Add !votes command, fix vote tally timing, and improve Kosmi stability

- Add !votes command (IRC + Kosmi) showing per-session and all-time vote
  breakdowns for the current game via new Jackbox API endpoints
  (GET sessions/{id}/games, sessions/{id}/votes, games/{id})
- Fix vote tally broadcasting: remove debounce timer, announce tallies
  only at game transitions or session end instead of after every vote
- Add !kreconnect IRC command to manually trigger Kosmi reconnection
- Add WebSocket ping/pong keepalive and write mutex to Kosmi client
  for connection stability
- Add watchConnection() auto-reconnect on unexpected Kosmi disconnects
- Remove old 2025-10-31 chat summaries; add votes command design doc

Made-with: Cursor

bridge/kosmi/graphql_ws_client.go

@@ -14,10 +14,14 @@ import (
 )
 
 const (
-	kosmiWSURL = "wss://engine.kosmi.io/gql-ws"
+	kosmiWSURL   = "wss://engine.kosmi.io/gql-ws"
 	kosmiHTTPURL = "https://engine.kosmi.io/"
-	userAgent  = "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36"
-	appVersion = "4364"
+	userAgent    = "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36"
+	appVersion   = "4364"
+
+	pingInterval = 30 * time.Second
+	pongTimeout  = 90 * time.Second
+	writeWait    = 10 * time.Second
 )
 
 // GraphQL-WS Protocol message types
@@ -40,6 +44,7 @@ type GraphQLWSClient struct {
 	messageCallback func(*NewMessagePayload)
 	connected       bool
 	mu              sync.RWMutex
+	writeMu         sync.Mutex
 	done            chan struct{}
 }
 
@@ -208,9 +213,17 @@ func (c *GraphQLWSClient) Connect() error {
 	c.connected = true
 	c.mu.Unlock()
 
+	// Set up ping/pong keepalive
+	conn.SetReadDeadline(time.Now().Add(pongTimeout))
+	conn.SetPongHandler(func(string) error {
+		conn.SetReadDeadline(time.Now().Add(pongTimeout))
+		return nil
+	})
+
 	c.log.Info("Native WebSocket client connected and ready")
 
-	// Start message listener
+	// Start keepalive pinger and message listener
+	go c.startPing()
 	go c.listenForMessages()
 
 	return nil
@@ -359,7 +372,10 @@ func (c *GraphQLWSClient) SendMessage(text string) error {
 		},
 	}
 
-	if err := c.conn.WriteJSON(msg); err != nil {
+	c.writeMu.Lock()
+	err := c.conn.WriteJSON(msg)
+	c.writeMu.Unlock()
+	if err != nil {
 		return fmt.Errorf("failed to send message: %w", err)
 	}
 
@@ -396,3 +412,33 @@ func (c *GraphQLWSClient) IsConnected() bool {
 	return c.connected
 }
 
+// Done returns a channel that is closed when the client disconnects
+func (c *GraphQLWSClient) Done() <-chan struct{} {
+	return c.done
+}
+
+// startPing sends WebSocket ping frames at a regular interval to keep the
+// connection alive and detect stale connections early.
+func (c *GraphQLWSClient) startPing() {
+	ticker := time.NewTicker(pingInterval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-ticker.C:
+			c.writeMu.Lock()
+			err := c.conn.WriteControl(
+				websocket.PingMessage, nil, time.Now().Add(writeWait),
+			)
+			c.writeMu.Unlock()
+			if err != nil {
+				c.log.Warnf("Ping failed, connection likely dead: %v", err)
+				c.conn.Close()
+				return
+			}
+		case <-c.done:
+			return
+		}
+	}
+}
+