Update app.py

app.py

@@ -91,15 +91,20 @@ def bot_streaming(message, history, max_new_tokens=2048):
         context = f"\nDocument context:\n{doc_state.current_doc_text}" if doc_state.current_doc_text else ""
         current_msg = f"{txt}{context}"
         messages.append({"role": "user", "content": [{"type": "text", "text": current_msg}, {"type": "image"}]})
-        
-        # Process with single image
+    else:
+        messages.append({"role": "user", "content": [{"type": "text", "text": txt}]})
+
+    # Apply chat template to messages
+    texts = processor.apply_chat_template(messages, add_generation_prompt=True)
+    
+    # Process inputs based on whether we have images
+    if doc_state.current_doc_images:
         inputs = processor(
             text=texts,
             images=doc_state.current_doc_images[0:1],  # Only use first image
             return_tensors="pt"
         ).to("cuda")
     else:
-        messages.append({"role": "user", "content": [{"type": "text", "text": txt}]})
         inputs = processor(text=texts, return_tensors="pt").to("cuda")
     
     streamer = TextIteratorStreamer(processor, skip_special_tokens=True, skip_prompt=True)