Update app.py

app.py

@@ -159,24 +159,28 @@ def chatbot_response(message, history):
     try:
         messages = [{"role": "system", "content": get_system_prompt()}]
         
-        # 添加历史(最近5轮)
+        # 添加历史(最近3轭) - 减少上下文长度
         if history:
-            for item in history[-5:]:
+            for item in history[-3:]:  # 从5轮改为3轮
                 if isinstance(item, (list, tuple)) and len(item) == 2:
                     messages.append({"role": "user", "content": item[0]})
-                    messages.append({"role": "assistant", "content": item[1]})
+                    # 截断过长的历史回复
+                    assistant_msg = item[1]
+                    if len(assistant_msg) > 1000:
+                        assistant_msg = assistant_msg[:1000] + "...[truncated]"
+                    messages.append({"role": "assistant", "content": assistant_msg})
         
         messages.append({"role": "user", "content": message})
         
         tool_calls_log = []
         
-        # LLM 调用循环(最多5轮工具调用)
-        for iteration in range(5):
+        # LLM 调用循环(最多3轮工具调用) - 减少迭代次数
+        for iteration in range(3):  # 从5轮改为3轮
             response = client.chat_completion(
                 messages=messages,
                 model="Qwen/Qwen2.5-72B-Instruct:novita",
                 tools=MCP_TOOLS,
-                max_tokens=2000,
+                max_tokens=1500,  # 从2000降到1500
                 temperature=0.5,
                 tool_choice="auto",
                 stream=False
@@ -194,16 +198,20 @@ def chatbot_response(message, history):
                     # 调用 MCP 工具
                     tool_result = call_mcp_tool(tool_name, tool_args)
                     
-                    # 限制返回结果大小，避免超长内容导致500错误
+                    # 大幅限制返回结果大小，避免超长内容导致500错误
                     result_str = json.dumps(tool_result, ensure_ascii=False)
-                    if len(result_str) > 4000:
-                        # 截断过长的结果
-                        tool_result_truncated = {"_truncated": True, "preview": result_str[:4000] + "..."}
+                    if len(result_str) > 2000:  # 从4000降到2000
+                        # 截断过长的结果，只保留关键信息
+                        if isinstance(tool_result, dict) and "text" in tool_result:
+                            # 如果是文本格式，截取前1500字符
+                            tool_result_truncated = {"text": tool_result["text"][:1500] + "...[truncated]", "_truncated": True}
+                        else:
+                            tool_result_truncated = {"_truncated": True, "preview": result_str[:1500] + "...[truncated]"}
                         result_for_llm = json.dumps(tool_result_truncated)
                     else:
                         result_for_llm = result_str
                     
-                    # 记录工具调用(包含结果)
+                    # 记录工具调用(包含完整结果用于UI显示)
                     tool_calls_log.append({"name": tool_name, "arguments": tool_args, "result": tool_result})
                     
                     messages.append({
@@ -250,7 +258,7 @@ def chatbot_response(message, history):
             messages=messages,
             model="Qwen/Qwen2.5-72B-Instruct:novita",
             tools=MCP_TOOLS,
-            max_tokens=2000,
+            max_tokens=1500,  # 从2000降到1500
             temperature=0.5,
             stream=True
         )