mirror of
https://github.com/multica-ai/multica.git
synced 2026-08-02 01:45:52 +02:00
6fc9ce2724
When a user's local daemon is offline or stops claiming tasks, the chat StatusPill used to sit on "排队中" / "Queued" indefinitely with no clue that anything was wrong (GH #2341). The runtime sweeper takes ~150s to flip runtime.status, and even when it does the only diagnostic was a single-word "Offline" label. - Add a STUCK_THRESHOLD_SECS=30 escalation: queued/dispatched past 30s now renders a static "Daemon not responding" / "Daemon 无响应" stage even when availability still reports online — covers the sweeper-lag window and the heartbeating-but-not-claiming edge case. - Rename the offline-stage label to "Runtime offline" / "Runtime 离线" so it points at the actual failure surface instead of the agent. - Append a `multica daemon logs -f` diagnostic hint to the offline banner so users have a concrete next step. - Extract pickStageKeys for testing and add unit coverage for the new stuck-detection branches and the "offline always wins" precedence. Co-authored-by: multica-agent <github@multica.ai>