fix: chat_with_tool_results不重复添加tool结果，修正消息格式

main_v2.py

@@ -1121,7 +1121,6 @@ async def websocket_endpoint(websocket: WebSocket, user_id: str):
                                         messages=history_with_tools,
                                         provider_config=agent_config['provider'],
                                         agent_config=agent_config['agent'],
-                                        tool_results=tool_results,
                                         enable_thinking=enable_thinking
                                     )
                         

services/agent_service.py

@@ -137,6 +137,9 @@ class AgentService:
                 'api_key': provider.api_key if provider else None,
                 'supports_thinking': provider.supports_thinking if provider else False,
                 'thinking_model': provider.thinking_model if provider else None,
+                'supports_vision': provider.supports_vision if provider else False,
+                'vision_model': provider.vision_model if provider else None,
+                'supports_function_calling': provider.supports_function_calling if provider else False,
                 'default_model': provider.default_model if provider else 'auto',
                 'max_tokens': provider.max_tokens if provider else 4096,
                 'temperature': provider.temperature if provider else 0.7,

services/llm_service.py

@@ -514,17 +514,15 @@ class LLMService:
         messages: List[Dict],
         provider_config: dict,
         agent_config: dict,
-        tool_results: List[Dict],
         enable_thinking: bool = True
     ) -> Tuple[str, Optional[str]]:
         """
-        第二阶段调用：将工具执行结果返回给LLM
+        第二阶段调用：使用包含工具调用和结果的完整消息历史
         
         Args:
-            messages: 对话历史（包含工具调用和结果）
+            messages: 已包含assistant tool_calls和tool结果的完整消息历史
             provider_config: LLM Provider配置
             agent_config: Agent配置
-            tool_results: 工具执行结果 [{"tool_call_id": "xxx", "content": "..."}]
         
         Returns:
             Tuple[str, Optional[str]]: (回复内容, 思考过程)
@@ -535,14 +533,8 @@ class LLMService:
         max_tokens = provider_config.get('max_tokens', 4096)
         temperature = agent_config.get('temperature_override') or provider_config.get('temperature', 0.7)
         
-        # 将工具结果添加到消息历史
+        # 消息历史已经包含了assistant的tool_calls和tool结果，直接使用
         final_messages = messages.copy()
-        for result in tool_results:
-            final_messages.append({
-                "role": "tool",
-                "tool_call_id": result['tool_call_id'],
-                "content": result['content']
-            })
         
         # 调用LLM生成最终回复
         url = f"{api_base.rstrip('/')}/chat/completions"
@@ -557,14 +549,14 @@ class LLMService:
             "max_tokens": max_tokens
         }
         
-        logger.info(f"工具结果返回LLM: url={url}, model={model}")
+        logger.info(f"工具结果返回LLM: url={url}, model={model}, 消息数={len(final_messages)}")
         
         try:
             async with httpx.AsyncClient(timeout=60.0) as client:
                 response = await client.post(url, headers=headers, json=payload)
                 
                 if response.status_code != 200:
-                    logger.error(f"API返回错误: status={response.status_code}")
+                    logger.error(f"API返回错误: status={response.status_code}, body={response.text[:500]}")
                     response.raise_for_status()
                 
                 data = response.json()