fix(server): set session to error state when sendTurn fails

[sachssem]

What Changed

When sendTurnForThread fails inside ProviderCommandReactor, the catchCause handler now also calls setThreadSession to transition the session into a terminal state, in addition to the existing failure activity.

• If an active provider binding exists: sets status: "error" (preserves binding for recovery).
• If no binding exists (e.g. provider mismatch before session start, or a previously stopped session): sets status: "stopped" to avoid creating a synthetic session without a backing ProviderService directory entry.

Both paths set lastError so the client can detect the failure.

Single-file production change in ProviderCommandReactor.ts. Four regression tests added covering existing-session, null-session, stopped-session failure paths, and recovery after a provider-switch error.

Why

After the client dispatches thread.turn.start, the HTTP call succeeds so turnStartSucceeded = true. The client then waits for a server-side signal via hasServerAcknowledgedLocalDispatch, which checks for phase === "running", threadError, or a session snapshot change.

When sendTurn fails server-side, the existing handler only appended a failure activity — it never updated the session. If the session was already bound and unchanged, none of the acknowledgment conditions were met: phase stayed "ready", threadError was null, and the session snapshot was identical. This left isSendBusy permanently stuck, the send button disabled, and "Working for …" hanging until Cmd+R.

Previous attempts to fix this (#1348, #1393, #1542) added client-side workarounds. This change addresses the root cause on the server by ensuring the session always transitions to a detectable state on failure.

Validation

• bun fmt — clean
• bun lint — clean
• 26/26 ProviderCommandReactor tests pass (4 new)
• Full orchestration + provider suite: 300/300 pass

Checklist

• This PR is small and focused
• I explained what changed and why

---

Note

Medium Risk
Changes error-handling in ProviderCommandReactor to mutate persisted session state on turn-start failures, which can affect client behavior and session reuse/recovery paths. Covered by new regression tests across existing-session and no-session scenarios, reducing risk but still touching orchestration state transitions.

Overview
Ensures thread.turn-start-requested failures no longer only append a provider.turn.start.failed activity: the reactor now also calls setThreadSession to move the thread session into a detectable terminal state and populate lastError.

On failure, sessions with an existing non-stopped binding are marked status: "error" (preserving binding for recovery), while cases with no real binding (including previously stopped sessions) are marked status: "stopped" to avoid creating a synthetic bound session. Tests are updated/added to assert these statuses, lastError, and that a subsequent correct-provider turn can recover after a provider-switch failure.

^{Reviewed by Cursor Bugbot for commit aee8e69. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Set thread session to error state when sendTurn fails in ProviderCommandReactor

• When a thread.turn.start fails, the reactor now updates the thread session: if an active session existed it transitions to status: 'error', otherwise it creates/updates a session with status: 'stopped'.
• In both cases, activeTurnId is cleared and lastError is set to 'Provider turn start failed: …', alongside the existing provider.turn.start.failed activity.
• New and updated tests in ProviderCommandReactor.test.ts cover the error, stopped, and recovery scenarios.
• Behavioral Change: previously a failed turn start left the session in an ambiguous state; it now always reflects the failure explicitly.

^{Macroscope summarized aee8e69.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 069ae7c7-d689-477a-87a3-89bd89940caa

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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 and usage tips.}

[macroscopeapp]

Approvability

Verdict: Approved

This is a focused bug fix that properly transitions session state to 'error' when sendTurn fails, fixing a hanging UI indicator. The code change is self-contained within an existing error handler, and 85% of the diff is comprehensive test coverage for the new behavior.

^{You can customize Macroscope's approvability policy. Learn more.}