fix(mattermost): preserve text-block boundaries in draft preview (#87322) by yetval · Pull Request #87449 · openclaw/openclaw

yetval · 2026-05-27T23:21:34Z

What Problem This Solves

Fixes #87322. Mattermost block previews could overwrite earlier assistant/tool blocks during one streamed turn, and streaming.mode: "off" with blockStreaming: true incorrectly suppressed normal block delivery.

Why This Change Was Made

Treat each visible assistant/tool activity as an ordered Mattermost preview generation, while keeping repeated tool updates on the current tool post.
Serialize callback starts for Mattermost so fire-and-forget assistant, reasoning, and tool events cannot overtake one another.
Preserve Responses item identity and phase metadata through OpenAI-compatible and proxy streams so commentary and final-answer boundaries stay stable.
Reconcile cumulative terminal output against only successfully published assistant previews, including response-prefix handling, so final delivery sends only the unpublished remainder.
Keep the existing streaming/config contract; no new config keys or runtime compatibility path.

User Impact

With Mattermost block streaming, earlier text and tool-progress blocks remain visible in source order instead of being overwritten. Reasoning placeholders remain transient, response prefixes appear once, long blocks retain configured chunking, and completed turns do not repeat preview text. Preview-off blockStreaming: true again produces ordinary separate block replies.

Evidence

Exact head: e40a51294973a0123d7c68dd341a5d34db5957cf
Hosted CI: https://github.com/openclaw/openclaw/actions/runs/29095426048
Real behavior workflow: https://github.com/openclaw/openclaw/actions/runs/29095424449
Proof lane: approved exact-head secretless fork CI; an additional sanitized AWS lease was canceled before host assignment because it was not required.
Fresh autoreview: clean; no actionable findings (0.99 confidence on the final test-fix commit, plus clean whole-patch review before freeze).
Static checks: targeted oxfmt, oxlint, git diff --check, and plugin-SDK API baseline verification passed.

Known proof gap: no approved authenticated Mattermost deployment was available for exact-head channel/UI validation. The focused runtime/HTTP and regression coverage exercises the delivery contract, but server-side rendering, notifications, and mobile push were not observed.

Fixes #87322

clawsweeper · 2026-05-27T23:23:32Z

Codex review: stale review; fresh review needed.

Summary
The latest durable ClawSweeper review was for head b8f9f5f21f94454984a80fc37d5c9a3e605e6dd8, but the PR head is now e40a51294973a0123d7c68dd341a5d34db5957cf. Its old verdict and PR readiness labels are no longer current.

Next step
Run or wait for a fresh ClawSweeper review on the current PR head.

Review history (4 earlier review cycles)

reviewed 2026-06-29T14:33:06.767Z sha 7ce7fa0 :: needs maintainer review before merge. :: none
reviewed 2026-07-05T23:25:10.289Z sha b052429 :: needs maintainer review before merge. :: none
reviewed 2026-07-05T23:33:18.099Z sha b052429 :: needs maintainer review before merge. :: none
reviewed 2026-07-10T11:57:46.744Z sha c3fe056 :: needs maintainer review before merge. :: none

clawsweeper · 2026-05-27T23:28:05Z

ClawSweeper PR egg

✨ Hatched: 🥚 common Velvet Merge Sprite

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: sniffs out flaky tests.
Image traits: location branch lighthouse; accessory lint brush; palette plum, gold, and soft gray; mood calm; pose balancing on a branch marker; shell starlit enamel shell; lighting subtle sparkle highlights; background miniature CI buoys.
Share on X: post this hatch
Copy: My PR egg hatched a 🥚 common Velvet Merge Sprite 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.

yetval · 2026-05-27T23:38:22Z

@clawsweeper re-review

clawsweeper · 2026-05-27T23:38:24Z

🦞🧹
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/26545206906
Updated: 2026-05-27T23:47:19.029Z

yetval · 2026-05-28T00:01:09Z

@clawsweeper re-review

Address P1 finding (draft-stream.ts:74-75): forceNewMessage is now async and awaits loop.flush() before clearing streamPostId; controller noteBoundary is async and the inbound callbacks (onAssistantMessageStart, onReasoningEnd, onToolStart) await it. New flushes a pending in-flight create before forcing a new post regression test holds the first POST /posts response and fires the boundary while the create is still in flight — asserts the next block lands as a separate POST (not a PUT). PR body Evidence section now shows a delayed-transport run (120ms POST hold, fastBoundary: true) where each boundary fires before the prior POST resolves; the timestamps show forceNewMessage() (awaiting pending flush) → resolved ~120ms later → next POST. AFTER pass: 5 POST, 0 PUT. Head: 353f83b.

clawsweeper · 2026-05-28T00:01:12Z

🦞🧹
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/26546052377
Updated: 2026-05-28T00:09:50.601Z

clawsweeper · 2026-06-15T13:12:15Z

🦞🧹
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/27549327432
Updated: 2026-06-15T13:33:22.999Z

yetval · 2026-06-15T15:12:44Z

@clawsweeper re-review

Addressed the boundary gate race P1 (draft-stream.ts:78). Head is now d6c9f85491.

Root cause: noteBoundary() left hasStreamedContent true while it awaited forceNewMessage(), then reset it to false after the await. Because the embedded subscriber dispatches both noteBoundary() and noteUpdate() fire-and-forget (src/agents/embedded-agent-subscribe.ts:281,1193, embedded-agent-subscribe.handlers.messages.ts:144,573,687), a noteUpdate() for the next block could land while a prior boundary was still awaiting; the older boundary then reset the flag to false, swallowing the streamed-content signal, so the following boundary skipped forceNewMessage() and the block after it edited into the previous post.

Fix: noteBoundary() now returns early when nothing streamed, then clears hasStreamedContent synchronously before awaiting forceNewMessage(), so a noteUpdate() during the await is preserved for the next boundary and no boundary is skipped.

Proof refreshed on current head, exercising the fire-and-forget path over live loopback HTTP (production stream + controller, real fetch round-trips, 127.0.0.1 only, placeholder token):

Gate clobber: block A then a fire-and-forget boundary, a noteUpdate() arriving while that boundary is pending, then a second boundary and block C. BEFORE (c8c90e0df8, gate unfixed): 2 posts, block C PUTs onto block B's post-2 and block B is lost. AFTER: 3 distinct posts (post-1, post-2, post-3).
Synchronous detach (prior P1): a fire-and-forget boundary with a partial arriving before it settles. BEFORE: PUT overwrites post-1. AFTER: new post-2, post-1 preserved.

New regression splits the next boundary when a noteUpdate arrives while the prior boundary is pending plus the synchronous-detach regression. Local: draft-stream 20/20, full extensions/mattermost/src/mattermost/ 29 files / 377/377, oxlint + oxfmt clean. Full before/after captures and reproduction details are in the updated PR body.

clawsweeper · 2026-06-15T15:12:47Z

🦞🧹
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: Superseded
Detail: A newer re-review for this item started before this run finished, so GitHub cancelled this older run. Check the latest ClawSweeper run for the current result.
Run: https://github.com/openclaw/clawsweeper/actions/runs/27557849493
Updated: 2026-06-15T15:41:59.799Z

clawsweeper · 2026-06-15T15:39:16Z

🦞👀
ClawSweeper picked this up.

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

yetval · 2026-06-15T15:49:32Z

@clawsweeper re-review

Addressed the callback-ordering P1 at monitor.ts:1950. Head is now cfb5a1fd05.

The monitor's onAssistantMessageStart and onReasoningEnd awaited previewBoundaryController.noteBoundary() before resetting the per-turn local state (lastPartialText, progressDraft). Since the embedded subscriber dispatches these callbacks with void, the next block's onPartialReply could run during that await with lastPartialText still holding the previous block's text; updateDraftFromPartial then drops a new-block partial that is a prefix of the prior block as a stale repaint.

Fix: both callbacks now start the split, reset local state synchronously, then await (const boundarySettled = noteBoundary(); lastPartialText = ""; progressDraft.reset(); ...; await boundarySettled;).

Proof refreshed on the current head through the real subscribeEmbeddedAgentSession dispatch (no direct controller calls), redaction-safe (127.0.0.1, placeholder token). With block A's POST in flight and block B's only partial a prefix of block A: OLD ordering drops block B (1 post); NEW ordering preserves it (2 posts). Full before/after in the PR body, alongside the synchronous-detach and gate-clobber embedded-path captures.

Local: draft-stream 20/20, monitor.inbound-system-event 10/10, full extensions/mattermost/src/mattermost/ 29 files / 377/377, oxlint + oxfmt clean.

Remaining items I believe are maintainer decisions rather than code defects, and I have flagged them in the PR body for sign-off: (1) the intentional visible delivery-shape change (one edited post to multiple posts for block-mode and preview-off blockStreaming); (2) the additive exported DraftStreamLoop.takePending type; (3) a real self-hosted Mattermost UI capture, which I cannot produce from this environment (loopback + real embedded-callback proof is provided instead); (4) the separate error-response draft-deletion artifact noted on #87322, which is out of scope for this fix and looks like a follow-up.

clawsweeper · 2026-06-15T15:49:36Z

🦞🧹
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/27558914325
Updated: 2026-06-15T16:07:42.123Z

yetval · 2026-06-15T16:13:54Z

@clawsweeper re-review

Re-requesting a clean review of the current head cfb5a1fd05. The durable verdict comment was just overwritten out of order: the cfb5a1fd05 review completed at 16:07, but a stale re-review of an earlier head (its evidence cites 56186cfe2512, before the synchronous-detach, gate, and monitor-ordering fixes) finished at 16:09 and clobbered it.

All three fire-and-forget races the prior verdicts flagged are fixed on cfb5a1fd05, and the detach is now synchronous through the whole chain:

Stream forceNewMessage bumps the generation and clears streamPostId synchronously before awaiting (the in-flight-create seal is preserved via a generation-keyed record).
Controller noteBoundary clears hasStreamedContent synchronously before awaiting forceNewMessage.
Monitor onAssistantMessageStart / onReasoningEnd start the split, reset lastPartialText / progressDraft synchronously, then await (monitor.ts:1949-1956). Because noteBoundary() runs its synchronous part (including forceNewMessage's synchronous detach) before any await, a fire-and-forget partial that arrives next deterministically opens a fresh post; it cannot flush against the old post.

Proof on the current head goes through the real subscribeEmbeddedAgentSession dispatch (genuine void callbacks, not direct controller calls), redaction-safe (127.0.0.1, placeholder token); before/after captures for all three races are in the PR body. Local: draft-stream 20/20, monitor.inbound-system-event 10/10, full extensions/mattermost/src/mattermost/ 377/377, oxlint + oxfmt clean.

The remaining items I believe are maintainer decisions rather than code defects (flagged in the PR body for sign-off): the intentional visible delivery-shape change, the additive exported DraftStreamLoop.takePending type, a real self-hosted Mattermost UI capture I cannot produce from this environment, and the separate error-response draft-deletion artifact on #87322 (follow-up).

clawsweeper · 2026-06-15T16:23:26Z

🦞🧹
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: Failed
Detail: The targeted re-review did not finish cleanly. Check the workflow run for details.
Run: https://github.com/openclaw/clawsweeper/actions/runs/27623195613
Updated: 2026-06-16T14:33:09.433Z

yetval · 2026-06-16T19:54:58Z

@clawsweeper re-run

clawsweeper · 2026-06-16T19:55:01Z

🦞🧹
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: Failed
Detail: The targeted re-review timed out or failed while waiting for exact-review Codex capacity. Retry after capacity recovers.
Run: https://github.com/openclaw/clawsweeper/actions/runs/27659239397
Updated: 2026-06-17T01:40:01.049Z

yetval · 2026-06-17T01:17:58Z

@clawsweeper re-run

vincentkoc · 2026-06-17T05:12:15Z

Clownfish 🐠 reef update

Thanks for the contribution here. Clownfish gave this branch a little current boost so the original PR can stay the canonical swim lane instead of opening a replacement.

Source PR: #87449
Validation: pnpm check:changed
The useful context and credit stay on this branch. tidy swim, same contributor trail.

fish notes: model gpt-5.5, reasoning medium; reviewed against 55dcfbf.

yetval · 2026-06-17T23:21:07Z

@clawsweeper re-run

clawsweeper · 2026-06-17T23:21:11Z

🦞🧹
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/27733105225
Updated: 2026-06-18T03:27:23.611Z

yetval · 2026-06-18T01:57:32Z

@clawsweeper re-run

…nclaw#87322)

Co-authored-by: Yuval Dinodia <[email protected]>

steipete · 2026-07-10T13:31:53Z

Land-ready at exact head e40a51294973a0123d7c68dd341a5d34db5957cf.

Improved the contributor fix into serialized Mattermost preview generations: visible text/tool blocks remain ordered, same-tool progress stays grouped, reasoning placeholders stay transient, and terminal output removes only successfully published assistant prefixes.
Preserved preview-off block delivery, configured chunking, live response-prefix rendering, final thread participation, and OpenAI-compatible Responses item/phase identity through proxy paths.
Hosted CI: https://github.com/openclaw/openclaw/actions/runs/29095426048
Real Behavior proof: https://github.com/openclaw/openclaw/actions/runs/29095424449
Proof lane: approved exact-head secretless fork CI; an additional sanitized AWS lease was canceled before host assignment because it was not required.
Fresh autoreview and focused compatibility/concurrency/terminal reviews: clean; no actionable findings.

Known gap: no approved authenticated Mattermost deployment was available for exact-head channel/UI validation. Source-blind behavior validation is blocked on the missing Mattermost URL/token and deployed runtime; server-side rendering, notifications, and mobile push were not observed.

Thanks @yetval for the original report-driven fix and persistence through review.

steipete · 2026-07-10T13:32:16Z

Merged via squash.

Prepared head SHA: e40a51294973a0123d7c68dd341a5d34db5957cf
Landed commit: b7ab62c2cf113316350122b13ec91a068a3f5233

openclaw-barnacle Bot added channel: mattermost Channel integration: mattermost size: S triage: mock-only-proof Candidate: PR proof only shows tests, mocks, snapshots, lint, typecheck, or CI. labels May 27, 2026

clawsweeper Bot added rating: 🦪 silver shellfish Thin PR readiness signal; proof, validation, or implementation needs work. status: 📣 needs proof The PR needs real behavior proof before ClawSweeper can clear the contributor ask. labels May 27, 2026

yetval force-pushed the fix/87322-mattermost-block-preview-boundaries branch from d969eef to 353f83b Compare May 27, 2026 23:59

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

openclaw-barnacle Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 28, 2026

yetval and others added 13 commits July 10, 2026 14:15

fix(mattermost): preserve text-block boundaries in draft preview (ope…

31fbcb5

…nclaw#87322)

fix(mattermost): preserve block preview boundaries

066c4c6

fix(mattermost): reset progress at tool boundaries

8ce087c

docs(plugin-sdk): refresh API baseline

75e86ec

fix(mattermost): group parallel tool previews

1ce2f1c

fix(mattermost): stage tool previews before awaits

8ae53be

fix(mattermost): satisfy tool boundary return contract

c8b975e

fix(mattermost): serialize preview block generations

72782bf

Co-authored-by: Yuval Dinodia <[email protected]>

fix(mattermost): preserve complete block previews

c3a4781

test(agents): align streaming assertions

5ecafa3

test(telegram): align reply pipeline mock

bdb71c9

fix(plugin-sdk): keep reply prefix options compatible

13e1f71

test(mattermost): exercise threaded final participation

e40a512

Uh oh!

Uh oh!

Conversation

yetval commented May 27, 2026 • edited by steipete 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

Uh oh!

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

Uh oh!

Uh oh!

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

Uh oh!

Hatch command

Uh oh!

yetval commented May 27, 2026

Uh oh!

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

Uh oh!

Uh oh!

yetval commented May 28, 2026

Uh oh!

clawsweeper Bot commented May 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yetval commented Jun 15, 2026

Uh oh!

clawsweeper Bot commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented Jun 15, 2026

Uh oh!

yetval commented Jun 15, 2026

Uh oh!

clawsweeper Bot commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yetval commented Jun 15, 2026

Uh oh!

clawsweeper Bot commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yetval commented Jun 16, 2026

Uh oh!

clawsweeper Bot commented Jun 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yetval commented Jun 17, 2026

Uh oh!

vincentkoc commented Jun 17, 2026

Uh oh!

yetval commented Jun 17, 2026

Uh oh!

clawsweeper Bot commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yetval commented Jun 18, 2026

Uh oh!

steipete commented Jul 10, 2026

Uh oh!

steipete commented Jul 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

yetval commented May 27, 2026 •

edited by steipete

Loading

clawsweeper Bot commented May 27, 2026 •

edited

Loading

clawsweeper Bot commented May 27, 2026 •

edited

Loading

clawsweeper Bot commented May 27, 2026 •

edited

Loading

clawsweeper Bot commented May 28, 2026 •

edited

Loading

clawsweeper Bot commented Jun 15, 2026 •

edited

Loading

clawsweeper Bot commented Jun 15, 2026 •

edited

Loading

clawsweeper Bot commented Jun 15, 2026 •

edited

Loading

clawsweeper Bot commented Jun 15, 2026 •

edited

Loading

clawsweeper Bot commented Jun 16, 2026 •

edited

Loading

clawsweeper Bot commented Jun 17, 2026 •

edited

Loading