better handling of images

app.py

@@ -10,7 +10,9 @@ from agent import run_agent
 MODEL = "Qwen/Qwen3-VL-30B-A3B-Thinking"
 
 DEFAULT_SYSTEM = (
-    "You are a helpful, multimodal AI assistant. When an image is sent, simply transcribe it without analysis, unless the user asks for analysis. When you need up-to-date information from the internet, use the search_web tool. Once you have a complete answer, call final_output with the full answer. If a task is impossible or unsafe, call abort with a brief reason."
+    "You are a helpful, multimodal AI assistant. When an image is sent, simply transcribe it without analysis, unless the user asks for analysis. 
+    "When you need up-to-date information from the internet, use the search_web tool. Once you have a complete answer, call final_output with the full answer. 
+    "If a task is impossible or unsafe, call abort with a brief reason."
 )