Update app.py

app.py

@@ -28,7 +28,7 @@ llama_cloud_api_key = os.environ.get("LLAMA_CLOUD_API_KEY")
 ACCOUNT_ID = os.environ.get("CLOUDFARE_ACCOUNT_ID")
 API_TOKEN = os.environ.get("CLOUDFLARE_AUTH_TOKEN")
 API_BASE_URL = "https://api.cloudflare.com/client/v4/accounts/a17f03e0f049ccae0c15cdcf3b9737ce/ai/run/"
-whisper_api = InferenceClient("openai/whisper-small", token=huggingface_token)
+whisper_api = InferenceClient("openai/whisper-large-v3", token=huggingface_token)
 
 print(f"ACCOUNT_ID: {ACCOUNT_ID}")
 print(f"CLOUDFLARE_AUTH_TOKEN: {API_TOKEN[:5]}..." if API_TOKEN else "Not set")
@@ -616,26 +616,42 @@ Write a detailed and complete response that answers the following user question:
         logging.info("Finished generating response")
 
 def transcribe(audio_file):
+    print(f"Received audio file: {audio_file}")
     if audio_file is None:
-        return ""
-    
-    with open(audio_file, "rb") as f:
-        audio_data = f.read()
-    
-    # Create a file-like object from the audio data
-    audio_file = io.BytesIO(audio_data)
-    audio_file.name = "audio.wav"  # The name is important for the API to recognize the file type
+        return "No audio file provided"
     
     try:
-        # Use the automatic_speech_recognition method
-        response = whisper_api.automatic_speech_recognition(audio=audio_file)
+        with open(audio_file, "rb") as f:
+            audio_data = f.read()
+        
+        print(f"Read {len(audio_data)} bytes from audio file")
+        
+        # Prepare the inputs
+        inputs = {
+            "audio": audio_data
+        }
+        
+        # Prepare the parameters (you can adjust these as needed)
+        parameters = {
+            "return_timestamps": False,
+            "language": "en"  # or "auto" for automatic language detection
+        }
+        
+        print("Calling automatic_speech_recognition")
+        
+        response = whisper_api.automatic_speech_recognition(
+            inputs=inputs,
+            parameters=parameters
+        )
+        
+        print(f"Received response: {response}")
         
-        # The response should be a dictionary with a 'text' key
         return response["text"] if isinstance(response, dict) and "text" in response else str(response)
     except Exception as e:
         print(f"Error in transcription: {str(e)}")
         return f"Error in transcription: {str(e)}"
 
+
 def vote(data: gr.LikeData):
     if data.liked:
         print(f"You upvoted this response: {data.value}")