cottongin 4fc7f08b24 Add connection resilience and reconnect commands for Kosmi and Jackbox ...

Kosmi WebSocket would silently die after hours/days with no reconnection.
Jackbox WebSocket failed to reconnect after API server restarts (stale JWT)
and leaked heartbeat goroutines on each reconnect cycle.

Kosmi changes:
- Add WebSocket ping/pong keepalive (30s ping, 90s read deadline)
- Send EventFailure on unexpected disconnect to trigger gateway reconnectBridge()
- Add intentionalDisconnect flag to prevent false failure events on clean shutdown
- Fix Disconnect() to be safe for reconnect cycles

Jackbox changes:
- Add read deadline (90s) to detect stale connections
- Fix heartbeat goroutine leak via per-connection listenDone channel
- Re-authenticate for fresh JWT before each reconnect attempt
- Add Manager.Reconnect() for on-demand teardown and rebuild

IRC commands:
- !kreconnect - reconnect Kosmi bridge
- !jreconnect - reconnect Jackbox WebSocket
- !reconnect  - reconnect all services (Kosmi + Jackbox)

Made-with: Cursor

2026-04-05 05:30:39 -04:00

17 KiB

Raw Blame History

View Raw