Update vlm.py

vlm.py

@@ -49,39 +49,21 @@ def encode_image(image_path):
 #
 # Build messages
 #
-def build_messages(message: dict, history: list[tuple]):
+def build_messages(message: dict, history: list[dict]):
     """Build messages given message & history from a **multimodal** chat interface.
     Args:
         message: dictionary with keys: 'text', 'files'
-        history: list of tuples with (message, response)
+        history: list of dictionaries
     
     Returns:
         list of messages (to be sent to the model)
     """
     logger.info(f"{message=}")
     logger.info(f"{history=}")
+    
     # Get the user's text and list of images
     user_text = message.get("text", "")
     user_images = message.get("files", [])  # List of images
-
-    # Build the message list including history
-    messages = []
-    combined_user_input = [] # Combine images and text if found in same turn.
-    for user_turn, bot_turn in history:
-        if isinstance(user_turn, tuple):  # Image input
-            image_content = [
-                {
-                    "type": "image_url",
-                    "image_url": f"data:image/jpeg;base64,{encode_image(image)}"
-                } for image in user_turn
-            ]
-            combined_user_input.extend(image_content)
-        elif isinstance(user_turn, str): # Text input
-            combined_user_input.append({"type": "text", "text": user_turn})
-        if combined_user_input and bot_turn:
-            messages.append({'role': 'user', 'content': combined_user_input})
-            messages.append({'role': 'assistant', 'content': [{"type": "text", "text": bot_turn}]})
-            combined_user_input = [] #reset the combined user input.
     
     # Build the user message's content from the provided message
     user_content = []
@@ -94,7 +76,9 @@ def build_messages(message: dict, history: list[tuple]):
                 "image_url": f"data:image/jpeg;base64,{encode_image(image)}"
             }
         )
-    
+
+    # Append to the history to create the new messages
+    messages = history
     messages.append({'role': 'user', 'content': user_content})
     logger.info(f"{messages=}")