Update app.py

app.py

@@ -40,6 +40,45 @@ def decode_base64_image(base64_str: str) -> Optional[Image.Image]:
         return None
 
 
+def save_images_to_disk(images: List[Image.Image], output_dir: str = "temp_images") -> List[str]:
+    """
+    Save PIL Image objects to disk and return their file paths.
+    
+    Args:
+        images (List[Image.Image]): List of PIL Image objects to save
+        output_dir (str): Directory where images will be saved (default: "temp_images")
+    
+    Returns:
+        List[str]: List of file paths where images were saved
+    
+    Raises:
+        Exception: If there's an error saving images to disk
+    """
+    try:
+        # Create output directory if it doesn't exist
+        os.makedirs(output_dir, exist_ok=True)
+        
+        saved_paths = []
+        for i, image in enumerate(images):
+            if image is None:
+                log.warning(f"Skipping None image at index {i}")
+                continue
+                
+            # Generate unique filename
+            filename = f"image_{uuid.uuid4().hex}.png"
+            file_path = os.path.join(output_dir, filename)
+            
+            # Save image to disk
+            image.save(file_path, "PNG")
+            saved_paths.append(file_path)
+            log.info(f"Saved image to: {file_path}")
+        
+        return saved_paths
+        
+    except Exception as e:
+        log.error(f"Error saving images to disk: {str(e)}")
+        raise
+
 app = FastAPI(title="Kimi Service", version="1.5.0")
 inference = Inference()
 router = APIRouter()
@@ -61,7 +100,13 @@ async def classify(request: ClassificationRequest):
             images.append(img)
         log.info(f"Decoded {len(images)} images successfully")
 
-        res = inference.classify_building(images)
+        # Save images and get their paths using a helper method
+        output_dir = os.environ.get("IMAGE_OUTPUT_DIR", "/tmp/temp_images")
+        saved_image_paths = save_images_to_disk(images, output_dir)
+
+        # Send images to inference
+        res = inference.classify_building(images, saved_image_paths)
+
         if res is None:
             raise HTTPException(status_code=500, detail="Classification failed")
         return res