Spaces:
Building
Building
Update websocket_handler.py
Browse files- websocket_handler.py +75 -63
websocket_handler.py
CHANGED
@@ -473,74 +473,86 @@ async def handle_audio_chunk(websocket: WebSocket, session: RealtimeSession, mes
|
|
473 |
# Check silence
|
474 |
silence_duration = session.silence_detector.update(decoded_audio)
|
475 |
|
476 |
-
|
477 |
-
|
478 |
-
|
479 |
-
|
480 |
-
|
481 |
-
|
482 |
-
|
483 |
-
|
484 |
-
|
485 |
-
|
486 |
-
|
487 |
-
|
488 |
-
|
489 |
-
|
490 |
-
|
491 |
-
try:
|
492 |
-
# Chunk counter - sadece önemli milestone'larda logla
|
493 |
-
if not hasattr(session, 'chunk_counter'):
|
494 |
-
session.chunk_counter = 0
|
495 |
-
session.chunk_counter += 1
|
496 |
-
|
497 |
-
if session.chunk_counter == 1:
|
498 |
-
log_info(f"🎤 Started streaming audio to STT", session_id=session.session.session_id)
|
499 |
-
elif session.chunk_counter % 100 == 0:
|
500 |
-
log_info(f"📊 Sent {session.chunk_counter} chunks to STT so far...", session_id=session.session.session_id)
|
501 |
|
502 |
-
|
503 |
-
|
504 |
-
|
505 |
-
|
506 |
-
|
507 |
|
508 |
-
|
509 |
-
|
510 |
-
|
511 |
-
"
|
512 |
-
"is_final": result.is_final,
|
513 |
-
"confidence": result.confidence
|
514 |
-
})
|
515 |
|
516 |
-
|
517 |
-
|
518 |
-
|
|
|
|
|
519 |
|
520 |
-
#
|
521 |
-
|
522 |
-
|
523 |
-
|
524 |
-
|
525 |
-
|
526 |
-
|
527 |
-
|
528 |
-
|
529 |
-
|
530 |
-
|
531 |
-
await process_user_input(websocket, session)
|
532 |
-
|
533 |
-
# STT'den final result geldiğinde audio buffer'ı ve transcription'ı resetle
|
534 |
-
await session.reset_for_new_utterance()
|
535 |
-
return # Bu audio chunk için işlem tamamlandı
|
536 |
|
537 |
-
|
538 |
-
|
539 |
-
|
540 |
-
|
541 |
-
|
542 |
-
|
543 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
544 |
|
545 |
async def handle_control_message(websocket: WebSocket, session: RealtimeSession, message: Dict[str, Any]):
|
546 |
"""Handle control messages"""
|
|
|
473 |
# Check silence
|
474 |
silence_duration = session.silence_detector.update(decoded_audio)
|
475 |
|
476 |
+
# Stream to STT if available
|
477 |
+
if session.stt_manager and session.state == ConversationState.LISTENING:
|
478 |
+
# Ensure streaming is active
|
479 |
+
if not session.is_streaming:
|
480 |
+
log_warning(f"⚠️ STT manager exists but streaming not active", session_id=session.session.session_id)
|
481 |
+
# Try to restart streaming
|
482 |
+
stt_initialized = await session.initialize_stt()
|
483 |
+
if not stt_initialized:
|
484 |
+
await websocket.send_json({
|
485 |
+
"type": "error",
|
486 |
+
"error_type": "stt_error",
|
487 |
+
"message": "STT streaming not available"
|
488 |
+
})
|
489 |
+
return
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
490 |
|
491 |
+
try:
|
492 |
+
# Chunk counter - sadece önemli milestone'larda logla
|
493 |
+
if not hasattr(session, 'chunk_counter'):
|
494 |
+
session.chunk_counter = 0
|
495 |
+
session.chunk_counter += 1
|
496 |
|
497 |
+
if session.chunk_counter == 1:
|
498 |
+
log_info(f"🎤 Started streaming audio to STT", session_id=session.session.session_id)
|
499 |
+
elif session.chunk_counter % 100 == 0:
|
500 |
+
log_info(f"📊 Sent {session.chunk_counter} chunks to STT so far...", session_id=session.session.session_id)
|
|
|
|
|
|
|
501 |
|
502 |
+
# STT'ye gönder ve sonuçları bekle
|
503 |
+
async for result in session.stt_manager.stream_audio(decoded_audio):
|
504 |
+
# Sadece anlamlı sonuçları logla
|
505 |
+
if result.text.strip(): # Boş olmayan text varsa
|
506 |
+
log_info(f"🎤 STT: '{result.text}' (final: {result.is_final})", session_id=session.session.session_id)
|
507 |
|
508 |
+
# Send transcription updates
|
509 |
+
await websocket.send_json({
|
510 |
+
"type": "transcription",
|
511 |
+
"text": result.text,
|
512 |
+
"is_final": result.is_final,
|
513 |
+
"confidence": result.confidence
|
514 |
+
})
|
515 |
+
|
516 |
+
if result.is_final:
|
517 |
+
session.current_transcription = result.text
|
518 |
+
log_info(f"✅ FINAL TRANSCRIPTION: '{result.text}'", session_id=session.session.session_id)
|
|
|
|
|
|
|
|
|
|
|
519 |
|
520 |
+
# Final transcription geldiğinde hemen işle
|
521 |
+
if session.current_transcription:
|
522 |
+
# State'i değiştir ve user input'u işle
|
523 |
+
await session.change_state(ConversationState.PROCESSING_STT)
|
524 |
+
await websocket.send_json({
|
525 |
+
"type": "state_change",
|
526 |
+
"from": "listening",
|
527 |
+
"to": "processing_stt"
|
528 |
+
})
|
529 |
+
|
530 |
+
# Process user input
|
531 |
+
await process_user_input(websocket, session)
|
532 |
+
|
533 |
+
# STT'den final result geldiğinde audio buffer'ı ve transcription'ı resetle
|
534 |
+
await session.reset_for_new_utterance()
|
535 |
+
return # Bu audio chunk için işlem tamamlandı
|
536 |
+
|
537 |
+
except Exception as e:
|
538 |
+
log_error(f"❌ STT streaming error", error=str(e), traceback=traceback.format_exc(), session_id=session.session.session_id)
|
539 |
+
await websocket.send_json({
|
540 |
+
"type": "error",
|
541 |
+
"error_type": "stt_error",
|
542 |
+
"message": f"STT error: {str(e)}"
|
543 |
+
})
|
544 |
+
|
545 |
+
except Exception as e:
|
546 |
+
log_error(
|
547 |
+
f"❌ Audio chunk handling error",
|
548 |
+
error=str(e),
|
549 |
+
traceback=traceback.format_exc(),
|
550 |
+
session_id=session.session.session_id
|
551 |
+
)
|
552 |
+
await websocket.send_json({
|
553 |
+
"type": "error",
|
554 |
+
"message": f"Audio processing error: {str(e)}"
|
555 |
+
})
|
556 |
|
557 |
async def handle_control_message(websocket: WebSocket, session: RealtimeSession, message: Dict[str, Any]):
|
558 |
"""Handle control messages"""
|