LangChain 实践之工具使用

作者：CSDN博客
参考：LangChain中文入门教程
LangChain官网
通过 Google 搜索并返回答案

1. import os
2. os.environ["OPENAI_API_KEY"]="xxx"
3. os.environ['SERPAPI_API_KEY']="xxx"from langchain.agents import load_tools
4. from langchain.agents import initialize_agent
5. from langchain.agents import AgentType
6. from langchain.llms import OpenAI
10. # First, let's load the language model we're going to use to control the agent.
11. llm = OpenAI(temperature=0)# chroma搜索# Next, let's load some tools to use. Note that the `llm-math` tool uses an LLM, so we need to pass that in.
12. tools = load_tools(["serpapi","llm-math"], llm=llm)# Finally, let's initialize an agent with the tools, the language model, and the type of agent we want to use.
13. agent = initialize_agent(tools, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True)# Now let's test it out!
14. agent.run("What was the  high temperature in SF yesterday in Fahrenheit? What 42 raised to the .023 power?")

复制代码

关于agent type 几个选项的含义:

zero-shot-react-description: 根据工具的描述和请求内容的来决定使用哪个工具（最常用）react-docstore: 使用 ReAct 框架和 docstore 交互, 使用Search 和Lookup 工具, 前者用来搜, 后者寻找term, 举例: Wipipedia 工具self-ask-with-search 此代理只使用一个工具: Intermediate Answer, 它会为问题寻找事实答案(指的非 gpt 生成的答案, 而是在网络中,文本中已存在的), 如 Google search API 工具conversational-react-description: 为会话设置而设计的代理, 它的prompt会被设计的具有会话性, 且还是会使用 ReAct 框架来决定使用来个工具, 并且将过往的会话交互存入内存

Gradio工具

stable fiffusion作图

1. from gradio_tools.tools import StableDiffusionTool
2. local_file_path = StableDiffusionTool().langchain.run("Please create a photo of a fox riding a skateboard")

复制代码

构建本地知识库问答机器人

1. from langchain.embeddings.openai import OpenAIEmbeddings
2. from langchain.vectorstores import Chroma
3. from langchain.text_splitter import CharacterTextSplitter
4. from langchain import OpenAI,VectorDBQA
5. from langchain.document_loaders import DirectoryLoader
6. from langchain.chains import RetrievalQA
8. # 加载文件夹中的所有txt类型的文件
9. loader = DirectoryLoader('../source_documents/', glob='*.txt')# 将数据转成 document 对象，每个文件会作为一个 document
10. documents = loader.load()# 初始化加载器
11. text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)# 切割加载的 document
12. split_docs = text_splitter.split_documents(documents)# 初始化 openai 的 embeddings 对象
13. embeddings = OpenAIEmbeddings()# 将 document 通过 openai 的 embeddings 对象计算 embedding 向量信息并临时存入 Chroma 向量数据库，用于后续匹配查询
14. docsearch = Chroma.from_documents(split_docs, embeddings)# 创建问答对象
15. qa = VectorDBQA.from_chain_type(llm=OpenAI(), chain_type="stuff", vectorstore=docsearch,return_source_documents=True)# 进行问答
16. query ="who the little prince meet first"
17. result = qa({"query": query})print(result)# 链式问答from langchain.chains.question_answering import load_qa_chain
18. docs = docsearch.similarity_search(query, include_metadata=True)
20. llm = OpenAI(temperature=0)
21. chain = load_qa_chain(llm, chain_type="stuff", verbose=True)
22. chain.run(input_documents=docs, question=query)

复制代码

视频问答

1. # 加载 youtube 频道
2. loader = YoutubeLoader.from_youtube_url('https://www.youtube.com/watch?v=9qq6HTr7Ocw')
3. loader = BiliBiliLoader(['https://www.bilibili.com/video/BV1xt411o7Xu/'])# loader = BiliBiliLoader(['https://www.bilibili.com/video/BV1Ch411j7Bb']) Return Empty transcript.# 加载blibili 频道# 将数据转成 document
4. documents = loader.load()# 初始化文本分割器
5. text_splitter = RecursiveCharacterTextSplitter(
6.   chunk_size=1000,
7.   chunk_overlap=20)# 分割 youtube documents
8. documents = text_splitter.split_documents(documents)# 初始化 openai embeddings
9. embeddings = OpenAIEmbeddings()# 将数据存入向量存储
10. vector_store = Chroma.from_documents(documents, embeddings)# 通过向量存储初始化检索器
11. retriever = vector_store.as_retriever()
13. system_template ="""
14. Use the following context to answer the user's question.
15. If you don't know the answer, say you don't, don't try to make it up. And answer in Chinese.
16. -----------
17. {context}
18. -----------
19. {chat_history}
20. """# 构建初始 messages 列表，这里可以理解为是 openai 传入的 messages 参数
21. messages =[
22.   SystemMessagePromptTemplate.from_template(system_template),
23.   HumanMessagePromptTemplate.from_template('{question}')]# 初始化 prompt 对象
24. prompt = ChatPromptTemplate.from_messages(messages)# 初始化问答链
25. qa = ConversationalRetrievalChain.from_llm(ChatOpenAI(temperature=0.1,max_tokens=2048),retriever,condense_question_prompt=prompt)
28. chat_history =[]whileTrue:
29.   question =input('问题：')# 开始发送问题 chat_history 为必须参数,用于存储对话历史
30.   result = qa({'question': question,'chat_history': chat_history})
31.   chat_history.append((question, result['answer']))print(result['answer'])

复制代码

原文地址：https://blog.csdn.net/weixin_38235865/article/details/130868767