Update websocket_handler.py

websocket_handler.py

@@ -176,23 +176,41 @@ class RealtimeSession:
         self.silence_threshold_ms = silence_threshold
         
     async def initialize_stt(self):
-        """Initialize STT provider"""
-        try:
-            self.stt_manager = STTFactory.create_provider()
-            if self.stt_manager:
-                config = ConfigProvider.get().global_config.stt_provider.settings
-                await self.stt_manager.start_streaming({
-                    "language": config.get("language", "tr-TR"),
-                    "interim_results": config.get("interim_results", True),
-                    "single_utterance": False,
-                    "enable_punctuation": config.get("enable_punctuation", True),
-                    "sample_rate": 16000,
-                    "encoding": "WEBM_OPUS"
-                })
-                log_info("STT manager initialized", session_id=self.session.session_id)
-                return True
-        except Exception as e:
-            log_error(f"Failed to initialize STT", error=str(e), session_id=self.session.session_id)
+    """Initialize STT provider"""
+    try:
+        self.stt_manager = STTFactory.create_provider()
+        if self.stt_manager and self.stt_manager.supports_realtime():
+            config = ConfigProvider.get().global_config.stt_provider.settings
+            
+            # STTConfig nesnesi oluştur, dict değil!
+            from stt_interface import STTConfig
+            stt_config = STTConfig(
+                language=config.get("language", "tr-TR"),
+                interim_results=config.get("interim_results", True),
+                single_utterance=False,
+                enable_punctuation=config.get("enable_punctuation", True),
+                sample_rate=16000,
+                encoding="WEBM_OPUS",
+                model=config.get("model", "latest_long"),
+                use_enhanced=config.get("use_enhanced", True),
+                # Voice Activity Detection
+                vad_enabled=True,
+                speech_timeout_ms=config.get("speech_timeout_ms", 2000),
+                # Noise reduction
+                noise_reduction_enabled=True,
+                noise_reduction_level=config.get("noise_reduction_level", 2)
+            )
+            
+            await self.stt_manager.start_streaming(stt_config)
+            log_info("STT manager initialized", session_id=self.session.session_id)
+            return True
+        else:
+            log_warning("STT provider does not support realtime", session_id=self.session.session_id)
+            return False
+    except Exception as e:
+        log_error(f"Failed to initialize STT", error=str(e), session_id=self.session.session_id)
+        # STT başarısız oldu, manager'ı None yap
+        self.stt_manager = None
         return False
     
     async def change_state(self, new_state: ConversationState):
@@ -252,11 +270,19 @@ async def websocket_endpoint(websocket: WebSocket, session_id: str):
     # Initialize STT
     stt_initialized = await realtime_session.initialize_stt()
     if not stt_initialized:
+        # STT başarısız oldu, kullanıcıya bildir ve bağlantıyı kapat
         await websocket.send_json({
             "type": "error",
-            "message": "STT initialization failed"
+            "message": "Speech-to-Text service initialization failed. Please check your configuration.",
+            "error_type": "stt_init_failed"
         })
-    
+        
+        # Cleanup ve close
+        await realtime_session.cleanup()
+        await websocket.close()
+        return
+
+    # STT başarılı, devam et
     try:
         while True:
             # Receive message
@@ -297,7 +323,17 @@ async def handle_audio_chunk(websocket: WebSocket, session: RealtimeSession, mes
         audio_data = message.get("data")
         if not audio_data:
             return
-        
+
+        # STT manager kontrolü
+        if not session.stt_manager:
+            log_warning("No STT manager available, ignoring audio chunk", session_id=session.session.session_id)
+            await websocket.send_json({
+                "type": "error",
+                "message": "Speech recognition not available",
+                "error_type": "stt_unavailable"
+            })
+            return
+            
         # Check for barge-in during TTS/audio playback
         if session.state in [ConversationState.PLAYING_AUDIO, ConversationState.PROCESSING_TTS]:
             await session.handle_barge_in()