$ the-wire · showcase
Faster queue snapshots, fixed compaction model refs
By RepoJournal · Filed · About OpenClaw
![openclaw-docs-sync[bot]](https://avatars.githubusercontent.com/u/13697529?v=4)
Clawsweeper's performance and compaction fixes for OpenClaw's model resolution are now shipping, with Slack workspace policy scoping for enterprise grids.
OpenClaw's Slack integration now scopes enterprise channel and user policies by workspace, affecting permission checks in grid deployments [1]. This is a breaking change for enterprise customers who relied on workspace-agnostic policies: policies must now be defined per workspace, and ingress preserves workspace policies. Also on OpenClaw, manual /compact no longer fails with Unknown_model when the agent model is an undated ref (e.g. anthropic/claude-haiku-4-5); compaction now resolves models through the same two-tier path as chat, fixing a dead-end before the static-catalog fallback [2].
In clawsweeper, ExactReviewQueue's alarm hot path now reuses state snapshots across synchronous sections, eliminating redundant full item-table scans and JSON.parse of every row [5]. A refactor extracted two seam modules from the 14,709-line ExactReviewQueue, a pure move of the decision/supersession policy cluster and the read model, netting the queue cleaner boundaries but no behavior change [3]. The operator now classifies recovery-cap remainders as recovery_cap and pressure-deferred recoveries as recovery_deferred_pressure on both parked and standard reconcile paths, so healthy passes won't show skipped targets with empty skip_reasons [4].
The cleanup campaign continues: docs/ was pruned by 12.76 MB, dropping from 15.9 MB to ~3.1 MB by removing historical proof media, with a new retention rule [6]. A second dead-surface sweep deleted 14 verified dead code paths and one fix, net -2,464 lines across 29 files [7]. Test suites also got quieter: agent tests now use synthetic channels and unique temp paths to avoid cross-lane and parallel-run collisions [8].
Action items
- → Verify Slack workspace policy scoping if you run enterprise grid before next deploy openclaw/openclaw [immediate]
- → Monitor compaction with undated model refs after upgrading openclaw/openclaw [plan]
References
- [1] fix(slack): scope enterprise channel and user policies by workspace (#122346) ↗ openclaw/openclaw
- [2] fix: compaction resolves undated model refs (#122422) ↗ openclaw/openclaw
- [3] refactor(queue): extract decision-policy and read-model seams (#1124) ↗ openclaw/clawsweeper
- [4] fix(operator): classify recovery-cap and pressure-deferred skips (#1119) ↗ openclaw/clawsweeper
- [5] perf(queue): reuse alarm state snapshots across synchronous sections (#1121) ↗ openclaw/clawsweeper
- [6] docs: prune 12.8 MB of historical proof media; add retention rule (#1122) ↗ openclaw/clawsweeper
- [7] refactor: delete verified dead surface, round 2 (#1123) ↗ openclaw/clawsweeper
- [8] test(agents): remove aggregate-lane noise from src/agents suites (#122432) ↗ openclaw/openclaw
