feat: harden audio device compatibility across machines

app/audio/stt.py

@@ -83,10 +83,12 @@ class SpeechRecognizer:
     def __init__(self):
         self.dg_client = None
         self.pa = None
+        self.audio_manager = None
         self.stream = None
         self.transcript = ""
         self.last_successful_operation = datetime.now()
         self._input_device_index = None
+        self._stream_sample_rate = SAMPLE_RATE
 
     def initialize(self):
         """Инициализация клиента Deepgram и PyAudio."""
@@ -103,9 +105,9 @@ class SpeechRecognizer:
             print(f"❌ Ошибка при создании клиента Deepgram: {e}")
             raise
 
-        audio_manager = get_audio_manager()
-        self.pa = audio_manager.get_pyaudio()
-        self._input_device_index = audio_manager.get_input_device_index()
+        self.audio_manager = get_audio_manager()
+        self.pa = self.audio_manager.get_pyaudio()
+        self._input_device_index = self.audio_manager.get_input_device_index()
         print("✅ Deepgram клиент готов")
         # Обновляем время последней успешной операции
         self.last_successful_operation = datetime.now()
@@ -131,18 +133,23 @@ class SpeechRecognizer:
     def _get_stream(self):
         """Открывает аудиопоток PyAudio, если он еще не открыт."""
         if self.stream is None:
-            kwargs = {}
-            if self._input_device_index is not None:
-                kwargs["input_device_index"] = self._input_device_index
-
-            self.stream = self.pa.open(
-                rate=SAMPLE_RATE,
-                channels=1,
-                format=pyaudio.paInt16,
-                input=True,
-                frames_per_buffer=4096,
-                **kwargs,
+            if self.audio_manager is None:
+                self.audio_manager = get_audio_manager()
+            self.stream, self._input_device_index, self._stream_sample_rate = (
+                self.audio_manager.open_input_stream(
+                    rate=SAMPLE_RATE,
+                    channels=1,
+                    format=pyaudio.paInt16,
+                    frames_per_buffer=4096,
+                    preferred_index=self._input_device_index,
+                    fallback_rates=[48000, 44100, 32000, 22050, 16000, 8000],
+                )
             )
+            if self._stream_sample_rate != SAMPLE_RATE:
+                print(
+                    f"⚠️ STT mic stream uses fallback rate={self._stream_sample_rate} "
+                    f"(requested {SAMPLE_RATE})"
+                )
         return self.stream
 
     async def _process_audio(
@@ -242,7 +249,7 @@ class SpeechRecognizer:
             smart_format=True,  # Расстановка знаков препинания
             encoding="linear16",
             channels=1,
-            sample_rate=SAMPLE_RATE,
+            sample_rate=self._stream_sample_rate,
             interim_results=True,
             utterance_end_ms=int(POST_SPEECH_SILENCE_TIMEOUT_SECONDS * 1000),
             vad_events=True,