Fix all hardcoded Kyutai model identifiers across codebase

- Update speech_processor.py default model from kyutai/stt-2.6b-en_fr-trfs to kyutai/stt-2.6b-en-trfs
- Update communication.py fallback model identifiers
- Update performance_optimizer.py speech model default
- Ensures consistent use of correct Kyutai model identifier throughout application
- Fixes speech model loading errors in HuggingFace Spaces deployment

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>

digipal/ai/communication.py

@@ -55,7 +55,7 @@ class AICommunication:
         self._model_loaded = False
         
         # Initialize speech processor
-        speech_model_id = self.kyutai_config.get('model_id', 'kyutai/stt-2.6b-en_fr-trfs')
+        speech_model_id = self.kyutai_config.get('model_id', 'kyutai/stt-2.6b-en-trfs')
         speech_device = self.kyutai_config.get('device', None)
         self.speech_processor = SpeechProcessor(speech_model_id, speech_device)
         self._speech_model_loaded = False
@@ -294,7 +294,7 @@ class AICommunication:
             return self.speech_processor.get_model_info()
         else:
             return {
-                'model_id': self.kyutai_config.get('model_id', 'kyutai/stt-2.6b-en_fr-trfs'),
+                'model_id': self.kyutai_config.get('model_id', 'kyutai/stt-2.6b-en-trfs'),
                 'loaded': False
             }
     

digipal/ai/speech_processor.py

@@ -259,7 +259,7 @@ class SpeechProcessor:
     Main speech processing class using Kyutai speech-to-text models.
     """
     
-    def __init__(self, model_id: str = "kyutai/stt-2.6b-en_fr-trfs", device: Optional[str] = None):
+    def __init__(self, model_id: str = "kyutai/stt-2.6b-en-trfs", device: Optional[str] = None):
         """
         Initialize speech processor with Kyutai model.
         

digipal/core/performance_optimizer.py

@@ -121,7 +121,7 @@ class LazyModelLoader:
                 # Direct loading without caching
                 return self._load_language_model(model_name)
     
-    def get_speech_model(self, model_name: str = "kyutai/stt-2.6b-en_fr-trfs") -> Tuple[Any, Any]:
+    def get_speech_model(self, model_name: str = "kyutai/stt-2.6b-en-trfs") -> Tuple[Any, Any]:
         """
         Get speech model with lazy loading.