📝 更新(word_action.py, woker_agent.py, chat_page.py, llm_util.py)：更正导入问题，调整工作代理逻辑。

actions/word_action.py

@@ -1,4 +1,4 @@
-from action_util import action, ActionBase
+from actions.action_base import action, ActionBase
 class WordAction(ActionBase):
     actions_description = "word基本操作"
     def __init__(self):

agent/woker_agent.py

@@ -1,55 +1,21 @@
-from langchain import hub
-from langchain.agents import create_react_agent, AgentExecutor
-from langchain_core.prompts import PromptTemplate, ChatMessagePromptTemplate
 from actions.action_util import ActionUtil
-from system_prompt import system_prompt
-from tools.tools_util import ToolsUtil
+from agent.system_prompt import system_prompt
 from utils.llm_util import LLM_Util
 
 
 class WorkerAgent:
-    def get_executor(self):
-        llm = LLM_Util().llm()
-        tools = ToolsUtil.get_tools()
-        prompt = hub.pull("langchain-ai/react-agent-template")
-        prompt.partial(instructions="")
-        agent = create_react_agent(llm=llm, tools=tools, prompt=prompt)
-        agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True, handle_parsing_errors=True)
-        return agent_executor
-
-    def run(self, question):
-        return self.get_executor().invoke({"input": question})["output"]
-
-    @staticmethod
-    def get_iter(question):
-        llm = LLM_Util().llm()
-        tools = ToolsUtil.get_tools()
-        example = """Thought: 我需要使用工具吗? 需要\nAction: 桌面路径\nAction Input: ""\nObservation: c:/path/develop\n\n
-Thought: 我需要使用工具吗? 需要\nAction: 打开应用\nAction Input: ""\nObservation: 打开成功\n\n
-Thought: 我需要使用工具吗? 不需要\nFinal Answer: 您的桌面上有以下文件\n
-"""
-        prompt = PromptTemplate(
-            template="### TOOLS ###\n{tools}#######\n"
-                     "### THOUGHT ###Thought:我需要使用工具吗? 需要\n"
-                     "Action:{tool_names}\n"
-                     "Action Input:Action的输入，如果没有参数请设置为""\n"
-                     "Observation:运行[ACTION]得到的结果\n\n"
-                     "当输出内容或者不需要使用工具，必须使用以下格式: Thought: 我需要使用工具吗? 不需要\nFinal Answer: [你的回复]######\n"
-                     "### EXAMPLE ###\n{example}######\n"
-                     "### PREVIOUS CONVERSATION HISTORY ###\n{chat_history}######\n"
-                     "### NEW INPUT ###\n{input}######\n{agent_scratchpad}",
-            input_variables=['agent_scratchpad', 'input', 'tool_names', 'tools'],
-            partial_variables={'chat_history': '', 'instructions': ""}
-        )
-        prompt = prompt.partial(example=example)
-        agent = create_react_agent(llm=llm, tools=tools, prompt=prompt)
-        agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True, handle_parsing_errors=True)
-        return agent_executor.iter({"input": question})
-
-    def run(self, question):
+    def __init__(self):
         action_descriptions = ""
-        for action in ActionUtil.get_actions():
+        for action_class in ActionUtil.get_actions():
+            action = action_class()
             action_descriptions += action.package_actions_description() + "\n"
-        messages = [{"content": system_prompt.substitute(python_code=action_descriptions), "role": "system"}]
-        messages.append({"content": question, "role": "user"})
-        LLM_Util().invoke(messages)
+        self.messages = [{"content": system_prompt.substitute(python_code=action_descriptions), "role": "system"}]
+
+
+    def run(self, question):
+        self.messages.append({"content": question, "role": "user"})
+        res = LLM_Util().invoke(self.messages)
+        self.messages.append({"content": res, "role": "assistant"})
+        return res
+
+

pages/chat_page.py

@@ -101,16 +101,8 @@ class WorkerThread(QThread):
 
     def run(self):
         try:
-            agent_iter = WorkerAgent().get_iter(self.text)
-            for step in agent_iter:
-                content = ""
-                if output := step.get("intermediate_step"):
-                    action, value = output[0]
-                    content = f"{action.tool} \n{value}"
-                elif step.get("output"):
-                    content = step["output"]
-                content = content.replace("```", "")
-                self.finished_signal.emit(content)
+            content = WorkerAgent().run(self.text)
+            self.finished_signal.emit(content)
         except Exception as e:
             traceback.print_exc(e)
 

utils/llm_util.py

@@ -20,6 +20,6 @@ class LLM_Util:
         if self.base_url == "":
             response = completion(model=self.model, api_key=self.api_key, messages=messages)
         else:
-            response = completion(model=self.model, base_url=self.base_url, api_key=self.key, messages=messages)
+            response = completion(model=self.model, base_url=self.base_url, api_key=self.api_key, messages=messages)
 
         return response.choices[0].message.content