Update app.py

app.py

@@ -94,7 +94,6 @@
     
 #     return Response("No audio generated", status_code=400)
 
-
 from fastapi import FastAPI, Response, HTTPException
 from fastapi.responses import FileResponse, JSONResponse
 from kokoro import KPipeline
@@ -146,7 +145,7 @@ def llm_chat_response(text, image_base64=None):
         
         logger.info("Initializing InferenceClient...")
         client = InferenceClient(
-            provider="sambanova",  # Specify provider if needed
+            provider="together",  # Updated to the provider shown in the sample
             api_key=HF_TOKEN
         )
         
@@ -160,7 +159,7 @@ def llm_chat_response(text, image_base64=None):
                     {"role": "system", "content": system_message},
                     {"role": "user", "content": [
                         {"type": "text", "text": text if text else "Describe what you see in the image in one line only"},
-                        {"type": "image", "source": {"data": f"data:image/jpeg;base64,{image_base64}"}}
+                        {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{image_base64}"}}
                     ]}
                 ]
             else:
@@ -179,35 +178,27 @@ def llm_chat_response(text, image_base64=None):
             
             logger.info(f"Received response from model")
             
-            # Handle potential different response formats
+            # Simplified response handling based on the sample code
             if not completion.choices or len(completion.choices) == 0:
                 logger.error("No choices returned from model.")
                 raise HTTPException(status_code=500, detail="Model returned no choices.")
             
-            # Extract the response message from the first choice
-            choice = completion.choices[0]
-            response_message = None
-            
-            if hasattr(choice, "message"):
-                response_message = choice.message
-            elif isinstance(choice, dict):
-                response_message = choice.get("message")
-            
-            if not response_message:
-                logger.error(f"Response message is empty: {choice}")
-                raise HTTPException(status_code=500, detail="Model response did not include a message.")
-            
-            content = None
-            if isinstance(response_message, dict):
-                content = response_message.get("content")
-            if content is None and hasattr(response_message, "content"):
-                content = response_message.content
-            
-            if not content:
-                logger.error(f"Message content is missing: {response_message}")
-                raise HTTPException(status_code=500, detail="Model message did not include content.")
-            
-            return content
+            # Extract the content directly using the expected format
+            try:
+                # Get message from first choice
+                message = completion.choices[0].message
+                
+                # Extract content from message
+                if hasattr(message, "content"):
+                    return message.content
+                elif isinstance(message, dict) and "content" in message:
+                    return message["content"]
+                else:
+                    logger.error(f"Unexpected message format: {message}")
+                    raise HTTPException(status_code=500, detail="Unexpected message format from model")
+            except Exception as e:
+                logger.error(f"Error extracting message content: {str(e)}")
+                raise HTTPException(status_code=500, detail=f"Failed to extract response content: {str(e)}")
             
         except Exception as e:
             logger.error(f"Error during model inference: {str(e)}")