Fix CUDA runtime error by forcing CPU-only execution

- Force device="cpu" to avoid CUDA driver version mismatch
- Add error handling with fallback to base model
- Resolves RuntimeError: CUDA driver version is insufficient

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

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

app/services/asr.py

@@ -7,11 +7,20 @@ WHISPER_MODEL_SIZE = os.getenv("WHISPER_MODEL_SIZE", "base")
 WHISPER_DEVICE = os.getenv("WHISPER_DEVICE", "cpu")
 WHISPER_COMPUTE_TYPE = os.getenv("WHISPER_COMPUTE_TYPE", "int8")
 
-_whisper = WhisperModel(
-    WHISPER_MODEL_SIZE,
-    device=WHISPER_DEVICE,
-    compute_type=WHISPER_COMPUTE_TYPE
-)
+try:
+    _whisper = WhisperModel(
+        WHISPER_MODEL_SIZE,
+        device="cpu",  # Force CPU to avoid CUDA issues
+        compute_type="int8"
+    )
+except Exception as e:
+    print(f"Error loading Whisper model: {e}")
+    # Fallback to base model with explicit CPU settings
+    _whisper = WhisperModel(
+        "base",
+        device="cpu",
+        compute_type="int8"
+    )
 
 def transcribe(audio_path: str) -> List[Dict]:
     """