clawsweeper Bot commented Jun 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented Jun 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as low quality.

openclaw-barnacle Bot commented Jul 15, 2026

Uh oh!

steipete commented Jul 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

tanshanshan commented May 26, 2026 •

edited by steipete

Loading

clawsweeper Bot commented May 26, 2026 •

edited

Loading

clawsweeper Bot commented May 26, 2026 •

edited

Loading

clawsweeper Bot commented May 27, 2026 •

edited

Loading

clawsweeper Bot commented Jun 3, 2026 •

edited

Loading

clawsweeper Bot commented Jun 3, 2026 •

edited

Loading

clawsweeper Bot commented Jun 3, 2026 •

edited

Loading

clawsweeper Bot commented Jun 5, 2026 •

edited

Loading