getting closer

handler.py

@@ -20,21 +20,20 @@ class EndpointHandler():
     def __call__(self, data):
         try:
 
-            logger.info('data is %s', data)
+            inputs = data.pop("inputs", None)
             text_input = None
+            image_data = None
 
-            logger.info('data is a dict %s', data)
-            inputs = data.pop("inputs", None)
-            text_input = inputs["text"] if "text" in inputs else None
-            image_data = inputs['image'] if 'image' in inputs else None
+            if isinstance(inputs, Image):
+                logger.info('image sent directly')
+                image = inputs
+            else:
+                text_input = inputs["text"] if "text" in inputs else None
+                image_data = inputs['image'] if 'image' in inputs else None
 
             if image_data is not None:
-                if isinstance(image_data, Image):
-                    logger.info('image is an image')
-                    image = image_data
-                else:
-                    logger.info('image is encoded')
-                    image = BytesIO(base64.b64decode(image_data))
+                logger.info('image is encoded')
+                image = BytesIO(base64.b64decode(image_data))
 
             if text_input:
                 processor = self.processor(text=text_input, return_tensors="pt", padding=True).to(device)