Update websocket_handler.py

websocket_handler.py

@@ -600,6 +600,10 @@ async def generate_and_stream_tts(
         # Generate audio
         audio_data = await tts_provider.synthesize(text)
         
+        # Debug log to check audio data
+        log_debug(f"Audio data type: {type(audio_data)}, length: {len(audio_data)}")
+        log_debug(f"First 10 bytes: {audio_data[:10]}")
+        
         # Change state to playing
         await session.change_state(ConversationState.PLAYING_AUDIO)
         await websocket.send_json({
@@ -620,13 +624,21 @@ async def generate_and_stream_tts(
             chunk = audio_data[i:i + chunk_size]
             chunk_index = i // chunk_size
             
+            # IMPORTANT: Ensure chunk is bytes before encoding
+            if isinstance(chunk, str):
+                # If chunk is already a string, it might be base64 already
+                chunk_base64 = chunk
+            else:
+                # Convert bytes to base64
+                chunk_base64 = base64.b64encode(chunk).decode('utf-8')
+            
             await websocket.send_json({
                 "type": "tts_audio",
-                "data": base64.b64encode(chunk).decode('utf-8'),
+                "data": chunk_base64,
                 "chunk_index": chunk_index,
                 "total_chunks": total_chunks,
                 "is_last": chunk_index == total_chunks - 1,
-                "mime_type": "audio/mpeg"  # MP3 format for ElevenLabs
+                "mime_type": "audio/mpeg"
             })
             
             # Small delay to prevent overwhelming the client
@@ -654,6 +666,7 @@ async def generate_and_stream_tts(
         log_error(
             f"TTS generation error",
             error=str(e),
+            traceback=traceback.format_exc(),
             session_id=session.session.session_id
         )
         await websocket.send_json({