🔧 Add FFmpeg system dependency and enhance ASR error handling

- Add packages.txt with ffmpeg dependency for Hugging Face Spaces
- Enhanced voice_service.py to check FFmpeg availability at startup
- Add graceful fallback to browser-native ASR when FFmpeg is missing
- Improve error handling for audio processing dependencies
- Add subprocess check for FFmpeg availability before Whisper initialization

packages.txt

@@ -0,0 +1 @@
+ffmpeg

voice_service.py

@@ -57,10 +57,20 @@ class VoiceService:
             if self.asr_provider == "whisper":
                 try:
                     import whisper
+                    import subprocess
+                    
                     # Verify it's the correct whisper package
                     if not hasattr(whisper, 'load_model'):
                         raise ImportError("Wrong whisper package - need openai-whisper")
                     
+                    # Check if FFmpeg is available
+                    try:
+                        subprocess.run(['ffmpeg', '-version'], capture_output=True, check=True)
+                        logger.info("✅ FFmpeg is available")
+                    except (subprocess.CalledProcessError, FileNotFoundError):
+                        logger.warning("⚠️ FFmpeg not found - Whisper may not work properly")
+                        raise ImportError("FFmpeg not available")
+                    
                     # Use base model for balance between speed and accuracy
                     self.whisper_model = whisper.load_model("base")
                     self.asr_available = True
@@ -215,8 +225,16 @@ class VoiceService:
                 return "Server-side ASR not available - using browser ASR"
                 
         except Exception as e:
-            logger.error(f"❌ ASR Error: {e}")
-            return None
+            error_msg = str(e).lower()
+            if 'ffmpeg' in error_msg or 'no such file or directory' in error_msg:
+                logger.error(f"❌ FFmpeg Error: {e}")
+                logger.info("🔄 FFmpeg not available - switching to browser-native ASR")
+                # Update provider to browser-native for future requests
+                self.asr_provider = "browser-native"
+                return None
+            else:
+                logger.error(f"❌ ASR Error: {e}")
+                return None
 
     def get_available_voices(self) -> Dict[str, Any]:
         """Get list of available TTS voices"""