fix(slack): drop working pill on codex out-of-credits warning#82
Merged
Conversation
When codex returns an out-of-credits sentinel (task_complete with last_agent_message=null + has_credits=false), the bridge correctly surfaces the warning at the channel root — but then immediately sets the WORKING pill on it. The pill refresh loop re-applies every 30s, so the channel keeps showing "is working…" against a state that's already finished. Operators see a stuck "Processing…" indicator that contradicts the warning right above it. Add a `terminal?: boolean` flag on SlackPost. Set it on the out-of-credits post in format.ts. In bridge.ts, when a `kind: "text"` post arrives with `terminal: true`: still clear the previous anchor's pill, still post the text as the new root, but skip the setStatus(WORKING_LABEL) call — no work is happening, no pill should advertise that it is. Could be extended to normal codex task_complete later (which also leaves a stale pill on the last agent_message anchor), but keeping the scope to the regression I can see in the screenshot for now. Claude-runtime parity is separate — Claude has its own end-of-turn signaling.
Contributor
Author
|
Extended in 048e59d — same The earlier commit only marked codex's out-of-credits sentinel; a screenshot caught the same UX bug on the Claude side after a normal end-of-turn: pill stays lit on the final answer, refresh loop keeps re-applying it, and on the next user message Slack overlays its default localized "Finding answers..." indicator on top. Changes:
One acknowledged gap: codex normal completion ( |
rogeriochaves
added a commit
that referenced
this pull request
Jun 1, 2026
…turn) After PR #82 only the codex out-of-credits sentinel set `terminal: true`, so the bridge always re-attached the "is working…" pill after every non-terminal text post. For Claude that meant the pill stayed up forever after the agent posted its final message and ended the turn — exactly what you see in the screenshot where the agent's review is posted but the channel still shows a pill below. When a transcript-line message carries `stop_reason: "end_turn"` (or `stop_sequence` / `refusal`), find the last text post we just emitted from that message and mark it terminal. The bridge already honours that flag by skipping the pill re-attach. `tool_use` / `max_tokens` / `pause_turn` stay non-terminal — the agent is still working. If a turn ends with no text post (e.g. the last block was a tool call), we leave the pill alone — that's the existing behaviour; rare in practice, and dropping a pill mid-tool would be more confusing than the two-minute Slack TTL it relies on otherwise.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
When codex hits its credit ceiling, the rollout emits `task_complete` with `last_agent_message=null` + `has_credits=false`. The bridge already surfaces a clear warning at the channel root:
But then it immediately sets the "is working…" pill on that anchor. The refresh loop re-applies every 30s, so the channel keeps showing a working indicator against a state that's already finished. Operators get a mixed signal — warning above, "Processing…" below.
This PR adds a `terminal?: boolean` flag on `SlackPost` and sets it on the out-of-credits emission. In `bridge.ts`, when a `kind: "text"` post arrives with `terminal: true`, the previous anchor's pill still gets cleared and the warning still posts as the new root — but the `setStatus(WORKING_LABEL)` call is skipped, so nothing claims work is in progress.
Why a flag, not a text match
Coupling the bridge to specific warning text would be fragile (the wording lives in `format.ts`, the pill logic lives in `bridge.ts`, and we'd silently drift). A flag on the post shape makes the intent explicit at the emission site and gives us a clean extension point for any future terminal sentinels (normal task_complete with output, Claude end-of-turn, etc.).
Test plan