Skip to content

fix(web): stop thinking when the coordinator replies - #120

Merged
etodanik merged 2 commits into
fix/posix-child-prebuildfrom
codex/stop-thinking-on-reply
Nov 3, 2026
Merged

etodanik merged 2 commits into
fix/posix-child-prebuildfrom
codex/stop-thinking-on-reply

Conversation

@etodanik

@etodanik etodanik commented Nov 3, 2026

Copy link
Copy Markdown
Collaborator

The thinking indicator could remain active after a coordinator reply
if the following message.run.ended frame was missed.

Treat connector.reply as completion in the existing UI reducer.
Keep reply text visible until the durable transcript catches up,
and allow thinking to restart for the next generation.

Changes are limited to the UI reducer and one regression test.

Stacked on #115.
Closes #118.

End the live reply on connector.reply so a missed message-end
frame cannot leave the thinking indicator active.

Add one regression test covering completion, text preservation,
and thinking restarting for the next generation.
@etodanik
etodanik force-pushed the codex/stop-thinking-on-reply branch from 9f1d472 to 372a454 Compare November 3, 2026 16:26
@etodanik
etodanik changed the base branch from main to fix/posix-child-prebuild November 3, 2026 16:26

@TheGreatAxios TheGreatAxios left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

One bug in the new terminal fold. Treating connector.reply as completion is right when the end frame never arrives, but that end frame is the normal path, not only the miss. It stamps a later endedAt, so a send between the two frames hides the thinking mark until the next run starts. Comment is on the diff.

Nothing else in this change looks wrong. Keeping the first end time still lets the next generation think: inference.start and message.run.started clear ended before the following reply stamps a new time.

Keep the first terminal timestamp when duplicate completion frames arrive, so an old message end cannot clear a newer send's waiting indicator. Cover duplicate completion and the next generation with a controlled clock.
@corbits-triage

corbits-triage Bot commented Nov 3, 2026

Copy link
Copy Markdown

Triage: Fail

Waiting on:

  • checks are still running: e2e-pipeline, lint, typecheck, test, format:check

Stacked on #115: only this pull request's own changes were checked.

@etodanik
etodanik merged commit 52ca312 into fix/posix-child-prebuild Nov 3, 2026
9 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

fix(web): thinking indicator remains active after a reply

2 participants