Langchain极简教程: 九、一个完整的RAG案例

系列文章

Langchain极简教程: 一、Hello Langchain
Langchain极简教程: 二、模型
Langchain极简教程: 三、数据连接
Langchain极简教程: 四、提示词
Langchain极简教程: 五、输出解析器
Langchain极简教程: 六、链
Langchain极简教程: 七、记忆组件
Langchain极简教程: 八、代理 (Agent)
Langchain极简教程: 九、一个完整的RAG案例

简介

这是该 LangChain 极简入门系列的最后一讲。我们将利用过去9讲学习的知识，来完成一个具备完整功能集的LLM应用。该应用基于 LangChain 框架，以某 PDF 文件的内容为知识库，提供给用户基于该文件内容的问答能力。

我们利用 LangChain 的QA chain，结合 Chroma 来实现PDF文档的语义化搜索。示例代码所引用的是AWS Serverless
Developer Guide，该PDF文档共84页。

本讲的完整代码请参考10_Example.jpynb

1. 安装必要的 Python 包

   pip install langchain-ollama==0.2.0
   pip install chromadb 
   pip install pymupdf 
   pip install tiktoken
   

2. 下载PDF文件AWS Serverless Developer Guide

   !wget https://docs.aws.amazon.com/pdfs/serverless/latest/devguide/serverless-core.pdf
   
   PDF_NAME = 'serverless-core.pdf'
   

3. 加载PDF文件

   from langchain_community.document_loaders import PyMuPDFLoader
   
   pdf_path = './serverless-core.pdf'
   docs = PyMuPDFLoader(pdf_path).load()
   
   print (f'There are {len(docs)} document(s) in {pdf_path}.')
   print (f'There are {len(docs[0].page_content)} characters in the first page of your document.')
   

4. 拆分文档并存储文本嵌入的向量数据

   from langchain_huggingface.embeddings import HuggingFaceEmbeddings
   from langchain_text_splitters import RecursiveCharacterTextSplitter
   from langchain_chroma.vectorstores import Chroma
   
   text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
   split_docs = text_splitter.split_documents(docs)
   # 为了提升检索的准确性，这里的 text2vec-base-chinese embedding模型可以替换成 stella_en_v5或者 nomic-ai/nomic-embed-text-v1 等英文模型
   embeddings = HuggingFaceEmbeddings(
       model_name="/shibing624/text2vec-base-chinese",
       model_kwargs={'device': 'cpu'},
       encode_kwargs={'normalize_embeddings': False}
   )
   
   vectorstore = Chroma.from_documents(split_docs, embeddings, collection_name="serverless_guide")
   

5. 基于OpenAI创建QA链

   from langchain_core.prompts import (
       ChatPromptTemplate,
       PromptTemplate,
       SystemMessagePromptTemplate,
       AIMessagePromptTemplate,
       HumanMessagePromptTemplate,
   )
   from langchain_core.runnables import RunnablePassthrough
   from langchain_core.output_parsers import StrOutputParser
   from langchain_openai import ChatOpenAI
   
   template = """You are an assistant for question-answering tasks. Use the following pieces of retrieved context to answer the question. If you don't know the answer, just say that you don't know. Use three sentences maximum and keep the answer concise.
   Context: {context} 
   Question: {question} 
   """
   
   retriever = vectorstore.as_retriever()
   
   def format_docs(docs):
       return "\n\n".join(doc.page_content for doc in docs)
   
   ## 利用 ChatPromptTemplate.from_messages
   # human_message_prompt = HumanMessagePromptTemplate(
   #     prompt=PromptTemplate(
   #         template=template,
   #         input_variables=["question", "context"],
   #     )
   # )
   
   # prompt = ChatPromptTemplate.from_messages([human_message_prompt])
   
   ## 利用 ChatPromptTemplate.from_template
   prompt = ChatPromptTemplate.from_template(template)
   
   llm = ChatOpenAI(temperature=0, model_name="qwen2.5:14b", openai_api_key='ollama', openai_api_base='http://localhost:11434/v1')
   
   rag_chain = (
       {"context": retriever | format_docs, "question": RunnablePassthrough()}
       | prompt
       | llm
       | StrOutputParser()
   )
   

6. 基于相关文档，利用QA链完成回答

   rag_chain.invoke("What is the use case of AWS Serverless?")

你可以看到如下的输出内容：

' \n\nThe AWS Serverless Application Model (SAM) CLI is used for creating, deploying, testing, and updating serverless applications from the command line. It allows developers to work more efficiently with serverless architectures by providing features to manage infrastructure and code locally before deploying them to the cloud.'

总结

本节课程中，我们利用所学的知识，完成了第一个完整的LLM应用。希望通过本系列的学习，大家能对 LangChain 框架的使用，有了基本的认识，并且掌握了框架核心组建的使用方法。

相关文档资料链接：

1. Python Langchain官方文档
2. Models - Langchain