Update websocket_handler.py

websocket_handler.py

@@ -376,19 +376,45 @@ async def handle_audio_chunk(websocket: WebSocket, session: RealtimeSession, mes
         
         # Stream to STT if available
         if session.stt_manager and session.state == ConversationState.LISTENING:
-            async for result in session.stt_manager.stream_audio(decoded_audio):
-                # Send transcription updates
+            # Ensure streaming is active
+            if not session.is_streaming:
+                log_warning(f"⚠️ STT manager exists but streaming not active", session_id=session.session.session_id)
+                # Try to restart streaming
+                stt_initialized = await session.initialize_stt()
+                if not stt_initialized:
+                    await websocket.send_json({
+                        "type": "error",
+                        "error_type": "stt_error",
+                        "message": "STT streaming not available"
+                    })
+                    return
+            
+            try:
+                log_debug(f"🎤 Streaming audio chunk to STT, size: {len(decoded_audio)} bytes", session_id=session.session.session_id)
+                
+                async for result in session.stt_manager.stream_audio(decoded_audio):
+                    log_debug(f"📝 STT result: {result.text}, final: {result.is_final}", session_id=session.session.session_id)
+                    
+                    # Send transcription updates
+                    await websocket.send_json({
+                        "type": "transcription",
+                        "text": result.text,
+                        "is_final": result.is_final,
+                        "confidence": result.confidence
+                    })
+                    
+                    if result.is_final:
+                        session.current_transcription = result.text
+                        log_info(f"📝 Final transcription: {result.text}", session_id=session.session.session_id)
+                        
+            except Exception as e:
+                log_error(f"❌ STT streaming error", error=str(e), traceback=traceback.format_exc(), session_id=session.session.session_id)
                 await websocket.send_json({
-                    "type": "transcription",
-                    "text": result.text,
-                    "is_final": result.is_final,
-                    "confidence": result.confidence
+                    "type": "error",
+                    "error_type": "stt_error",
+                    "message": f"STT error: {str(e)}"
                 })
                 
-                if result.is_final:
-                    session.current_transcription = result.text
-                    log_info(f"📝 Final transcription: {result.text}", session_id=session.session.session_id)
-        
         # Process if silence detected and we have transcription
         if silence_duration > session.silence_threshold_ms and session.current_transcription:
             log_info(