fix: Kimi模型伪工具调用格式过滤

修复Kimi-K2.5模型在第二轮调用时输出伪工具调用格式的问题：
- 添加系统提示告诉模型直接根据工具结果回答
- 过滤 <|tool_calls_section_begin|> 等内部格式标记
- 清理多余空行

版本: v3.0.1

main_v2.py

@@ -33,7 +33,7 @@ logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 # 创建应用
-app = FastAPI(title="AI对话系统 v2.0", version="2.0.0")
+app = FastAPI(title="AI对话系统 v3.0", version="3.0.1")
 
 # 静态文件和模板
 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
@@ -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 +533,17 @@ 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']
-            })
+        
+        # 添加提示：告诉模型直接根据工具结果回答，不要再调用工具
+        # 添加一个系统级别的提示
+        tool_hint = {
+            "role": "system",
+            "content": "请根据工具返回的结果直接回答用户的问题，不要再调用任何工具或搜索。如果结果不足以回答问题，请根据现有信息给出最好的回答，并说明信息的局限性。"
+        }
+        # 在工具结果之后添加提示
+        final_messages.append(tool_hint)
         
         # 调用LLM生成最终回复
         url = f"{api_base.rstrip('/')}/chat/completions"
@@ -557,19 +558,50 @@ 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)}")
+        # 打印消息内容（调试）
+        for i, msg in enumerate(final_messages):
+            role = msg.get('role')
+            content_preview = str(msg.get('content', ''))[:100] if msg.get('content') else 'None'
+            if role == 'tool':
+                logger.info(f"消息[{i}] role={role}, tool_call_id={msg.get('tool_call_id')}, content长度={len(msg.get('content',''))}")
+            elif role == 'assistant' and msg.get('tool_calls'):
+                logger.info(f"消息[{i}] role={role}, tool_calls={len(msg['tool_calls'])}")
+            else:
+                logger.info(f"消息[{i}] role={role}, content={content_preview}...")
         
         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()
                 content = data['choices'][0]['message']['content']
                 
+                # 过滤掉伪工具调用格式（某些模型如Kimi会输出这种内部格式）
+                # 模式：<|tool_calls_section_begin|>...<|tool_calls_section_end|>
+                import re
+                tool_pattern = r'<\|tool_calls_section_begin\|>.*?<\|tool_calls_section_end\|>'
+                content = re.sub(tool_pattern, '', content, flags=re.DOTALL)
+                
+                # 也过滤单个 tool_call 格式
+                tool_call_pattern = r'<\|tool_call_begin\|>.*?<\|tool_call_end\|>'
+                content = re.sub(tool_call_pattern, '', content, flags=re.DOTALL)
+                
+                # 清理可能残留的格式标记
+                content = content.replace('<|tool_calls_section_begin|>', '')
+                content = content.replace('<|tool_calls_section_end|>', '')
+                content = content.replace('<|tool_call_begin|>', '')
+                content = content.replace('<|tool_call_end|>', '')
+                content = content.replace('<|tool_call_argument_begin|>', '')
+                content = content.replace('<|tool_call_argument_end|>', '')
+                
+                # 清理多余空行
+                content = re.sub(r'\n{3,}', '\n\n', content).strip()
+                
                 return content, None
                 
         except Exception as e: