feat: VAD-based barge-in during TTS playback

satellite/tts.py

@@ -4,7 +4,7 @@ import subprocess
 import threading
 from elevenlabs import VoiceSettings
 
-from .config import AUDIO_SINK, AGENTS, log
+from .config import AUDIO_SINK, AGENTS, SILENCE_THRESHOLD, log
 
 ELEVENLABS_API_KEY = os.getenv("ELEVENLABS_API_KEY", "")
 ELEVENLABS_MODEL = os.getenv("ELEVENLABS_MODEL", "eleven_flash_v2_5")
@@ -40,6 +40,43 @@ def is_speaking() -> bool:
         return _current_process is not None and _current_process.poll() is None
 
 
+_barge_in_flag = threading.Event()
+
+def start_barge_in_listener():
+    """Запускает фоновый поток VAD — если услышал голос во время TTS, ставит флаг barge-in."""
+    _barge_in_flag.clear()
+
+    def _listen():
+        import pyaudio
+        import numpy as np
+        try:
+            audio = pyaudio.PyAudio()
+            stream = audio.open(format=pyaudio.paInt16, channels=1, rate=16000,
+                                input=True, frames_per_buffer=1024)
+            warmup = 8  # ~0.5s прогрев чтобы не словить эхо начала TTS
+            i = 0
+            while is_speaking():
+                data = stream.read(1024, exception_on_overflow=False)
+                i += 1
+                if i < warmup:
+                    continue
+                amplitude = np.abs(np.frombuffer(data, dtype=np.int16)).mean()
+                if amplitude > SILENCE_THRESHOLD * 1.5:  # порог чуть выше чем для записи
+                    _barge_in_flag.set()
+                    stop_speaking()
+                    break
+            stream.stop_stream()
+            audio.terminate()
+        except Exception:
+            pass
+
+    t = threading.Thread(target=_listen, daemon=True)
+    t.start()
+    return t
+
+def was_barge_in() -> bool:
+    return _barge_in_flag.is_set()
+
 def _mpv_cmd() -> list[str]:
     """Команда mpv для воспроизведения из stdin"""
     mpv_bin = os.getenv("MPV_PATH", "mpv")