langchain agent 使用外部工具示例

示例1

室外计算机器与天气搜索引擎，但在使用天气查询的时候好像不可以用了

提示词模版：LangSmith

import os
from dotenv import load_dotenv
from langchain_community.llms import Tongyi
load_dotenv('key.env')  # 指定加载 env 文件
key = os.getenv('DASHSCOPE_API_KEY')  # 获得指定环境变量
DASHSCOPE_API_KEY = os.environ["DASHSCOPE_API_KEY"]  # 获得指定环境变量
llm = Tongyi(temperature=1)


from langchain import hub
from langchain.agents import load_tools
from langchain.agents import AgentExecutor
from langchain.agents.output_parsers import ReActSingleInputOutputParser
from langchain.agents.format_scratchpad import format_log_to_str
from langchain.tools.render import render_text_description


llm_with_stop = llm.bind(stop=["\nObservation"])

# 准备我们的工具：这里用到 DuckDuckGo 搜索引擎，和一个基于 LLM 的计算器
tools = load_tools(["ddg-search", "llm-math"], llm=llm)

# 准备核心提示词：这里从 LangChain Hub 加载了 ReAct 模式的提示词，并填充工具的文本描述
prompt = hub.pull("hwchase17/react-json")
prompt = prompt.partial(
    tools=render_text_description(tools),
    tool_names=", ".join([t.name for t in tools]),
)

# 构建 Agent 的工作链：这里最重要的是把中间步骤的结构要保存到提示词的 agent_scratchpad 中
agent = (
    {
        "input": lambda x: x["input"],
        "agent_scratchpad": lambda x: format_log_to_str(x["intermediate_steps"]),
    }
    | prompt
    | llm_with_stop
    | ReActSingleInputOutputParser()
)

# 构建 Agent 执行器：执行器负责执行 Agent 工作链，直至得到最终答案（的标识）并输出回答
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
print(agent_executor.invoke({"input": "今天上海和北京的气温差几度？"}))

示例2

使用 playwright 工具访问网页，提取上面内容，大模型总结网页内容

pip install playwright

playwright install

提示词模版：LangSmith

import os
from dotenv import load_dotenv
from langchain_community.llms import Tongyi
from langchain import hub
from langchain.agents import AgentExecutor
from langchain.agents.format_scratchpad import format_log_to_str
from langchain.agents.output_parsers import JSONAgentOutputParser
from langchain_community.chat_models.openai import ChatOpenAI
from langchain_community.agent_toolkits import PlayWrightBrowserToolkit
from langchain_community.tools.playwright.utils import create_async_playwright_browser
from langchain.tools.render import render_text_description_and_args

load_dotenv('key.env')  # 指定加载 env 文件
key = os.getenv('DASHSCOPE_API_KEY')  # 获得指定环境变量
DASHSCOPE_API_KEY = os.environ["DASHSCOPE_API_KEY"]  # 获得指定环境变量
llm = Tongyi(temperature=1)

# 避免 Jupyter Notebook 产生 EventLoop 问题
import nest_asyncio
nest_asyncio.apply()

llm_with_stop = llm.bind(stop=["\nObservation"])

async def run_agent():
    async_browser = create_async_playwright_browser()
    browser_toolkit = PlayWrightBrowserToolkit.from_browser(async_browser=async_browser)
    tools = browser_toolkit.get_tools()

    prompt = hub.pull("hwchase17/react-multi-input-json")
    prompt = prompt.partial(
        tools=render_text_description_and_args(tools),
        tool_names=", ".join([t.name for t in tools]),
    )

    agent = (
        {
            "input": lambda x: x["input"],
            "agent_scratchpad": lambda x: format_log_to_str(x["intermediate_steps"]),
        }
        | prompt
        | llm_with_stop
        | JSONAgentOutputParser()
    )
    agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

    result = await agent_executor.ainvoke({"input": "请访问这个网页并总结一下上面的内容：https://v.qq.com/"})
    print(result)
    return result

# 在主程序中调用异步函数并等待其完成
if __name__ == "__main__":
    import asyncio
    asyncio.run(run_agent())