Update websocket_handler.py

websocket_handler.py

@@ -368,91 +368,93 @@ class RealtimeSession:
             
         log_info(f"✅ Reset for new utterance complete", session_id=self.session.session_id)
 
-    async def handle_control_message(websocket: WebSocket, session: RealtimeSession, message: Dict[str, Any]):
-        """Handle control messages"""
-        action = message.get("action")
-        config = message.get("config", {})
+
+# ========================= Backend-Frontend Messaging =========================
+async def handle_control_message(websocket: WebSocket, session: RealtimeSession, message: Dict[str, Any]):
+    """Handle control messages"""
+    action = message.get("action")
+    config = message.get("config", {})
+    
+    log_debug(f"🎮 Control message", action=action, session_id=session.session.session_id)
+    
+    if action == "start_session":
+        # Session configuration
+        await websocket.send_json({
+            "type": "session_config",
+            "session_id": session.session.session_id,
+            "config": {
+                "silence_threshold_ms": session.silence_threshold_ms,
+                "audio_chunk_size": session.audio_chunk_size,
+                "supports_barge_in": False  # Barge-in devre dışı
+            }
+        })
+        
+    elif action == "end_session" or action == "stop_session":
+        # Clean up and close
+        await session.cleanup()
+        await websocket.close()
+        
+    elif action == "interrupt":
+        # Barge-in devre dışı - ignore
+        log_warning(f"⚠️ Interrupt request ignored (barge-in disabled)", session_id=session.session.session_id)
+        
+    elif action == "reset":
+        # Reset conversation state
+        await session.reset_for_new_utterance()
+        await session.stop_stt_streaming()
+        await session.change_state(ConversationState.IDLE)
+        await websocket.send_json({
+            "type": "state_change",
+            "from": session.state.value,
+            "to": "idle"
+        })
         
-        log_debug(f"🎮 Control message", action=action, session_id=session.session.session_id)
+    elif action == "audio_ended":
+        # Audio playback ended on client
+        log_info(f"🎵 Client reported audio ended, current state: {session.state.value}", session_id=session.session.session_id)
         
-        if action == "start_session":
-            # Session configuration
+        if session.state == ConversationState.PLAYING_AUDIO:
+            # State'i listening'e çevir
+            await session.change_state(ConversationState.LISTENING)
             await websocket.send_json({
-                "type": "session_config",
-                "session_id": session.session.session_id,
-                "config": {
-                    "silence_threshold_ms": session.silence_threshold_ms,
-                    "audio_chunk_size": session.audio_chunk_size,
-                    "supports_barge_in": False  # Barge-in devre dışı
-                }
+                "type": "state_change",
+                "from": "playing_audio",
+                "to": "listening"
             })
             
-        elif action == "end_session" or action == "stop_session":
-            # Clean up and close
-            await session.cleanup()
-            await websocket.close()
-            
-        elif action == "interrupt":
-            # Barge-in devre dışı - ignore
-            log_warning(f"⚠️ Interrupt request ignored (barge-in disabled)", session_id=session.session.session_id)
+            # STT'yi başlat
+            log_info(f"🎤 Starting STT after audio playback ended", session_id=session.session.session_id)
             
-        elif action == "reset":
-            # Reset conversation state
-            await session.reset_for_new_utterance()
-            await session.stop_stt_streaming()
-            await session.change_state(ConversationState.IDLE)
-            await websocket.send_json({
-                "type": "state_change",
-                "from": session.state.value,
-                "to": "idle"
-            })
+            # Önce mevcut STT varsa temizle
+            if session.stt_manager:
+                await session.stop_stt_streaming()
+                await asyncio.sleep(0.1)  # Kısa bekleme
             
-        elif action == "audio_ended":
-            # Audio playback ended on client
-            log_info(f"🎵 Client reported audio ended, current state: {session.state.value}", session_id=session.session.session_id)
+            # Yeni STT başlat
+            success = await session.initialize_stt()
             
-            if session.state == ConversationState.PLAYING_AUDIO:
-                # State'i listening'e çevir
-                await session.change_state(ConversationState.LISTENING)
+            # STT hazır olduğunda sinyal gönder
+            if success and session.is_streaming:
+                log_info(f"✅ Sending STT ready signal", session_id=session.session.session_id)
                 await websocket.send_json({
-                    "type": "state_change",
-                    "from": "playing_audio",
-                    "to": "listening"
+                    "type": "stt_ready",
+                    "message": "STT is ready to receive audio"
                 })
-                
-                # STT'yi başlat
-                log_info(f"🎤 Starting STT after audio playback ended", session_id=session.session.session_id)
-                
-                # Önce mevcut STT varsa temizle
-                if session.stt_manager:
-                    await session.stop_stt_streaming()
-                    await asyncio.sleep(0.1)  # Kısa bekleme
-                
-                # Yeni STT başlat
-                success = await session.initialize_stt()
-                
-                # STT hazır olduğunda sinyal gönder
-                if success and session.is_streaming:
-                    log_info(f"✅ Sending STT ready signal", session_id=session.session.session_id)
-                    await websocket.send_json({
-                        "type": "stt_ready",
-                        "message": "STT is ready to receive audio"
-                    })
-                else:
-                    log_error(f"❌ STT initialization failed", session_id=session.session.session_id)
-                    await websocket.send_json({
-                        "type": "error",
-                        "error_type": "stt_init_failed",
-                        "message": "Failed to initialize STT after audio playback"
-                    })
             else:
-                log_warning(f"⚠️ audio_ended received but state is not playing_audio: {session.state.value}", session_id=session.session.session_id)
-                
-        elif action == "restart_stt":
-            # Manual STT restart request
-            log_info(f"🔄 Manual STT restart requested", session_id=session.session.session_id)
-            await session.stop_stt_streaming()
-            await session.restart_stt_if_needed()
+                log_error(f"❌ STT initialization failed", session_id=session.session.session_id)
+                await websocket.send_json({
+                    "type": "error",
+                    "error_type": "stt_init_failed",
+                    "message": "Failed to initialize STT after audio playback"
+                })
+        else:
+            log_warning(f"⚠️ audio_ended received but state is not playing_audio: {session.state.value}", session_id=session.session.session_id)
+            
+    elif action == "restart_stt":
+        # Manual STT restart request
+        log_info(f"🔄 Manual STT restart requested", session_id=session.session.session_id)
+        await session.stop_stt_streaming()
+        await session.restart_stt_if_needed()
 
 # ========================= MAIN HANDLER =========================
 async def websocket_endpoint(websocket: WebSocket, session_id: str):