Update app.py

app.py

@@ -97,6 +97,8 @@
 # app.py
 import os
 import logging
+import base64
+import requests
 from typing import Optional
 from fastapi import FastAPI, HTTPException
 from fastapi.responses import JSONResponse
@@ -137,16 +139,34 @@ def llm_chat_response(text: str, image_url: Optional[str] = None) -> str:
         )
         
         # Build the messages payload dynamically.
-        # If only text is provided, we append a default instruction.
+        # For text-only queries, append a default instruction.
         message_content = [{
             "type": "text",
             "text": text + ("" if image_url else " describe in one line only")
         }]
         
         if image_url:
+            logger.info("Downloading and converting image to base64 data URI...")
+            # Download the image from the URL
+            image_response = requests.get(image_url)
+            if image_response.status_code != 200:
+                logger.error("Failed to download image from URL")
+                raise HTTPException(status_code=500, detail="Failed to download image from provided URL")
+            image_bytes = image_response.content
+            
+            # Determine the MIME type
+            mime_type = image_response.headers.get("Content-Type")
+            if not mime_type or not mime_type.startswith("image/"):
+                logger.error("Invalid image MIME type")
+                raise HTTPException(status_code=500, detail="Invalid image MIME type")
+            
+            # Encode image in base64 and format as a data URI
+            base64_image = base64.b64encode(image_bytes).decode("utf-8")
+            data_uri = f"data:{mime_type};base64,{base64_image}"
+            
             message_content.append({
                 "type": "image_url",
-                "image_url": {"url": image_url}
+                "image_url": {"url": data_uri}
             })
         
         messages = [{
@@ -161,18 +181,15 @@ def llm_chat_response(text: str, image_url: Optional[str] = None) -> str:
             max_tokens=500
         )
         
-        # Debug log the raw response for troubleshooting.
         logger.info(f"Raw model response: {completion}")
         
-        # Ensure we have a valid response.
         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 message; use dict get to avoid NoneType errors.
-        response_message = None
-        # Some responses may be dicts or objects; try both approaches.
+        # Extract the message content from the first choice.
         choice = completion.choices[0]
+        response_message = None
         if hasattr(choice, "message"):
             response_message = choice.message
         elif isinstance(choice, dict):
@@ -182,11 +199,9 @@ def llm_chat_response(text: str, image_url: Optional[str] = None) -> str:
             logger.error(f"Response message is empty: {choice}")
             raise HTTPException(status_code=500, detail="Model response did not include a message.")
         
-        # Extract the content from the message.
         content = None
         if isinstance(response_message, dict):
             content = response_message.get("content")
-        # If for some reason it's not a dict, try attribute access.
         if content is None and hasattr(response_message, "content"):
             content = response_message.content
         
@@ -233,3 +248,4 @@ async def method_not_allowed_handler(request, exc):
         content={"error": "Method not allowed. Please check the API documentation."}
     )
 
+