eb35bc5dc9
The runtime sweeper was publishing task:failed events without a WorkspaceID, causing them to be silently dropped by the WS listener. This meant frontends never received notification when stale/orphaned tasks were failed by the sweeper — the live log card kept showing "Agent is working" and the agent status remained "working" indefinitely. - Look up workspace_id from issue table for each swept task - Set WorkspaceID on published events so they reach the correct WS room - Reconcile agent status after sweeping so agents return to "idle" |
||
|---|---|---|
| .. | ||
| cmd | ||
| internal | ||
| migrations | ||
| pkg | ||
| go.mod | ||
| go.sum | ||
| sqlc.yaml | ||