fix(ui): deduplicate assistant messages on chat.status final event by SebTardif · Pull Request #86646 · openclaw/openclaw

SebTardif · 2026-05-25T21:36:06Z

Summary

Fix duplicate assistant messages appearing in WebChat when loadChatHistory and the final chat event race.

Guard: isTailDuplicate uses messageDisplaySignature to detect when the same message arrives via both history load and the live event stream.

Cross-run preservation: The dedup guard is applied only to same-run final/aborted events (the real race condition). Cross-run finals (sub-agent announcements, separate runs) are always appended because they may legitimately produce the same text. The renderer's existing duplicateCount collapse handles visual dedup for consecutive identical messages.

Stale stream fallback prevention: When a duplicate terminal payload is detected (final or aborted), the message is treated as fully handled. The code no longer falls through to the stream fallback, which could contain stale or different text from an earlier delta burst.

Changed files

ui/src/ui/controllers/chat.ts - isTailDuplicate helper, guards on same-run final/aborted/stream-fallback paths; cross-run path always appends; duplicate terminal payloads skip stream fallback
ui/src/ui/controllers/chat.test.ts - 7 new tests: same-run dedup, cross-run preservation, stream-fallback dedup, aborted dedup, stale-stream-on-final-duplicate, stale-stream-on-aborted-duplicate

Verification

Terminal: pnpm vitest run ui/src/ui/controllers/chat.test.ts

 ✓ |unit-ui| ui/src/ui/controllers/chat.test.ts (98 tests) 22ms

 Test Files  1 passed (1)
      Tests  98 passed (98)

Behavioral checks:

Same-run final with matching tail -> skipped (race condition dedup)
Same-run final duplicate + stale chatStream -> stream text NOT appended (P2 fix)
Same-run aborted duplicate + stale chatStream -> stream text NOT appended (P2 fix)
Cross-run final with matching tail -> appended (legitimate repeat, renderer handles display collapse)
Stream-fallback with matching tail -> skipped (same-run race)
Aborted message with matching tail -> skipped (same-run race)
Different content -> always appended regardless of run

Closes #85771

Real behavior proof

Behavior addressed: (1) Duplicate assistant messages appear in WebChat when loadChatHistory and the final chat event race. The isTailDuplicate guard detects when the same message arrives via both paths. (2) When the duplicate terminal message was detected, the code previously fell through to the stream fallback. If chatStream contained a different partial or stale value from an earlier delta burst, that stale text was appended instead of treating the duplicate terminal as fully handled. Both the final and aborted branches had this issue.
Real environment tested: macOS 15.5, Node 26.0.0, OpenClaw source checkout at /tmp/wt-86646. UI code exercises the production handleChatEvent function from ui/src/ui/controllers/chat.ts via npx tsx.
Exact steps or command run after this patch:

Step 1. Ran the full test suite to confirm all 98 tests pass (including 2 new stale-stream tests):

cd /tmp/wt-86646 && npx vitest run ui/src/ui/controllers/chat.test.ts

Step 2. Ran a proof script that imports the production handleChatEvent and exercises three scenarios:

cd /tmp/wt-86646 && npx tsx /tmp/proof-86646.ts

The proof script creates real ChatState objects with stale chatStream values and duplicate terminal payloads, then asserts the stale stream text is never appended.

Evidence after fix:

Test suite output (98 tests, all passing):

$ npx vitest run ui/src/ui/controllers/chat.test.ts
 ✓ |unit-ui| ui/src/ui/controllers/chat.test.ts (98 tests) 22ms
 Test Files  1 passed (1)
      Tests  98 passed (98)

Production proof output:

$ npx tsx /tmp/proof-86646.ts

=== Test 1: duplicate final with stale chatStream ===
  PASS  returns 'final'
  PASS  chatMessages length stays 1
  PASS  stale stream text NOT appended
  PASS  original message preserved

=== Test 2: duplicate aborted with stale chatStream ===
  PASS  returns 'aborted'
  PASS  chatMessages length stays 1
  PASS  stale stream text NOT appended
  PASS  original message preserved

=== Test 3: cross-run final still appends new message ===
  PASS  chatMessages length is 2
  PASS  new cross-run message appended
  PASS  active stream preserved

=== Summary: 11 passed, 0 failed ===

Observed result after fix: The production handleChatEvent correctly handles duplicate terminal payloads without falling through to the stale stream fallback. When a final or aborted payload matches the chat tail (history-load/final-event race), the message is treated as "already visible" and skipped, with no stream text appended. When a cross-run final arrives from a different runId, it is always appended regardless of tail content. The fix restructures the conditional nesting so that isTailDuplicate returning true stays within the "valid terminal message" branch instead of falling through to else if (chatStream).
What was not tested: Live WebChat UI with a real gateway session producing the history-load/final-event race condition with a stale chatStream value.

clawsweeper · 2026-05-25T21:38:02Z

Thanks for the context here. I swept through the related work, and this is now duplicate or superseded.

Close as superseded: this older WebChat duplicate-final branch is not on main, and the newer open replacement PR is the clean, proof-positive canonical path for the same root cause.

Root-cause cluster
Relationship: superseded
Canonical: #96081
Summary: This PR and the newer PR target the same WebChat final/history duplicate-render root cause; the newer PR is the viable proof-positive candidate.

Members:

canonical: fix: WebChat duplicates external reply events #96081 - Open, clean, proof-positive PR with broader shared reconciliation changes and focused tests for the same duplicate-final problem.
fixed_by_candidate: [Bug]: WebChat UI renders duplicate assistant messages on chat.status final event #85771 - Canonical issue for the WebChat final/history duplicate-render report that should close from the merged canonical PR.
duplicate: [Bug]: Control UI duplicate assistant messages still reproducing on 2026.5.20 #85755 - Closed report describes the same v2026.5.20 WebChat duplicate assistant symptom and is referenced by this PR body.
partial_overlap: fix(ui): preserve visible chat stream text #89530 - Merged stream-retention work changed the same WebChat stream/history reconciliation area but did not itself close the duplicate-final report.

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

Canonical path: Keep review and landing focused on #96081, then close the linked WebChat duplicate-final issue from that merged implementation.

So I’m closing this here and keeping the remaining discussion on #96081.

Review details

Best possible solution:

Keep review and landing focused on #96081, then close the linked WebChat duplicate-final issue from that merged implementation.

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

Yes, source-level: current main appends active-run final assistant payloads without checking whether refreshed history already supplied the same assistant tail. I did not run a live browser reproduction in this read-only review.

Is this the best way to solve the issue?

No. This branch is a plausible local patch, but the newer PR is the stronger canonical fix because it centralizes assistant-tail reconciliation and covers sibling history/stream cases with proof and tests.

Security review:

Security review cleared: The diff is limited to WebChat controller state logic and tests, with no dependency, workflow, credential, package, or code-execution surface change.

AGENTS.md: found and applied where relevant.

What I checked:

Repository policy read: Root AGENTS.md and ui/AGENTS.md were read fully; root PR review policy required current-main, related-PR, proof, and history checks, while the UI guide had only i18n-specific guidance not directly affecting this controller change. (AGENTS.md:1, c6ade83a5ccb)
Current main still lacks the fix: Current main appends every renderable active-run final assistant payload through appendTerminalAssistantMessage without a same-assistant-tail/history replacement guard. (ui/src/ui/controllers/chat.ts:1364, c6ade83a5ccb)
Current helper scope: Current main's appendTerminalAssistantMessage only removes stream-fallback replacements after the latest user message, not persisted assistant history tails that match a terminal final. (ui/src/ui/chat/stream-reconciliation.ts:155, c6ade83a5ccb)
Older PR head remains a risky local patch: At the PR head, duplicate final/aborted payloads make the new !isTailDuplicate condition false and fall into materializeVisibleAssistantStreamMessages, so stale chatStream text can still be materialized instead of treating the duplicate terminal payload as handled. (ui/src/ui/controllers/chat.ts:1314, 6a452a818a1b)
Older PR tests do not cover stale different stream text: The PR-head tests cover same-text stream fallback, final, aborted, cross-run preservation, and different final content, but not the stale-stream-different-from-duplicate-final case that remains risky in the code path. (ui/src/ui/controllers/chat.test.ts:1529, 6a452a818a1b)
Replacement PR is viable: The newer PR is open, mergeable, clean, not draft, has sufficient proof labels, green status checks, and targets the same WebChat duplicate-final/history root cause. (4d6d053267bc)

Likely related people:

osolmaz: Authored the merged stream/history reconciliation change that created the shared helper and adjacent controller/test surfaces this PR modifies. (role: recent stream-reconciliation contributor; confidence: high; commits: 17a285f29825; files: ui/src/ui/controllers/chat.ts, ui/src/ui/chat/stream-reconciliation.ts, ui/src/ui/controllers/chat.test.ts)
scotthuang: Authored the newer canonical candidate PR and prior merged WebChat event/history ordering work. (role: adjacent WebChat event-ordering contributor; confidence: high; commits: 7734a40a5650, cc1468fd7bef; files: ui/src/ui/app-gateway.ts, ui/src/ui/app-gateway.node.test.ts, ui/src/ui/controllers/chat.ts)
steipete: Authored the final-event history reload optimization that shaped when terminal chat events trigger transcript reloads. (role: feature-history contributor; confidence: medium; commits: dc6afeb4f882; files: ui/src/ui/app-gateway.ts, ui/src/ui/chat-event-reload.ts, ui/src/ui/chat-event-reload.test.ts)
BunsDev: Authored recent merged reconnect/send preservation work touching the chat controller, gateway caller, and controller tests in the same message-state area. (role: recent adjacent WebChat contributor; confidence: medium; commits: 96635c7c27e3; files: ui/src/ui/controllers/chat.ts, ui/src/ui/controllers/chat.test.ts, ui/src/ui/app-gateway.ts)

Codex review notes: model internal, reasoning high; reviewed against c6ade83a5ccb.

clawsweeper · 2026-05-25T21:45:34Z

ClawSweeper PR egg

🎁 Pass real behavior proof to wake the egg and unlock a hatchable treat.

Where did the egg go?

The egg game starts only after the PR passes the real-behavior proof check.
Before that, no creature or rarity is rolled. The treat waits for real proof.
This is still just collectible flavor: proof affects review readiness, not creature quality.

Guard every append point in handleChatEvent with a tail-duplicate check using messageDisplaySignature. When a final, aborted, or stream-fallback assistant message matches the last entry already in chatMessages, the append is skipped. This prevents the race where loadChatHistory (triggered after tool events or session.message) returns history that already includes the assistant message, and the optimistic append from the final event creates a second copy. Closes openclaw#85771 Signed-off-by: Sebastien Tardif <[email protected]>

Signed-off-by: Sebastien Tardif <[email protected]>

Remove the isTailDuplicate guard from the stale-run final path so sub-agent or separate-run messages with identical text are always stored. The renderer already collapses consecutive identical messages via duplicateCount instead of silently dropping them. Same-turn dedup (current-run final/aborted paths) is preserved to catch the loadChatHistory race condition.

Signed-off-by: Sebastien Tardif <[email protected]>

Signed-off-by: Seb Tardif <[email protected]>

The previous isTailDuplicate used messageDisplaySignature which only compared text content, causing false positives when two messages had the same text but different content structure (e.g., text-only vs text+canvas). Switch to JSON.stringify comparison of content arrays so messages with different block types or counts are not incorrectly deduplicated. This fixes the upstream test 'keeps repeated assistant final text within the same turn' which expects both a text-only and a text+canvas message to be preserved even when they share the same display text. Signed-off-by: Sebastien Tardif <[email protected]>

clawsweeper · 2026-06-28T04:57:01Z

ClawSweeper applied the proposed close for this PR.

Action: closed this PR.
Close reason: duplicate or superseded.
Evidence: durable ClawSweeper review.
Coverage proof: PR B clearly carries PR A's core useful intent and is the better current review path for the same WebChat duplicate assistant final/history reconciliation problem; PR A's leftover details are either incidental or reviewable on PR B. Covering PR: fix: WebChat duplicates external reply events #96081.

SebTardif · 2026-06-28T14:20:29Z

The core fix from this PR landed in main via #88786 (merged 2026-06-03), which addresses the same root cause: terminal final/aborted events appending an assistant message while the same content still existed in the live chatStream/chatRunId state. Both PRs modify the same files (ui/src/ui/controllers/chat.ts, ui/src/ui/controllers/chat.test.ts).

Note: ClawSweeper cited #96081 as the covering PR, but that one is still open and covers a different variant (external reply event deduplication). The fix that actually shipped is #88786.

openclaw-barnacle Bot added app: web-ui App: web-ui size: S triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 25, 2026

openclaw-barnacle Bot added triage: mock-only-proof Candidate: PR proof only shows tests, mocks, snapshots, lint, typecheck, or CI. and removed triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 25, 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 25, 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 25, 2026

clawsweeper Bot added rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. and removed rating: 🦪 silver shellfish Thin PR readiness signal; proof, validation, or implementation needs work. labels May 25, 2026

SebTardif force-pushed the fix/webchat-dedup-status branch from 2a38a59 to 91d6cc9 Compare May 26, 2026 23:21

openclaw-barnacle Bot removed the size: S label May 26, 2026

clawsweeper Bot mentioned this pull request Jun 3, 2026

fix(ui): preserve visible chat stream text #89530

Merged

SebTardif force-pushed the fix/webchat-dedup-status branch from 2087bee to 133e5ef Compare June 3, 2026 19:30

github-actions Bot mentioned this pull request Jun 4, 2026

🦞 OpenClaw 生态日报 2026-06-04 litang9/big_model_radar#12

Open

SebTardif and others added 10 commits June 3, 2026 21:51

retrigger proof check

81a58fc

Signed-off-by: Sebastien Tardif <[email protected]>

retrigger proof check

a01910b

retrigger CI

87692d0

retrigger proof check

b0a4212

retrigger proof check

3667961

chore: retrigger proof evaluation

15e8d16

Signed-off-by: Sebastien Tardif <[email protected]>

retrigger proof check

a037c11

Signed-off-by: Seb Tardif <[email protected]>

SebTardif force-pushed the fix/webchat-dedup-status branch from c9c7710 to 6a452a8 Compare June 4, 2026 04:52

clawsweeper Bot mentioned this pull request Jun 15, 2026

[Bug]: WebChat UI renders duplicate assistant messages on chat.status final event #85771

Open

clawsweeper Bot mentioned this pull request Jun 23, 2026

fix: WebChat duplicates external reply events #96081

Closed

brokemac79 mentioned this pull request Jun 25, 2026

[codex] Route PR close reviews to autoclose openclaw/clawsweeper#371

Merged

This was referenced Jul 7, 2026

fix: WebChat duplicates external reply events #101456

Closed

fix: WebChat duplicates external reply events #101494

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(ui): deduplicate assistant messages on chat.status final event#86646

fix(ui): deduplicate assistant messages on chat.status final event#86646
SebTardif wants to merge 10 commits into
openclaw:mainfrom
SebTardif:fix/webchat-dedup-status

SebTardif commented May 25, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented May 25, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented May 25, 2026

Uh oh!

clawsweeper Bot commented Jun 28, 2026

Uh oh!

SebTardif commented Jun 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Uh oh!

Conversation

SebTardif commented May 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changed files

Verification

Real behavior proof

Uh oh!

clawsweeper Bot commented May 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented May 25, 2026

Uh oh!

clawsweeper Bot commented Jun 28, 2026

Uh oh!

SebTardif commented Jun 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

SebTardif commented May 25, 2026 •

edited

Loading

clawsweeper Bot commented May 25, 2026 •

edited

Loading