File size: 20,917 Bytes
890fc3a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
81c06f1
890fc3a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
86714dd
890fc3a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
"""
WebSocket Handler for Real-time STT/TTS with Barge-in Support
"""
from fastapi import WebSocket, WebSocketDisconnect
from typing import Dict, Any, Optional
import json
import asyncio
import base64
from datetime import datetime
from collections import deque
from enum import Enum
import numpy as np
import traceback

from session import Session, session_store
from config_provider import ConfigProvider
from chat_handler import handle_new_message, handle_parameter_followup
from stt_factory import STTFactory
from tts_factory import TTSFactory
from logger import log_info, log_error, log_debug, log_warning

# ========================= CONSTANTS =========================
# Default values - will be overridden by config
DEFAULT_SILENCE_THRESHOLD_MS = 2000
DEFAULT_AUDIO_CHUNK_SIZE = 4096
DEFAULT_ENERGY_THRESHOLD = 0.01
DEFAULT_AUDIO_BUFFER_MAX_SIZE = 1000

# ========================= ENUMS =========================
class ConversationState(Enum):
    IDLE = "idle"
    LISTENING = "listening"
    PROCESSING_STT = "processing_stt"
    PROCESSING_LLM = "processing_llm"
    PROCESSING_TTS = "processing_tts"
    PLAYING_AUDIO = "playing_audio"

# ========================= CLASSES =========================
class AudioBuffer:
    """Thread-safe circular buffer for audio chunks"""
    def __init__(self, max_size: int = DEFAULT_AUDIO_BUFFER_MAX_SIZE):
        self.buffer = deque(maxlen=max_size)
        self.lock = asyncio.Lock()
        
    async def add_chunk(self, chunk_data: str):
        """Add base64 encoded audio chunk"""
        async with self.lock:
            decoded = base64.b64decode(chunk_data)
            self.buffer.append(decoded)
            
    async def get_all_audio(self) -> bytes:
        """Get all audio data concatenated"""
        async with self.lock:
            return b''.join(self.buffer)
    
    async def clear(self):
        """Clear buffer"""
        async with self.lock:
            self.buffer.clear()
    
    def size(self) -> int:
        """Get current buffer size"""
        return len(self.buffer)


class SilenceDetector:
    """Detect silence in audio stream"""
    def __init__(self, threshold_ms: int = DEFAULT_SILENCE_THRESHOLD_MS, energy_threshold: float = DEFAULT_ENERGY_THRESHOLD):
        self.threshold_ms = threshold_ms
        self.energy_threshold = energy_threshold
        self.silence_start = None
        self.sample_rate = 16000
        
    def update(self, audio_chunk: bytes) -> int:
        """Update with new audio chunk and return silence duration in ms"""
        if self.is_silence(audio_chunk):
            if self.silence_start is None:
                self.silence_start = datetime.now()
            silence_duration = (datetime.now() - self.silence_start).total_seconds() * 1000
            return int(silence_duration)
        else:
            self.silence_start = None
            return 0
    
    def is_silence(self, audio_chunk: bytes) -> bool:
        """Check if audio chunk is silence"""
        try:
            # Convert bytes to numpy array (assuming 16-bit PCM)
            audio_data = np.frombuffer(audio_chunk, dtype=np.int16)
            
            # Calculate RMS energy
            if len(audio_data) == 0:
                return True
                
            rms = np.sqrt(np.mean(audio_data.astype(float) ** 2))
            normalized_rms = rms / 32768.0  # Normalize for 16-bit audio
            
            return normalized_rms < self.energy_threshold
        except Exception as e:
            log_warning(f"Silence detection error: {e}")
            return False
    
    def reset(self):
        """Reset silence detection"""
        self.silence_start = None


class BargeInHandler:
    """Handle user interruptions during TTS playback"""
    def __init__(self):
        self.active_tts_task: Optional[asyncio.Task] = None
        self.is_interrupting = False
        self.lock = asyncio.Lock()
    
    async def start_tts_task(self, coro):
        """Start a cancellable TTS task"""
        async with self.lock:
            # Cancel any existing task
            if self.active_tts_task and not self.active_tts_task.done():
                self.active_tts_task.cancel()
                try:
                    await self.active_tts_task
                except asyncio.CancelledError:
                    pass
            
            # Start new task
            self.active_tts_task = asyncio.create_task(coro)
            return self.active_tts_task
    
    async def handle_interruption(self, current_state: ConversationState):
        """Handle barge-in interruption"""
        async with self.lock:
            self.is_interrupting = True
            
            # Cancel TTS if active
            if self.active_tts_task and not self.active_tts_task.done():
                log_info("Barge-in: Cancelling active TTS")
                self.active_tts_task.cancel()
                try:
                    await self.active_tts_task
                except asyncio.CancelledError:
                    pass
            
            # Reset flag after short delay
            await asyncio.sleep(0.5)
            self.is_interrupting = False


class RealtimeSession:
    """Manage a real-time conversation session"""
    def __init__(self, session: Session):
        self.session = session
        self.state = ConversationState.IDLE
        
        # Get settings from config
        config = ConfigProvider.get().global_config.stt_provider.settings
        
        # Initialize with config values or defaults
        silence_threshold = config.get("speech_timeout_ms", DEFAULT_SILENCE_THRESHOLD_MS)
        energy_threshold = config.get("energy_threshold", DEFAULT_ENERGY_THRESHOLD)
        buffer_max_size = config.get("audio_buffer_max_size", DEFAULT_AUDIO_BUFFER_MAX_SIZE)
        
        self.audio_buffer = AudioBuffer(max_size=buffer_max_size)
        self.silence_detector = SilenceDetector(
            threshold_ms=silence_threshold,
            energy_threshold=energy_threshold
        )
        self.barge_in_handler = BargeInHandler()
        self.stt_manager = None
        self.current_transcription = ""
        self.is_streaming = False
        self.lock = asyncio.Lock()
        
        # Store config for later use
        self.audio_chunk_size = config.get("audio_chunk_size", DEFAULT_AUDIO_CHUNK_SIZE)
        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)
        return False
    
    async def change_state(self, new_state: ConversationState):
        """Change conversation state"""
        async with self.lock:
            old_state = self.state
            self.state = new_state
            log_debug(
                f"State change: {old_state.value}{new_state.value}",
                session_id=self.session.session_id
            )
    
    async def handle_barge_in(self):
        """Handle user interruption"""
        await self.barge_in_handler.handle_interruption(self.state)
        await self.change_state(ConversationState.LISTENING)
    
    async def reset_for_new_utterance(self):
        """Reset for new user utterance"""
        await self.audio_buffer.clear()
        self.silence_detector.reset()
        self.current_transcription = ""
    
    async def cleanup(self):
        """Clean up resources"""
        try:
            if self.stt_manager:
                await self.stt_manager.stop_streaming()
            log_info(f"Cleaned up realtime session", session_id=self.session.session_id)
        except Exception as e:
            log_warning(f"Cleanup error", error=str(e), session_id=self.session.session_id)


# ========================= MAIN HANDLER =========================
async def websocket_endpoint(websocket: WebSocket, session_id: str):
    """Main WebSocket endpoint for real-time conversation"""
    await websocket.accept()
    log_info(f"WebSocket connected", session_id=session_id)
    
    # Get session
    session = session_store.get_session(session_id)
    if not session:
        await websocket.send_json({
            "type": "error",
            "message": "Session not found"
        })
        await websocket.close()
        return
    
    # Mark as realtime session
    session.is_realtime_session = True
    session_store.update_session(session)
    
    # Initialize conversation
    realtime_session = RealtimeSession(session)
    
    # Initialize STT
    stt_initialized = await realtime_session.initialize_stt()
    if not stt_initialized:
        await websocket.send_json({
            "type": "error",
            "message": "STT initialization failed"
        })
    
    try:
        while True:
            # Receive message
            message = await websocket.receive_json()
            message_type = message.get("type")
            
            if message_type == "audio_chunk":
                await handle_audio_chunk(websocket, realtime_session, message)
                
            elif message_type == "control":
                await handle_control_message(websocket, realtime_session, message)
                
            elif message_type == "ping":
                # Keep-alive ping
                await websocket.send_json({"type": "pong"})
                
    except WebSocketDisconnect:
        log_info(f"WebSocket disconnected", session_id=session_id)
    except Exception as e:
        log_error(
            f"WebSocket error",
            error=str(e),
            traceback=traceback.format_exc(),
            session_id=session_id
        )
        await websocket.send_json({
            "type": "error",
            "message": str(e)
        })
    finally:
        await realtime_session.cleanup()


# ========================= MESSAGE HANDLERS =========================
async def handle_audio_chunk(websocket: WebSocket, session: RealtimeSession, message: Dict[str, Any]):
    """Handle incoming audio chunk with barge-in support"""
    try:
        audio_data = message.get("data")
        if not audio_data:
            return
        
        # Check for barge-in during TTS/audio playback
        if session.state in [ConversationState.PLAYING_AUDIO, ConversationState.PROCESSING_TTS]:
            await session.handle_barge_in()
            await websocket.send_json({
                "type": "control",
                "action": "stop_playback"
            })
            log_info(f"Barge-in detected", session_id=session.session.session_id, state=session.state.value)
        
        # Change state to listening if idle
        if session.state == ConversationState.IDLE:
            await session.change_state(ConversationState.LISTENING)
            await websocket.send_json({
                "type": "state_change",
                "from": "idle",
                "to": "listening"
            })
        
        # Add to buffer - don't lose any audio
        await session.audio_buffer.add_chunk(audio_data)
        
        # Decode for processing
        decoded_audio = base64.b64decode(audio_data)
        
        # Check silence
        silence_duration = session.silence_detector.update(decoded_audio)
        
        # 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
                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
        
        # Process if silence detected and we have transcription
        if silence_duration > session.silence_threshold_ms and session.current_transcription:
            log_info(
                f"User stopped speaking",
                session_id=session.session.session_id,
                silence_ms=silence_duration,
                text=session.current_transcription
            )
            await process_user_input(websocket, session)
            
    except Exception as e:
        log_error(
            f"Audio chunk handling error",
            error=str(e),
            traceback=traceback.format_exc(),
            session_id=session.session.session_id
        )
        await websocket.send_json({
            "type": "error",
            "message": f"Audio processing error: {str(e)}"
        })


async def handle_control_message(websocket: WebSocket, session: RealtimeSession, message: Dict[str, Any]):
    """Handle control messages"""
    action = message.get("action")
    config = message.get("config", {})
    
    log_debug(f"Control message", action=action, session_id=session.session.session_id)
    
    if action == "start_session":
        # Session configuration
        await websocket.send_json({
            "type": "session_started",
            "session_id": session.session.session_id,
            "config": {
                "silence_threshold_ms": session.silence_threshold_ms,
                "audio_chunk_size": session.audio_chunk_size,
                "supports_barge_in": True
            }
        })
        
    elif action == "end_session":
        # Clean up and close
        await session.cleanup()
        await websocket.close()
        
    elif action == "interrupt":
        # Handle explicit interrupt
        await session.handle_barge_in()
        await websocket.send_json({
            "type": "control",
            "action": "interrupt_acknowledged"
        })
        
    elif action == "reset":
        # Reset conversation state
        await session.reset_for_new_utterance()
        await session.change_state(ConversationState.IDLE)
        await websocket.send_json({
            "type": "state_change",
            "from": session.state.value,
            "to": "idle"
        })
        
    elif action == "audio_ended":
        # Audio playback ended on client
        if session.state == ConversationState.PLAYING_AUDIO:
            await session.change_state(ConversationState.IDLE)
            await websocket.send_json({
                "type": "state_change",
                "from": "playing_audio",
                "to": "idle"
            })


# ========================= PROCESSING FUNCTIONS =========================
async def process_user_input(websocket: WebSocket, session: RealtimeSession):
    """Process complete user input"""
    try:
        user_text = session.current_transcription
        if not user_text:
            await session.reset_for_new_utterance()
            await session.change_state(ConversationState.IDLE)
            return
        
        log_info(f"Processing user input", text=user_text, session_id=session.session.session_id)
        
        # State: STT Processing
        await session.change_state(ConversationState.PROCESSING_STT)
        await websocket.send_json({
            "type": "state_change",
            "from": "listening",
            "to": "processing_stt"
        })
        
        # Send final transcription
        await websocket.send_json({
            "type": "transcription",
            "text": user_text,
            "is_final": True,
            "confidence": 0.95
        })
        
        # State: LLM Processing
        await session.change_state(ConversationState.PROCESSING_LLM)
        await websocket.send_json({
            "type": "state_change",
            "from": "processing_stt",
            "to": "processing_llm"
        })
        
        # Add to chat history
        session.session.add_message("user", user_text)
        
        # Get LLM response based on session state
        if session.session.state == "collect_params":
            response_text = await handle_parameter_followup(session.session, user_text)
        else:
            response_text = await handle_new_message(session.session, user_text)
        
        # Add response to history
        session.session.add_message("assistant", response_text)
        
        # Send text response
        await websocket.send_json({
            "type": "assistant_response",
            "text": response_text
        })
        
        # Generate TTS if enabled
        tts_provider = TTSFactory.create_provider()
        if tts_provider:
            await session.change_state(ConversationState.PROCESSING_TTS)
            await websocket.send_json({
                "type": "state_change",
                "from": "processing_llm",
                "to": "processing_tts"
            })
            
            # Generate TTS with barge-in support
            tts_task = session.barge_in_handler.start_tts_task(
                generate_and_stream_tts(websocket, session, tts_provider, response_text)
            )
            
            try:
                await tts_task
            except asyncio.CancelledError:
                log_info("TTS cancelled due to barge-in", session_id=session.session.session_id)
        else:
            # No TTS, go back to idle
            await session.change_state(ConversationState.IDLE)
            await websocket.send_json({
                "type": "state_change",
                "from": "processing_llm",
                "to": "idle"
            })
        
        # Reset for next input
        await session.reset_for_new_utterance()
        
    except Exception as e:
        log_error(
            f"Error processing user input",
            error=str(e),
            traceback=traceback.format_exc(),
            session_id=session.session.session_id
        )
        await websocket.send_json({
            "type": "error",
            "message": f"Processing error: {str(e)}"
        })
        await session.reset_for_new_utterance()
        await session.change_state(ConversationState.IDLE)


async def generate_and_stream_tts(
    websocket: WebSocket,
    session: RealtimeSession,
    tts_provider,
    text: str
):
    """Generate and stream TTS audio with cancellation support"""
    try:
        # Generate audio
        audio_data = await tts_provider.synthesize(text)
        
        # Change state to playing
        await session.change_state(ConversationState.PLAYING_AUDIO)
        await websocket.send_json({
            "type": "state_change",
            "from": "processing_tts",
            "to": "playing_audio"
        })
        
        # Stream audio in chunks
        chunk_size = session.audio_chunk_size
        total_chunks = (len(audio_data) + chunk_size - 1) // chunk_size
        
        for i in range(0, len(audio_data), chunk_size):
            # Check for cancellation
            if asyncio.current_task().cancelled():
                break
                
            chunk = audio_data[i:i + chunk_size]
            chunk_index = i // chunk_size
            
            await websocket.send_json({
                "type": "tts_audio",
                "data": base64.b64encode(chunk).decode('utf-8'),
                "chunk_index": chunk_index,
                "total_chunks": total_chunks,
                "is_last": chunk_index == total_chunks - 1
            })
            
            # Small delay to prevent overwhelming the client
            await asyncio.sleep(0.01)
        
        log_info(
            f"TTS streaming completed",
            session_id=session.session.session_id,
            text_length=len(text),
            audio_size=len(audio_data)
        )
        
    except asyncio.CancelledError:
        log_info("TTS streaming cancelled", session_id=session.session.session_id)
        raise
    except Exception as e:
        log_error(
            f"TTS generation error",
            error=str(e),
            session_id=session.session.session_id
        )
        await websocket.send_json({
            "type": "error",
            "message": f"TTS error: {str(e)}"
        })