Update chat_handler.py

chat_handler.py

@@ -335,25 +335,19 @@ class ChatHandler:
                         }
                     ))
                     
-                    # Only add separate media display if:
-                    # 1. We extracted a valid URL 
-                    # 2. The result text is NOT just a URL (to avoid double rendering)
-                    # 3. The result text doesn't contain the exact URL
-                    if media_url:
-                        # Check if the result is essentially just the URL (with possible whitespace)
-                        cleaned_result = result_text.strip()
-                        is_just_url = (cleaned_result == media_url or 
-                                     cleaned_result.startswith('http') and len(cleaned_result.split()) == 1)
-                        
-                        # Only add separate media if the result is more than just the URL
-                        if not is_just_url and media_url not in result_text:
-                            logger.info(f"🎯 Adding separate media display for: {media_url}")
-                            chat_messages.append(ChatMessage(
-                                role="assistant", 
-                                content={"path": media_url}
-                            ))
+                    # Only add separate media display if the tool result does NOT contain 
+                    # any Gradio file data structures that would be auto-rendered
+                    if media_url and not self._contains_gradio_file_structure(result_text):
+                        logger.info(f"🎯 Adding separate media display for: {media_url}")
+                        chat_messages.append(ChatMessage(
+                            role="assistant", 
+                            content={"path": media_url}
+                        ))
+                    else:
+                        if media_url:
+                            logger.info(f"🚫 Skipping separate media - tool result contains Gradio file structure")
                         else:
-                            logger.info(f"🚫 Skipping separate media - result is just URL: {is_just_url}, contains URL: {media_url in result_text}")
+                            logger.info(f"🚫 No media URL extracted")
                         
                 else:
                     # Add error message for failed tool call
@@ -385,6 +379,62 @@ class ChatHandler:
         
         return chat_messages
     
+    def _contains_gradio_file_structure(self, text: str) -> bool:
+        """Check if the text contains ANY Gradio file data structures that would be auto-rendered"""
+        
+        # Check for key indicators of Gradio file structures
+        gradio_indicators = [
+            # Gradio FileData type indicators
+            "'_type': 'gradio.FileData'",
+            '"_type": "gradio.FileData"',
+            'gradio.FileData',
+            
+            # File structure patterns
+            "'path':",
+            '"path":',
+            "'url':",
+            '"url":',
+            "'orig_name':",
+            '"orig_name":',
+            "'mime_type':",
+            '"mime_type":',
+            'is_stream',
+            'meta_type',
+            
+            # Common file result patterns
+            "{'image':",
+            '{"image":',
+            "{'audio':",
+            '{"audio":',
+            "{'video':",
+            '{"video":',
+            "{'file':",
+            '{"file":',
+            
+            # List patterns that typically contain file objects
+            "[{'image'",
+            '[{"image"',
+            "[{'audio'",
+            '[{"audio"',
+            "[{'video'",
+            '[{"video"',
+            "[{'file'",
+            '[{"file"'
+        ]
+        
+        # If we find multiple indicators, it's likely a Gradio file structure
+        indicator_count = sum(1 for indicator in gradio_indicators if indicator in text)
+        
+        # Also check for simple URL patterns (for audio case)
+        is_simple_url = (text.strip().startswith('http') and 
+                        len(text.strip().split()) == 1 and 
+                        any(ext in text.lower() for ext in ['.wav', '.mp3', '.mp4', '.png', '.jpg', '.jpeg', '.gif', '.svg', '.webm', '.ogg']))
+        
+        result = indicator_count >= 2 or is_simple_url
+        logger.debug(f"📋 File structure check: {indicator_count} indicators, simple_url: {is_simple_url}, result: {result}")
+        
+        return result
+    
     def _get_native_system_prompt(self) -> str:
         """Get system prompt for Claude without MCP servers"""
         return f"""You are Claude Sonnet 4, a helpful AI assistant with native multimodal capabilities. You can have conversations, answer questions, help with various tasks, and provide information on a wide range of topics.