Update websocket_handler.py

websocket_handler.py

@@ -313,25 +313,35 @@ class RealtimeSession:
             self.last_stt_stop_time = datetime.now()
 
     async def restart_stt_if_needed(self):
-        """Restart STT if it's not active"""
+        """Restart STT streaming if needed"""
         try:
-            # Sadece LISTENING state'inde ve WebSocket aktifse restart yap
-            if not self.is_streaming and self.is_websocket_active and self.state == ConversationState.LISTENING:
-                log_info(f"🔄 Restarting STT stream (session #{self.stt_session_count} -> #{self.stt_session_count + 1})", 
-                        session_id=self.session.session_id)
-                
-                # Yeni session başlat (initialize_stt zaten stop_stt_streaming'i çağırıyor)
-                stt_initialized = await self.initialize_stt()
-                if stt_initialized:
-                    log_info(f"✅ STT stream restarted successfully", session_id=self.session.session_id)
-                    return True
-                else:
-                    log_error(f"❌ Failed to restart STT stream", session_id=self.session.session_id)
+            # STT yoksa baştan oluştur
+            if not self.stt_manager:
+                await self.create_stt_manager()
+                if not self.stt_manager:
+                    log_error(f"❌ Failed to create STT manager", session_id=self.session.session_id)
                     return False
+            
+            # Streaming başlat
+            config = ConfigProvider.get().global_config.stt_provider.settings
+            stt_config = {
+                'language': self.get_stt_language(),
+                'interim_results': True,
+                'single_utterance': False,  # Continuous listening için False
+                'enable_punctuation': True,
+                'sample_rate': 16000,
+                'encoding': 'LINEAR16'  # WEBM_OPUS yerine LINEAR16 kullan
+            }
+            
+            await self.stt_manager.start_streaming(stt_config)
+            self.is_streaming = True
+            
+            log_info(f"✅ STT streaming started successfully with clean state", session_id=self.session.session_id)
             return True
+            
         except Exception as e:
-            log_error(f"❌ Error restarting STT", error=str(e), traceback=traceback.format_exc(), 
-                     session_id=self.session.session_id)
+            log_error(f"❌ Failed to restart STT", error=str(e), traceback=traceback.format_exc(), session_id=self.session.session_id)
+            self.is_streaming = False
             return False
     
     async def change_state(self, new_state: ConversationState):
@@ -513,12 +523,8 @@ async def handle_audio_chunk(websocket: WebSocket, session: RealtimeSession, mes
             
             if session.chunk_counter == 1:
                 log_info(f"🎤 Started streaming audio to STT", session_id=session.session.session_id)
-                # İlk chunk'ta format kontrolü yap
-                if len(decoded_audio) >= 4:
-                    if decoded_audio[:4] == b'\x1a\x45\xdf\xa3':
-                        log_info(f"✅ Valid WEBM header detected", session_id=session.session.session_id)
-                    else:
-                        log_warning(f"⚠️ Unknown audio format, first 4 bytes: {decoded_audio[:4].hex()}", session_id=session.session.session_id)
+                # İlk chunk log'u - format kontrolü kaldırıldı
+                log_info(f"📤 First chunk - size: {len(decoded_audio)} bytes", session_id=session.session.session_id)
             elif session.chunk_counter % 100 == 0:
                 log_info(f"📊 Sent {session.chunk_counter} chunks to STT so far...", session_id=session.session.session_id)
             
@@ -545,11 +551,14 @@ async def handle_audio_chunk(websocket: WebSocket, session: RealtimeSession, mes
                         
                         # State'i değiştir
                         await session.change_state(ConversationState.PROCESSING_STT)
-                        await websocket.send_json({
-                            "type": "state_change",
-                            "from": "listening",
-                            "to": "processing_stt"
-                        })
+                        
+                        # State change mesajı gönder
+                        if session.is_websocket_active:
+                            await websocket.send_json({
+                                "type": "state_change",
+                                "from": "listening",
+                                "to": "processing_stt"
+                            })
                         
                         # Process user input
                         await process_user_input(websocket, session)
@@ -577,7 +586,7 @@ async def handle_audio_chunk(websocket: WebSocket, session: RealtimeSession, mes
             "error_type": "audio_error",
             "message": f"Audio processing error: {str(e)}"
         })
-
+        
 # ========================= MAIN HANDLER =========================
 async def websocket_endpoint(websocket: WebSocket, session_id: str):
     """Main WebSocket endpoint for real-time conversation"""