Update chat_handler.py

chat_handler.py

@@ -172,45 +172,42 @@ class ChatHandler:
     def _prepare_hf_messages(self, history: List, uploaded_file_urls: List[str] = None) -> List[Dict[str, Any]]:
         """Convert history (ChatMessage or dict) to HF OpenAI-compatible format with multimodal support"""
         messages: List[Dict[str, Any]] = []
-
+        
         # Get optimal context settings for current model/provider
         if self.mcp_client.current_model and self.mcp_client.current_provider:
             context_settings = AppConfig.get_optimal_context_settings(
-                self.mcp_client.current_model,
+                self.mcp_client.current_model, 
                 self.mcp_client.current_provider,
                 len(self.mcp_client.get_enabled_servers())
             )
-            max_history = context_settings["recommended_history_limit"]
+            max_history = context_settings['recommended_history_limit']
         else:
             max_history = 20  # Fallback
-
-        # Limit history
+        
+        # Convert history to HF API format (text only for context)
         recent_history = history[-max_history:] if len(history) > max_history else history
-
-        vision_supported = AppConfig.supports_vision_model(self.mcp_client.current_model) if self.mcp_client.current_model else False
-        last_role: Optional[str] = None
-
+        
+        last_role = None
         for msg in recent_history:
-            # Normalize message
-            if hasattr(msg, "role"):
+            # Handle both ChatMessage objects and dictionary format for backward compatibility
+            if hasattr(msg, 'role'):  # ChatMessage object
                 role = msg.role
                 content = msg.content
-            elif isinstance(msg, dict) and "role" in msg:
-                role = msg.get("role")
-                content = msg.get("content")
+            elif isinstance(msg, dict) and 'role' in msg:  # Dictionary format
+                role = msg.get('role')
+                content = msg.get('content')
             else:
-                continue
-
+                continue  # Skip invalid messages
+                
             if role == "user":
-                part: Optional[Dict[str, Any]] = None
+                # Build multimodal user messages with parts
+                part = None
                 if isinstance(content, dict) and "path" in content:
                     file_path = content.get("path", "")
                     if isinstance(file_path, str) and file_path.startswith("http") and AppConfig.is_image_file(file_path):
-                        if vision_supported:
-                            part = {"type": "image_url", "image_url": {"url": file_path}}
-                        else:
-                            part = None  # skip images for text-only models
+                        part = {"type": "image_url", "image_url": {"url": file_path}}
                     else:
+                        # Non-image or non-URL: fallback to text description
                         part = {"type": "text", "text": f"[File: {file_path}]"}
                 elif isinstance(content, (list, tuple)):
                     part = {"type": "text", "text": f"[List: {str(content)[:50]}...]"}
@@ -219,17 +216,18 @@ class ChatHandler:
                 else:
                     part = {"type": "text", "text": str(content)}
 
-                if part is not None:
-                    if messages and last_role == "user" and isinstance(messages[-1].get("content"), list):
-                        messages[-1]["content"].append(part)
-                    elif messages and last_role == "user" and isinstance(messages[-1].get("content"), str):
-                        existing_text = messages[-1]["content"]
-                        messages[-1]["content"] = [{"type": "text", "text": existing_text}, part]
-                    else:
-                        messages.append({"role": "user", "content": [part]})
-                    last_role = "user"
+                if messages and last_role == "user" and isinstance(messages[-1].get("content"), list):
+                    messages[-1]["content"].append(part)
+                elif messages and last_role == "user" and isinstance(messages[-1].get("content"), str):
+                    # Convert existing string content to parts and append
+                    existing_text = messages[-1]["content"]
+                    messages[-1]["content"] = [{"type": "text", "text": existing_text}, part]
+                else:
+                    messages.append({"role": "user", "content": [part]})
+                last_role = "user"
 
             elif role == "assistant":
+                # Assistant content remains text for chat.completions API
                 if isinstance(content, dict):
                     text = f"[Object: {str(content)[:50]}...]"
                 elif isinstance(content, (list, tuple)):
@@ -240,7 +238,7 @@ class ChatHandler:
                     text = str(content)
                 messages.append({"role": "assistant", "content": text})
                 last_role = "assistant"
-
+        
         return messages
     
     def _call_hf_api(self, messages: List[Dict[str, Any]], uploaded_file_urls: List[str] = None) -> List[ChatMessage]:
@@ -646,4 +644,4 @@ IMPORTANT NOTES:
 - ALWAYS provide a descriptive message before the JSON tool call
 - After tool execution, you can provide additional context or ask if the user needs anything else
 Current time: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}
-Current model: {self.mcp_client.current_model} via {self.mcp_client.current_provider}"""
+Current model: {self.mcp_client.current_model} via {self.mcp_client.current_provider}"""