diff --git a/llmops/llm_tools/log_analysis/log_analysis_workflow.py b/llmops/llm_tools/log_analysis/log_analysis_workflow.py
index 5062336e163b34e3bfd24a67ee05c926288ef910..abab8ea34dc6d6a4cdd5f369c5704aec7f49bf8c 100644
--- a/llmops/llm_tools/log_analysis/log_analysis_workflow.py
+++ b/llmops/llm_tools/log_analysis/log_analysis_workflow.py
@@ -4,89 +4,25 @@
 # * Author: yzy_dev <yuzhiyuan@kylinos.cn> 
 # * Date: Thu Nov 28 15:00:41 2024 +0800 
 # 日志分析
-import os
 
-import Agently
 from Agently.Agent.Agent import Agent
 
-from llmops.config.config import init_config
-from llmops.llm_tools.log_analysis.agentfactory import agentfactory
-
-
-def logworkflow(user_input,agent: Agent):
-    main_workflow = Agently.Workflow()
-    log_agent = agent
-
-    # @main_workflow.chunk("user_input")
-    # def user_input(inputs, storage):
-    #     storage.set(
-    #         "user_input",
-    #         # input("[请输入你的问题]: ")
-    #         user_input
-    #     )
-
-    # 返回shell脚本
-    @main_workflow.chunk("assistant_reply")
-    def assistant_shell_reply(inputs, storage):
-        print("分析中.....")
-        # assistant_reason_reply = (
-        #     agent
-        #     .info("回答尽量精简")
-        #     .user_info("用户是服务器运维人员，操作系统是macOs")
-        #     .input({
-        #         "user_input": storage.get("user_input"),
-        #     })
-        #     .instruct([
-        #         "replay中输出语言为中文",
-        #         f"请根据：{user_input}找出日志的问题",
-        #         "用简洁的方式说明问题"
-        #         "操作系统是maoOs"
-        #     ])
-        #     .output({
-        #         "replay": ("str", "对{user_input}进行解释"),
-        #         "improvement": ([("str", "根据用户输入{input}输出相应脚本处理用户的要求"), "找出改进方案"])
-        #     })
-        #     .start_async()
-        # )
-        # storage.set("replay", assistant_reason_reply.get("replay"))
-        # print("*" * 50)
-        # print("一.大模型分析：")
-        # print(assistant_reason_reply.get("replay"))
-        # print("*" * 50)
-        # print("二.生成的脚本内容:")
-        # storage.set("shell", assistant_reason_reply.get("improvement"))
-        # print(assistant_reason_reply.get("improvement"))
-        # print("*" * 50)
-        # return assistant_reason_reply.get("improvement")
-        assistant_reason_reply = (
-            log_agent
-            .input(user_input)
-            .info("用中文对系统日志进行解释，结果用字典形式例如：{reason:'解释'}")
-            .instruct("目标语言：中文")
-            .output({
-                "info_list": [
-                    {
-                        "知识对象": ("str", "分析{input}问题时，需要了解相关知识的具体对象"),
-                        "关键知识点": ("str", "分析{input}问题时，需要了解的关键知识")
-                    }
-                ],
-                "analyse": ("str", "根据{info_list}出现的问题进行解释和分析"),
-                "reasons": [("str", "{info_list}根据找出报错的原因并按列表格式输出，添加编号")]
-            })
-            .start()
-        )
-        return assistant_reason_reply
-
-    (
-        main_workflow
-        # .connect_to("user_input")
-        .connect_to("assistant_reply")
-        .connect_to("END")
+def generate(agent, user_input):
+    # 假设 agent 的生成器返回逐步生成的数据
+    generator = (
+        agent
+        .input(user_input)
+        .info("用中文对系统日志进行解释，结果用字典形式例如：{reason:'解释'}")
+        .instruct("目标语言：中文")
+        .output({
+            "analyse": ("str", "根据{user_input}出现的问题进行分析")
+            # "reasons": [("str", "{user_input}根据找出报错的原因并按列表格式输出，添加编号")]
+        })
+        .get_instant_generator()
     )
-    # print(main_workflow.draw())
-    result = main_workflow.start()
-    return result
-
+    for chunk in generator:
+        # 生成的数据以 'data: ' 开头，按格式发送
+        # delta = chunk["delta"]
+        # analyse = chunk["analyse"]
+        yield f"data: {chunk}\n\n"
 
-# r = logworkflow("Can not connect to MySQL server. Too many connections -mysql 1040错误")
-# print(r['default'])
diff --git a/llmops/router/log_analysis_router.py b/llmops/router/log_analysis_router.py
index 1d29bce6a163a30e863237df397367bf35f97f98..54cab9859d5c75cbad525af4f425f8f9957d7107 100644
--- a/llmops/router/log_analysis_router.py
+++ b/llmops/router/log_analysis_router.py
@@ -3,15 +3,30 @@
 # * See LICENSE file for more details.
 # * Author: yzy_dev <yuzhiyuan@kylinos.cn> 
 # * Date: Thu Nov 28 15:00:41 2024 +0800 
-from flask import Blueprint, request, current_app
+from Agently.Agent.Agent import Agent
+from flask import Blueprint, request,current_app, Response
+
 from llmops.controller.log_analysis_controller import LogAnalysisController
+from llmops.llm_tools.log_analysis.log_analysis_workflow import generate
+from llmops.utils.response import fail
 
+aasf = "asdfsadf"
 log_analysis_router = Blueprint("log_analysis", __name__)
-logAnalysisController = LogAnalysisController()
 
 
-@log_analysis_router.route("/log_analysis", methods=["POST"])
-def log_analysis():
+@log_analysis_router.route("/log_analysis/stream", methods=["POST"])
+def log_analysis_stream():
+    # 从查询参数获取 user_input
     agent = current_app.config["AGENT"]
-    user_input = request.get_json("user_input")
-    return logAnalysisController.log_analysis(user_input, agent)
+    user_input = request.form.get('user_input')
+    print(user_input)
+    if not user_input:
+        return fail("error:user_input 参数是必需的")
+    # 返回流式响应
+    return Response(generate(agent, user_input),
+                    content_type="text/event-stream",
+                    status=200,
+                    mimetype="text/event-stream",
+                    headers={"Cache-Control": "no-cache", "Connection": "keep-alive"})
+
+