new

.gitignore

@@ -6,3 +6,6 @@
 # Ignore environment-specific files
 .venv
 config.yaml
+
+# Ignore Python cache files
+**/__pycache__/

README.md

@@ -43,6 +43,17 @@ print(objective.progress)  # 输出应该是两个关键成果进度的平均值
 ```
 
 # 工具
-使用selenium工具操作浏览器进行网络搜索和内容爬取。
+使用selenium工具操作浏览器进行网络搜索和内容爬取。以下是一个简单的示例：
 
+```python
+from selenium import webdriver
 
+driver = webdriver.Firefox()
+driver.get("http://www.python.org")
+assert "Python" in driver.title
+driver.quit()
+```
+# 安装和运行
+```commandline
+pip install -r requirements.txt
+```

agent/agent_base.py

@@ -1,5 +1,4 @@
-import openai
-import yaml
+from tools.llm_tools import LLMTools
 
 
 class AgentBase:
@@ -14,7 +13,6 @@ class AgentBase:
             None
         """
         self.role = role
-        
 
     def set_role(self, new_role):
         """
@@ -28,34 +26,5 @@ class AgentBase:
         """
         self.role = new_role
 
-    
-
     def call_gpt(self, input_text):
-        """
-        Make a call to the OpenAI chat API.
-
-        Args:
-            input_text (str): The input text for the chat API.
-
-        Returns:
-            str: The generated text from the chat API.
-        """
-        # Load API configuration from YAML file
-        with open("config.yaml", "r") as file:
-            config = yaml.safe_load(file).get("openai")
-
-        openai.api_key = config.get("api_key")
-        openai.base_url = config.get("api_url")
-
-        response = openai.chat.completions.create(
-            model="gpt-3.5-turbo",  # GPT-3.5 model
-            messages=[
-                {"role": "system", "content": self.role},
-                {"role": "user", "content": input_text}
-            ],
-            temperature=0.2
-        )
-
-        generated_text = response.choices[0].message.content
-
-        return generated_text
+        return LLMTools().run(param={"role": self.role, "content": input_text})

agent/manager_agent.py

@@ -3,28 +3,27 @@ from agent.agent_base import AgentBase
 from agent.plan_agent import PlanAgent
 from work_principle.okr_principle import OKR_Object
 import logging
-logging.basicConfig(level=logging.INFO)
+
 class ManagerAgent(AgentBase):
     def __init__(self):
         super().__init__("你是一名总经理，负责与客户对齐目标，与计划拆解者对齐关键结果")
         self.logger = logging.getLogger(__name__)
 
     # 与用户对齐目标，填充缺失的信息
-    def optimization_Object(self, okr_object:OKR_Object):
+    def optimization_Object(self, okr_object: OKR_Object):
         # todo 待加入 PDCA循环规则
         for i in okr_object.task.content:
             call_openai_res = {'isOk': 'no', 'content': ''}
-            while call_openai_res["isOk"]=="no" :
+            while call_openai_res["isOk"] == "no":
                 prompt = f"这是一个任务描述:'{okr_object.raw_user_task}'。你觉得这个任务描述具备{i['descriptions']}吗？如果具备返回格式如下：{{\"isOk\":\"yes\", \"content\":\"提炼出任务的{i['descriptions']}\"}}，如果不具备返回格式如下：{{\"isOk\":\"no\",\"content\":\"返回不具备的原因并给出完善建议\"}}"
-                self.logger.info(prompt)
                 call_openai_res = json.loads(self.call_gpt(prompt))
-                self.logger.info(call_openai_res)
-                if call_openai_res["isOk"]=="no":
-                    okr_object.raw_user_task  = okr_object.raw_user_task + f"，{i['target']}：" +input(f"【警告】{call_openai_res['content']}\n请您补充信息：")
+                if call_openai_res["isOk"] == "no":
+                    okr_object.raw_user_task = okr_object.raw_user_task + f"，{i['target']}：" + input(
+                        f"【警告】{call_openai_res['content']}\n请您补充信息：")
             i['content'] = call_openai_res["content"]
 
-    
+
     # 与计划拆解者对齐关键结果，填充缺失的信息
-    def assign_task_to_plan_agent(self, okr_object:OKR_Object):
+    def assign_task_to_plan_agent(self, okr_object: OKR_Object):
         okr_object.task.raw_user_task
         PlanAgent()

agent/plan_agent.py

@@ -1,15 +1,23 @@
+import logging
 from agent.agent_base import AgentBase
+from tools.tools_base import ToolsBase
 from work_principle.okr_principle import OKR_Object
 
 class PlanAgent(AgentBase):
     def __init__(self):
-        super().__init__("你是一名计划拆解者，负责对OKR中的O进行拆解并制定KR，向总经理汇报")
+        self.logger = logging.getLogger(__name__)
+        super().__init__('你是一名计划拆解者，利用工具将object拆解成key_result，请你返回json格式的内容，比如["使用工具a进行xx操作", "使用工具b进行xx操作"]')
 
+    # 根据object和已有的工具能力拆解成key result，key
+    def aligning(self, okr_object: OKR_Object):
+        raw = okr_object.raw_user_task
+        tools_list = [i().get_info() for i in self.get_subclasses(ToolsBase)]
+        r = self.call_gpt(f"object为'{raw}'， 工具列表为{tools_list}")
+        print(r)
+        self.logger.info(f"Alignment result: {r}")
 
-     # 与用户对齐目标，填充缺失的信息
-    def aligning(self, okr_object:OKR_Object):
-        pass
-
-
-
-    
+    def get_subclasses(self, cls):
+        subclasses = []
+        for subclass in cls.__subclasses__():
+            subclasses.append(subclass)
+        return subclasses

main.py

@@ -1,6 +1,10 @@
 from agent.manager_agent import ManagerAgent
+from agent.plan_agent import PlanAgent
 from work_principle.okr_principle import OKR_Object
+import logging
 
+# 设置日志
+logging.basicConfig(level=logging.INFO)
 
 class AutoMate:
     def __init__(self):
@@ -10,11 +14,14 @@ class AutoMate:
     def rule_define(self):
         # 与用户对齐任务
         while True:
-            o_kr = OKR_Object("对比一下copilot和curson谁更好用，比较提示词数量、安装易用性，给出不少于100字的文章")
+            o_kr = OKR_Object("因为想要增加编程效率，对比一下copilot和curson谁更好用，比较提示词数量、安装易用性，给出不少于100字的文章")
             ManagerAgent().optimization_Object(o_kr)
             r = input(f"最终对齐的任务是：{o_kr.raw_user_task}，一切都OK对吧？y/n")
             if r == "y":
                 break
+        
+        # 让计划拆解者拆解任务
+        PlanAgent().aligning(o_kr)
 
 
 

tools/llm_tools.py

@@ -1,7 +1,39 @@
-
+import openai
+import yaml
 from tools.tools_base import ToolsBase
 
 
 class LLMTools(ToolsBase):
-    def get_describe(self):
-        return "利用大模型进行回答"
+    def __init__(self):
+        self.name = "llm_tools"
+        self.description = "利用大模型进行回答，入参格式为字典：{role:大模型的角色, content:问题}"
+    
+    def run(self, param=None):
+        """
+        Make a call to the OpenAI chat API.
+
+        Args:
+            param: The input text for the chat API.
+
+        Returns:
+            str: The generated text from the chat API.
+        """
+        with open("config.yaml", "r") as file:
+            config = yaml.safe_load(file).get("openai")
+
+        openai.api_key = config.get("api_key")
+        openai.base_url = config.get("api_url")
+        messages = []
+        if "role" in param:
+            messages.append({"role": "system", "content": param["role"]})
+        messages.append({"role": "user", "content": param["content"]})
+
+        response = openai.chat.completions.create(
+            model="gpt-3.5-turbo",  # GPT-3.5 model
+            messages=messages,
+            temperature=0.2
+        )
+
+        generated_text = response.choices[0].message.content
+
+        return generated_text

tools/tools_base.py

@@ -1,8 +1,10 @@
 class ToolsBase:
-    def get_describe(self):
-        return ["内容总结", "感性信息"]
+    def __init__(self):
+        self.name = ""
+        self.description = ""
+        
+    def get_info(self):
+        return {"name": self.name, "description": self.description}
 
-
-    def run(self):
+    def run(self, param=None):
         pass
- 

tools/web_browser_tools.py

@@ -12,10 +12,11 @@ from selenium.webdriver.common.by import By
 from tools.tools_base import ToolsBase
 
 class WebBrowser(ToolsBase):
-    def get_describe(self):
-        return ["最新信息", "链接", "具体针对性信息", "专业信息"]
+    def __init__(self):
+        self.name = "web_browser"
+        self.description = "利用浏览器进行搜索，入参格式为字符串"
 
-    def run(self):
+    def run(self, param=None):
         # Load browser configuration from YAML file
         with open("config.yaml", "r") as file:
             config = yaml.safe_load(file)

work_principle/object_components.py

@@ -3,7 +3,7 @@ class ObjectComponents:
         self.content = [
             {"target": "任务主体", "content": "", "descriptions": "任务主体"},
             {"target": "执行动作", "content": "", "descriptions": "执行动作"},
-            {"target": "任务背景", "content": "", "descriptions": "任务背景或原因"},
+            {"target": "任务背景", "content": "", "descriptions": "任务背景，为了xx、因为xx"},
             # "how": {"content": "", "descriptions": "这个任务如何做", "example": ["通过线上直播发布会，合作伙伴渠道和社交媒体宣传", "通过增加人员、改善流程和引入新的服务软件"]},
             {"target": "任务完成程度", "content": "", "descriptions": "任务要做到什么程度，给出具体的几个要求"}
         ]