Merge remote-tracking branch 'refs/remotes/origin/main'

core/__init__.py

@@ -1,4 +1,20 @@
-from realtime_voice_service.core.session import CallSession, SessionState
-from realtime_voice_service.core.vad import BaseVAD, SileroVADDetector, VADFrameResult
-
 __all__ = ["BaseVAD", "CallSession", "SessionState", "SileroVADDetector", "VADFrameResult"]
+
+
+def __getattr__(name: str):
+    if name in {"CallSession", "SessionState"}:
+        from realtime_voice_service.core.session import CallSession, SessionState
+
+        return {
+            "CallSession": CallSession,
+            "SessionState": SessionState,
+        }[name]
+    if name in {"BaseVAD", "SileroVADDetector", "VADFrameResult"}:
+        from realtime_voice_service.core.vad import BaseVAD, SileroVADDetector, VADFrameResult
+
+        return {
+            "BaseVAD": BaseVAD,
+            "SileroVADDetector": SileroVADDetector,
+            "VADFrameResult": VADFrameResult,
+        }[name]
+    raise AttributeError(name)

core/session.py

@@ -2153,13 +2153,7 @@ class CallSession:
             timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S_%f")
             safe_session_id = re.sub(r"[^A-Za-z0-9_.-]+", "_", self.session_id)
             filename = os.path.join(dump_dir, f"utterance_{safe_session_id}_{timestamp}.wav")
-            dump_sample_rate_hz = 8000
-            if self.transport.sample_rate_hz != dump_sample_rate_hz:
-                LOGGER.warning(
-                    "realtime session %s audio dump writing raw bytes with 8000Hz header while transport sample_rate=%s",
-                    self.session_id,
-                    self.transport.sample_rate_hz,
-                )
+            dump_sample_rate_hz = self.transport.sample_rate_hz
 
             with wave.open(filename, "wb") as wav_file:
                 wav_file.setnchannels(1)

core/vad.py

@@ -54,7 +54,7 @@ class SileroVADDetector(BaseVAD):
     def __init__(
         self,
         *,
-        sample_rate_hz: int = 8000,
+        sample_rate_hz: int = 16000,
         threshold: float = 0.5,
         negative_threshold: float | None = None,
         speech_end_silence_ms: int = 1600,