Fix choppy/distorted audio to live caller

- Mute host mic forwarding while TTS is streaming to prevent interleaving
  both audio sources into the same playback buffer
- Replace nearest-neighbor downsampling with box-filter averaging on both
  server (host mic) and browser (caller mic) for anti-aliased resampling

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend/services/audio.py

@@ -472,9 +472,10 @@ class AudioService:
                 if not self._host_send_callback:
                     return
                 mono = indata[:, record_channel]
-                # Simple decimation to ~16kHz
+                # Downsample to ~16kHz with averaging (anti-aliased)
                 if step > 1:
-                    mono = mono[::step]
+                    n = len(mono) // step * step
+                    mono = mono[:n].reshape(-1, step).mean(axis=1)
 
                 host_accum.append(mono.copy())
                 host_accum_samples[0] += len(mono)

backend/services/caller_service.py

@@ -19,6 +19,7 @@ class CallerService:
         self._caller_counter: int = 0
         self._lock = threading.Lock()
         self._websockets: dict[str, any] = {}  # caller_id -> WebSocket
+        self.streaming_tts: bool = False  # True while TTS audio is being streamed
 
     def add_to_queue(self, caller_id: str, name: str):
         with self._lock:
@@ -139,6 +140,7 @@ class CallerService:
         if not ws:
             return
 
+        self.streaming_tts = True
         try:
             audio = np.frombuffer(pcm_data, dtype=np.int16).astype(np.float32) / 32768.0
             if sample_rate != 16000:
@@ -160,6 +162,8 @@ class CallerService:
 
         except Exception as e:
             print(f"[Caller] Failed to stream audio: {e}")
+        finally:
+            self.streaming_tts = False
 
     async def notify_caller(self, caller_id: str, message: dict):
         """Send JSON control message to caller"""