fix(web): first reply in a new thread shows without a reload by realies · Pull Request #27570 · pingdotgg/t3code

realies · 2026-10-30T12:10:54Z

Problem

On Windows desktop (0.1.6-nightly.20261030.3917), the first reply in a new thread never appears in the chat view. The sidebar updates and the turn completes, but the conversation shows only the user's message. Switching away and back shows "Send a message to start the conversation". Only a reload (Ctrl+R) or a restart shows the reply. Later messages in the same thread work. The reporter does not see this on macOS or Linux.

Change

useThreadLifecycle read runStopAtom(activeThreadRef) while the route was still a draft. That mounted the thread state for an id the server had not created yet. The bounded snapshot request returned thread_not_found, so makeEnvironmentThreadState marked the thread deleted and parked its subscription. The resume cache kept that state after the thread was created.

The run-stop lookup now waits for the thread shell, as ChatView already does for its own detail reads through resolveThreadDetailRef. For a server thread, activeThreadShell is never null, so the Stop control behaves the same once the thread exists.

Scope and approval

This is a small, focused fix for an obvious bug, so I didn't open an issue first. It changes one guard in one hook, with no product-behavior change. A draft has no runs to stop, so the Stop control has nothing to show before the thread exists.

Verification

Tested on Windows 11 with vp run dev (isolated home) and the Replay (dev) provider, Showcase scenario:

Click New thread, type "go", press Enter.
Without the fix: the server log shows threadBoundedSnapshot failing with thread_not_found about 1.6s before orchestrationV2.dispatch.threadCreate. The turn completes (the sidebar shows the generated title), but the conversation shows only "go".
With the fix: the reply streams in live and stays.

Before	After

(The Changes count in the after shot comes from an unrelated broken git index in my local checkout. The machine name is redacted.)

vp run check:changed passes: format, lint, knip, typecheck and React compiler. No test was added. The fix is wiring in a hook with no existing test, and a ChatView browser test of the draft-to-thread handoff would need much more setup than the change. Happy to add one if you want it.

Not checked: macOS and Linux. The cause looks platform-independent, so I don't know why it only shows up on Windows. It may depend on request timing.

Fixed with Claude Opus 5.5 in Claude Code.

macroscopeapp · 2026-10-30T12:12:59Z

Approvability

Verdict: Approved at cf48159

Macroscope's review found this PR approvable — The run-stop subscription now waits until a new thread exists, preventing premature loading from hiding its first reply. Existing thread controls remain unchanged, and the reported Windows reproduction confirms replies appear without reloading.

^{You can add or adjust custom eligibility rules. Learn more.}

The run-stop lookup loaded the draft's thread before it existed. The miss cached the thread as deleted, so the chat view never showed the reply. Wait for the thread shell before loading it.

coderabbitai · 2026-10-30T12:17:03Z

Important

Review skipped

Review was skipped as selected files did not have any reviewable changes.

Run configuration

Source mode: Auto
Repository: pingdotgg/t3code
Branch: main
File: .coderabbit.config.ts
Commit: 32ecd71
Layers: Repository TS config
Review profile: CHILL
Plan: Advanced
Run ID: 27e92940-042e-4d8f-9111-5acd24d4ab08

Commits

Reviewing files that changed from the base of the PR and between cf48159 and 52b678f.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

No actionable comments were generated in the recent review.

Recent review info

Run configuration

Source mode: Auto
Repository: pingdotgg/t3code
Branch: main
File: .coderabbit.config.ts
Commit: 32ecd71
Layers: Repository TS config
Review profile: CHILL
Plan: Advanced
Run ID: 823b66a1-844b-4898-8418-2861f0fef87d

Commits

Reviewing files that changed from the base of the PR and between 32ecd71 and cf48159.

Files selected for processing: 1

apps/web/src/components/chat/useThreadLifecycle.tsx

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 9 remain after this review.

Walkthrough

The hook now reads the run-stop atom only when both the active thread reference and thread shell are available. Otherwise, it uses NO_RUN_STOP_ATOM.

Changes: Thread lifecycle

Layer / File(s)	Summary
Gate run-stop atom reads `apps/web/src/components/chat/useThreadLifecycle.tsx`	The hook reads the run-stop atom when both the active thread reference and shell are available. If either is unavailable, it uses `NO_RUN_STOP_ATOM`.

Priority: Pending latest changes

Estimated code review effort: 1 (Trivial) | ~5 minutes

Change: Bug fix

Possibly related issues

pingdotgg/t3code#27482 — Reports an empty thread view after a snapshot request precedes thread creation. The PR changes a run-stop atom read that may cause an early lookup, but does not change cache or subscription handling.
pingdotgg/t3code#27443 — Reports a thread history request before server creation. The PR changes an early run-stop atom read, but does not change the reported deleted-state or subscription handling.
pingdotgg/t3code#27527 — Reports an empty new-thread view after a draft-time snapshot request. The PR gates the run-stop atom read on shell availability; the issue also describes a background-service setup.

Possibly related PRs

pingdotgg/t3code#27522 — Changes the same useThreadLifecycle run-stop read to avoid requesting a reserved thread ID before the server lists it.
pingdotgg/t3code#25477 — Uses creation-gated thread detail lookup on mobile to avoid caching a missing thread before creation.

Suggested reviewers: juliusmarminge, macroscopeapp

Merge Risk: Minimal · previous review cf481

The change safely gates run-stop reads until the thread shell is available, so the first reply can appear without introducing a verified regression.

Pre-merge checks: 4

Check name	Status	Explanation
Title check		The title clearly and concisely describes the primary fix: making the first reply in a new thread visible without requiring a reload.
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.
Description check		The description includes all required sections: Problem, Change, Scope and approval, and Verification. It explains the reproduction, root cause, fix, approval rationale, verification steps, screenshot…

Full details: Description check

Explanation

The description includes all required sections: Problem, Change, Scope and approval, and Verification. It explains the reproduction, root cause, fix, approval rationale, verification steps, screenshots, checks run, limitations, and agent usage.

Finishing Touches

🧪 Generate unit tests (beta)

Create a new PR

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

realies · 2026-10-30T13:05:37Z

Closing as a duplicate of #27522, which makes the same fix and adds a test.

For the record, this isn't Windows-only. It came in with #27252: nightly 20261029.3892 is clean and 20261030.3917 has it, on both Windows and macOS desktop. My earlier "not on macOS" was from a Mac still running 3892. On 3917 the macOS server log shows the same sequence: the draft's /threads/<id>/bounded request returns 404 before threadCreate, and nothing loads the thread until a reload.

github-actions Bot added vouch:unvouched PR author is not yet trusted in the VOUCHED list. size:XS 0-9 changed lines (additions + deletions). labels Oct 30, 2026

macroscopeapp Bot approved these changes Oct 30, 2026

View reviewed changes

realies force-pushed the fix/draft-first-reply-hidden branch from cf48159 to ab03ad4 Compare October 30, 2026 12:14

fix(web): first reply in a new thread shows without a reload

52b678f

The run-stop lookup loaded the draft's thread before it existed. The miss cached the thread as deleted, so the chat view never showed the reply. Wait for the thread shell before loading it.

realies force-pushed the fix/draft-first-reply-hidden branch from ab03ad4 to 52b678f Compare October 30, 2026 12:14

realies closed this Oct 30, 2026

This was referenced Oct 30, 2026

fix(web): new threads show the agent's output without a reload #27609

Closed

fix(web): new threads stream the agent's reply without a reload #27611

Closed

Navigation Menu

Repository navigation

fix(web): first reply in a new thread shows without a reload - #27570

fix(web): first reply in a new thread shows without a reload#27570
realies wants to merge 1 commit into
pingdotgg:mainfrom
realies:fix/draft-first-reply-hidden

realies commented Oct 30, 2026 •

edited

Loading

Uh oh!

macroscopeapp Bot commented Oct 30, 2026

Uh oh!

coderabbitai Bot commented Oct 30, 2026 •

edited

Loading

Review skipped

Uh oh!

realies commented Oct 30, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

realies commented Oct 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Change

Scope and approval

Verification

Uh oh!

macroscopeapp Bot commented Oct 30, 2026

Approvability

Uh oh!

coderabbitai Bot commented Oct 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

realies commented Oct 30, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

realies commented Oct 30, 2026 •

edited

Loading

coderabbitai Bot commented Oct 30, 2026 •

edited

Loading