fix(webchat): preserve refresh-visible history and composer state by spacegeologist · Pull Request #83992 · openclaw/openclaw

spacegeologist · 2026-05-19T04:49:42Z

Summary

Refs #83344.

This intentionally does not auto-close #83344 because session preview/title metadata remains out of scope for this PR.

Addresses two refresh-time WebChat state gaps from #83344:

expands the raw transcript window used by chat.history before visible-message projection/pagination, so recent visible conversation can be backfilled even when the latest raw transcript tail is mostly silent/internal traffic
persists WebChat composer draft text and queued messages per gateway/session in sessionStorage, then restores them after refresh/session switching

Notes

Uses sessionStorage instead of localStorage as the safer privacy default: drafts/queued prompts survive refresh in the same tab, but do not persist indefinitely across browser restarts.
Queued attachments are restored only when their payload data is available. If a queued attachment cannot be serialized safely, the queue item is not persisted rather than being restored without the attachment.
Session preview/title metadata is intentionally out of scope for this PR.

Real behavior proof

Behavior or issue addressed: WebChat refresh keeps visible chat history available after a tool-heavy/raw-tail transcript, and preserves same-tab composer draft plus queued messages for the same gateway/session.
Real environment tested: Local OpenClaw Control UI from this PR served by Vite at http://127.0.0.1:5173, connected in Google Chrome to a local loopback WebSocket gateway at ws://127.0.0.1:18789.
Exact steps or command run after this patch:
1. Started the patched Control UI with npm --prefix ui run dev -- --host 127.0.0.1 --port 5173 --force.
2. Started a local loopback WebSocket gateway that returns real gateway frames for connect, sessions.list, chat.history, and chat.send, and keeps the first chat.send run active.
3. Opened Chrome to http://127.0.0.1:5173/chat?session=main.
4. Verified history showed visible history before tool-heavy tail and visible assistant reply restored after refresh.
5. Sent start long run proof, then queued queued proof after refresh while that run stayed active.
6. Typed an unsent draft draft proof after refresh.
7. Reloaded the browser tab and waited for the Control UI to reconnect to the same gateway/session.

Evidence after fix: Copied live output from the browser accessibility tree after reload included:

URL: 127.0.0.1:5173/chat?session=main
VERSION v proof-local-gateway
Gateway status: Online
visible history before tool-heavy tail
visible assistant reply restored after refresh
start long run proof
Queued ( 1 )
queued proof after refresh
text entry area Value: draft proof after refresh
Run status: In progress
Queue message

Observed result after fix: The refreshed WebChat kept the same visible history messages, restored the queued message under Queued (1), and restored the unsent draft in the composer for session main. The queued message stayed in the queue rather than being delivered early on reconnect.
What was not tested: The gateway Vitest target did not run locally because root node_modules does not contain vitest/package.json after the full pnpm workspace install stalled; session preview/title metadata remains intentionally out of scope for this PR.

Tests

Added gateway coverage for chat.history backfilling visible messages when the raw tail is mostly NO_REPLY.
Added UI coverage for composer draft/queue restore, gateway/session scoping, clearing, attachment payload handling, and lifecycle persistence.
Ran npm --prefix ui test -- src/ui/chat/composer-persistence.test.ts src/ui/app-lifecycle.node.test.ts:
- 2 test files passed
- 8 tests passed
Ran git diff --check.
Ran a lightweight Node smoke check for ui/src/ui/chat/composer-persistence.ts.
Attempted OPENCLAW_GATEWAY_PROJECT_SHARDS=1 node scripts/run-vitest.mjs run --config test/vitest/vitest.gateway.config.ts src/gateway/server.chat.gateway-server-chat-b.test.ts; it failed before running tests because root node_modules is missing vitest/package.json in this local environment.

Full targeted Vitest was not run locally because dependency installation for the full 128-workspace pnpm repo stalled in this environment.

clawsweeper · 2026-05-19T04:50:46Z

Codex review: needs maintainer review before merge. Reviewed May 31, 2026, 4:20 AM ET / 08:20 UTC.

Summary
The PR widens WebChat refresh history backfill and adds per-gateway/session persistence for composer drafts and queued messages in browser sessionStorage.

PR surface: Source +459, Tests +380. Total +839 across 11 files.

Reproducibility: yes. at source level: current main stores chatMessage and chatQueue in component memory and reads a narrow raw transcript tail before visible projection, matching the reported refresh loss mode. I did not run a live browser before/after repro in this read-only review.

Review metrics: 2 noteworthy metrics.

Browser Storage Surface: 1 sessionStorage key family added. The new store retains user-authored WebChat draft and queue data in the browser, so maintainers should review the privacy boundary before merge.
Gateway History Read Window: 1 chat.history path widened from limit+1 to limit*20+21 raw lines/messages. The broader bounded read is the core history fix and changes per-refresh Gateway transcript I/O on a hot path.

Merge readiness
Overall: 🦐 gold shrimp
Proof: 🦞 diamond lobster
Patch quality: 🦐 gold shrimp
Result: needs maintainer review before merge.

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

Rank-up moves:

Get explicit maintainer acceptance or a narrowing patch for sessionStorage retention of drafts, queued prompts, local command args, and attachment data URLs.
[P2] Let the latest head's required checks finish green before merge.

Mantis proof suggestion
A short browser recording would give maintainers direct visual proof of reload-restored history, draft text, and queue behavior. 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 reload restores recent visible history, unsent draft text, and a queued prompt without sending it early.

Risk before merge

[P1] Browser sessionStorage will retain user-authored draft text, queued prompt text, local slash-command arguments, and serializable attachment data URLs until the browser clears the tab session; maintainers need to accept that privacy boundary.
[P1] Restored queues can now drain after refresh once a fresh sessions.list result says the selected session is idle, so queued prompts or local slash commands may be delivered later than the user last saw them.
[P1] The PR intentionally does not finish the session preview/title metadata part of [Bug]: WebChat refresh can lose visible conversation history when transcript tail is mostly internal tool traffic #83344, so it should not be treated as the complete fix for that issue cluster.
[P1] Latest head checks were still queued or in progress when inspected through the GitHub check-runs API, so required CI should settle before merge.

Maintainer options:

Approve Current Retention Boundary
Merge after required checks are green if maintainers explicitly accept same-tab sessionStorage retention for drafts, queued text, local command args, and serializable attachment payloads.
Narrow Persisted Payloads First
Before merge, remove attachment data URL persistence or add an explicit clearing/UX policy if maintainers do not want queued files retained in browser storage.
Pause Queue Replay Semantics
Pause the PR if restored queued prompts or local slash commands should require an explicit user action after refresh instead of auto-draining on fresh idle proof.

Next step before merge

[P2] The next action is maintainer judgment on privacy and replay semantics, not an autonomous repair against a clear code defect.

Security
Needs attention: The diff changes the browser-side sensitive-data retention boundary by storing drafts, queued prompts, local command args, and serializable attachment payloads in sessionStorage.

Review details

Best possible solution:

Land the focused history/composer fix only after maintainers explicitly accept or narrow the browser-storage and restored-queue replay boundaries, then keep the linked issues open or updated for remaining preview/title metadata work.

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

Yes at source level: current main stores chatMessage and chatQueue in component memory and reads a narrow raw transcript tail before visible projection, matching the reported refresh loss mode. I did not run a live browser before/after repro in this read-only review.

Is this the best way to solve the issue?

Unclear as a product decision: the code path is a plausible focused fix, but the durable answer depends on whether maintainers accept sessionStorage retention and auto-drain semantics for restored queues. If not, the safer direction is to narrow persisted payloads or require explicit replay after refresh.

AGENTS.md: found and applied where relevant.

Codex review notes: model gpt-5.5, reasoning high; reviewed against b81adc6202b2.

Label changes

Label changes:

add proof: sufficient: Contributor real behavior proof is sufficient. The PR body contains after-fix live browser output from a local Control UI connected to a loopback gateway showing restored visible history, a queued message, and an unsent draft after reload.

Label justifications:

P2: The PR addresses a normal-priority WebChat state-loss bug with limited blast radius but real user impact.
merge-risk: 🚨 session-state: The patch persists and restores WebChat draft and queue state across refresh and session switching.
merge-risk: 🚨 message-delivery: The patch can replay restored queued prompts and local slash commands after refresh once fresh session state reports idle.
merge-risk: 🚨 security-boundary: The patch stores prompt text, command args, and serializable attachment data URLs in browser sessionStorage.
rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🦞 diamond lobster and patch quality is 🦐 gold shrimp.
status: ⏳ waiting on author: ClawSweeper has contributor-facing work open and is waiting for author action. Sufficient (live_output): The PR body contains after-fix live browser output from a local Control UI connected to a loopback gateway showing restored visible history, a queued message, and an unsent draft after reload.
proof: sufficient: Contributor real behavior proof is sufficient. The PR body contains after-fix live browser output from a local Control UI connected to a loopback gateway showing restored visible history, a queued message, and an unsent draft after reload.

Evidence reviewed

PR surface:

Source +459, Tests +380. Total +839 across 11 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	7	468	9	+459
Tests	4	381	1	+380
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	11	849	10	+839

Security concerns:

[medium] Accept browser storage retention — ui/src/ui/chat/composer-persistence.ts:326
persistChatComposerState writes drafts and serialized queue items to sessionStorage, and serializeChatAttachment includes data URLs when available. That may be acceptable for same-tab refresh recovery, but it should be explicitly accepted or narrowed because prompts and attachments can contain sensitive data.
Confidence: 0.88

Acceptance criteria:

[P1] npm --prefix ui test -- src/ui/chat/composer-persistence.test.ts src/ui/app-lifecycle.node.test.ts.
[P1] node scripts/run-vitest.mjs run --config test/vitest/vitest.gateway.config.ts src/gateway/server.chat.gateway-server-chat-b.test.ts.
[P1] git diff --check.

What I checked:

Repository policy read: Root AGENTS.md plus UI and Gateway scoped AGENTS.md files were read; their guidance makes persisted state, message delivery, security boundaries, and Gateway hot paths merge-sensitive review surfaces. (AGENTS.md:1, b81adc6202b2)
Current main composer state is memory-only: Current main initializes chatMessage and chatQueue as component state, so a full browser refresh rebuilds those values empty without a storage-backed restore path. (ui/src/ui/app.ts:245, b81adc6202b2)
Current main session switch only preserves in-memory queues: Current main saves and restores chatQueue by session during in-app switching, then clears chatMessage; that does not survive tab refresh. (ui/src/ui/app-render.helpers.ts:136, b81adc6202b2)
Current main history read is too narrow for tool-heavy tails: Current chat.history reads only max+1 raw local messages before visible-message projection, which matches the linked report's source-level failure mode when the raw tail is mostly silent/internal records. (src/gateway/server-methods/chat.ts:2441, b81adc6202b2)
PR widens bounded raw history reads: The PR changes chat.history to use resolveSessionHistoryTailReadOptions(max), reading a bounded raw window before visible projection while preserving maxBytes. (src/gateway/server-methods/chat.ts:2443, 5fc2e6c63972)
PR stores composer snapshots in sessionStorage: The new persistence path stores draft text and serialized queue items per gateway/session, including attachment data URLs when available. (ui/src/ui/chat/composer-persistence.ts:326, 5fc2e6c63972)

Likely related people:

steipete: Current-main blame for the relevant WebChat refresh/session-switch/Gateway history surfaces points to Peter Steinberger, the live PR is assigned to steipete, and the PR head commits in the provided context are authored by steipete. (role: recent area contributor and likely follow-up owner; confidence: high; commits: 5c38c0c76d27, 25a06c8e94e8, 5fc2e6c63972; files: ui/src/ui/app-chat.ts, ui/src/ui/app-render.helpers.ts, ui/src/ui/chat/composer-persistence.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.

spacegeologist · 2026-05-19T10:40:54Z

@clawsweeper re-review

clawsweeper · 2026-05-19T10:42:42Z

🦞🧹
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: 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/26092101877
Updated: 2026-05-19T10:48:15.368Z

spacegeologist · 2026-05-19T10:53:36Z

@clawsweeper re-review

clawsweeper · 2026-05-19T10:55:15Z

🦞🧹
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/26093079233
Updated: 2026-05-19T11:14:39.572Z

spacegeologist · 2026-05-19T11:01:20Z

@clawsweeper re-review

…enclaw#83992) WebChat now stores/restores composer draft and queued sends across refresh, scoped by gateway/session/agent. It skips in-flight/steered sends, restores after agent scope hydration, waits for fresh idle session proof before draining restored sends, and backfills visible chat history when the raw tail contains silent/context entries. Refs openclaw#83344 Co-authored-by: Zee Zheng <zheng.zuo0@gmail.com>

openclaw-barnacle Bot added app: web-ui App: web-ui gateway Gateway runtime size: L triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 19, 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 19, 2026

spacegeologist marked this pull request as ready for review May 19, 2026 10:39

spacegeologist force-pushed the fix/webchat-refresh-history-composer branch from 6bb9097 to 77025f3 Compare May 19, 2026 10:47

openclaw-barnacle Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 19, 2026

This was referenced May 21, 2026

feat(ui): preserve chat draft per session on switch #47992

Closed

fix(ui): preserve WebChat messages across session switches #81037

Closed