fix(native-chat): scope composer sends to the active turn

[AmethystLiang]

Summary

Related issue: #8401. Canonical first-flush PR: #8418. This is an independent composer-lifecycle follow-up; it does not change transcript reading, polling, transport, or pagination.

Native chat could leave delayed PTY writes and optimistic messages owned by a prior composer target. Repeated prompts could also reconcile against an older identical turn. This PR scopes a send to the active turn by:

• returning cancellable handles for delayed prompt/Enter/image writes and cancelling them on target change, unmount, interrupt, Stop, or failed ownership;
• giving pending sends globally unique IDs and recording the authoritative transcript boundary visible at send time;
• reconciling repeated identical prompts by occurrence after that boundary, including bounded-history and remote-clock fallback;
• cancelling interactive-answer sends when their PTY target changes;
• reconciling attachment-only sends without leaving optimistic ghosts;
• keeping pending optimistic bubbles out of the streaming-preview derivation.

Screenshots

No visual change.

Testing

• pnpm lint (full lint/custom gates in review; changed-file lint rerun on the rebased head)
• pnpm typecheck (full typecheck in review; web typecheck rerun on the rebased head)
• Full repository pnpm test was not run; 48 focused files / 408 tests passed in review, then the exact rebased 7-file suite passed 67/67
• pnpm build was not run
• Added regressions for detach/Stop/target-change cancellation, repeated prompts, bounded history, launch prompts, attachment-only sends, interactive answers, remote clock domains, remount IDs, and optimistic streaming

AI Review Report

Two separate gpt-5.6-sol xhigh agents implemented and adversarially reviewed origin/main..HEAD. The second pass found and fixed five lifecycle gaps: Stop bypassing owned-send cancellation, interactive answers surviving PTY rebinding, cancelled sends lingering in the optimistic cache, attachment-only sends not reconciling, and renderer-clock fallback after an authoritative boundary was paged out.

The review covered macOS, Linux, and Windows input behavior; local and SSH/runtime PTY ownership; delayed timer cleanup; target changes and remounts; repeated/identical prompts; bounded history; image paths; performance; and overlap with #8418, #8157, and #8288. No platform-specific shortcut, path separator, shell behavior, or Electron branch was introduced.

Security Audit

No new command execution surface, IPC/RPC method, dependency, secret, auth, or credential handling. User text and image paths continue through the existing terminal-input path; this change reduces the period in which queued input can outlive its owning pane. Pending state remains renderer-local and bounded by the existing cache/pruning rules.

Notes

• Independent of fix(native-chat): retry not-yet-flushed transcripts instead of settling into a permanent error #8418/perf(native-chat): coalesce concurrent transcript reads #8157/fix(codex): resume sessions from authoritative homes #8288 and intentionally excludes transcript transport, pagination identity, split routing, paired-host mode, and accessibility work.
• Residual risk: before any authoritative transcript boundary exists, extreme SSH clock skew can conservatively retain a duplicate optimistic bubble rather than hiding the wrong turn.
• X handle: @JinjingLiang

[coderabbitai]

📝 Walkthrough

Walkthrough

Native chat send functions now return cancellable handles with settlement timing. A lifecycle hook tracks and cancels pending sends across interruptions, PTY changes, unmounts, and settlement. Pending message matching now uses content keys, transcript boundaries, and occurrence counts for text and image submissions. Composer and view logic use this lifecycle, remove canceled optimistic entries, and include pending messages in streaming derivation. Tests cover cancellation, lifecycle behavior, occurrence reconciliation, pruning, visibility, and streaming previews.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 35.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title matches the main change: scoping native chat composer sends to the active turn.
Description check	✅ Passed	The description follows the template and covers summary, screenshots, testing, AI review, security, and notes, including cross-platform review.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

🧹 Nitpick comments (1)

src/renderer/src/components/native-chat/NativeChatView.tsx (1)

309-333: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Duplicate pendingSendsAsMessages computation.

pendingSendsAsMessages(pending, sessionAfterCommandBoundaries.messages) is called twice with identical arguments — once inside the streamingText memo (Line 310) and again inside sessionWithPending (Line 330). Consider computing it once (e.g. in its own useMemo) and reusing the result in both places to avoid recomputation and future drift between the two call sites.

♻️ Proposed refactor

+  const pendingMessages = useMemo(
+    () => pendingSendsAsMessages(pending, sessionAfterCommandBoundaries.messages),
+    [pending, sessionAfterCommandBoundaries.messages]
+  )
   const streamingText = useMemo(() => {
-    const pendingMessages = pendingSendsAsMessages(pending, sessionAfterCommandBoundaries.messages)
     return deriveNativeChatStreamingText({
       messages:
         pendingMessages.length > 0
           ? [...sessionAfterCommandBoundaries.messages, ...pendingMessages]
           : sessionAfterCommandBoundaries.messages,
       previewText: hookPreview,
       working: hookWorking
     })
-  }, [sessionAfterCommandBoundaries.messages, pending, hookPreview, hookWorking])
+  }, [sessionAfterCommandBoundaries.messages, pendingMessages, hookPreview, hookWorking])
   const sessionWithPending = useMemo<typeof session>(() => {
     if (pending.length === 0 && commandMarkers.length === 0 && !streamingText) {
       return sessionAfterCommandBoundaries
     }
     return {
       ...sessionAfterCommandBoundaries,
       messages: [
         ...sessionAfterCommandBoundaries.messages,
         ...commandMarkersAsMessages(commandMarkers),
         ...(streamingText ? [nativeChatStreamingMessage(streamingText)] : []),
-        ...pendingSendsAsMessages(pending, sessionAfterCommandBoundaries.messages)
+        ...pendingMessages
       ]
     }
-  }, [sessionAfterCommandBoundaries, pending, commandMarkers, streamingText])
+  }, [sessionAfterCommandBoundaries, pendingMessages, commandMarkers, streamingText])

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 4dd7623a-e123-427b-b36c-f66af9f1715b

📥 Commits

Reviewing files that changed from the base of the PR and between 7b3103f and 69ec4af.

📒 Files selected for processing (14)

• src/renderer/src/components/native-chat/NativeChatComposer.test.tsx
• src/renderer/src/components/native-chat/NativeChatComposer.tsx
• src/renderer/src/components/native-chat/NativeChatView.tsx
• src/renderer/src/components/native-chat/native-chat-pending-occurrence.test.ts
• src/renderer/src/components/native-chat/native-chat-pending-occurrence.ts
• src/renderer/src/components/native-chat/native-chat-pending.test.ts
• src/renderer/src/components/native-chat/native-chat-pending.ts
• src/renderer/src/components/native-chat/native-chat-runtime-send.test.ts
• src/renderer/src/components/native-chat/native-chat-runtime-send.ts
• src/renderer/src/components/native-chat/use-native-chat-interactive-send.test.tsx
• src/renderer/src/components/native-chat/use-native-chat-interactive-send.ts
• src/renderer/src/components/native-chat/use-native-chat-send-lifecycle.test.tsx
• src/renderer/src/components/native-chat/use-native-chat-send-lifecycle.ts
• src/shared/native-chat-streaming.test.ts

[AmethystLiang]

Addressed the outside-diff CodeRabbit nit in d87459b by memoizing pending-message derivation once and reusing it for streaming preview and session assembly; transcript, marker, streaming-preview, optimistic-send order and pending reconciliation inputs are unchanged.

Validation: 8 focused NativeChatView/composer/pending/send-lifecycle/streaming files, 71 tests passed; web typecheck, changed-file oxlint + React Doctor, oxfmt check, git diff check, and max-lines ratchet passed. React Doctor retains five pre-existing state/effect advisories outside this edit; the generic docstring warning is intentionally not actioned because AGENTS.md requires brief why-comments instead of mechanical narration, and there were no inline threads to resolve.

@coderabbitai review

[coderabbitai]

@AmethystLiang Thanks for the detailed update. Kicking off a fresh review to verify the memoized pending-message derivation is correctly reused for both streaming preview and session assembly, and that the transcript/marker/streaming-preview/optimistic-send ordering and pending reconciliation inputs remain unchanged as described.

---

✏️ Learnings added

Learnt from: AmethystLiang
URL: https://github.com/stablyai/orca/pull/8568

Timestamp: 2026-07-13T19:48:24.448Z
Learning: In this repository (orca), AGENTS.md requires brief "why"-comments in code rather than mechanical narration/docstrings. When CodeRabbit flags generic docstring-style warnings, prefer brief rationale comments over verbose narrative docstrings, consistent with AGENTS.md guidance.

Note: Learnings are effective only in the context of similar code segments. To apply general coding guidelines, please configure review instructions. You can manage existing learnings in the app.

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.