我使用langchain.document_loaders加载pdf文件:
loader = DirectoryLoader( './files/', glob='*.pdf', loader_cls=PyPDFLoader)
然后分割文档,创建嵌入,存储并加载它们:
docsearch = Chroma.from_documents(texts, embeddings, persist_directory=persist_directory)...docsearch = Chroma(persist_directory, embedding_function=embeddings ) retriever = docsearch.as_retriever( search_kwargs={"k": 5})docs = retriever.get_relevant_documents( query )len( docs)
我得到了正确的响应,但源文档数量为0。
回答:
错误出现在:
docsearch = Chroma(persist_directory, embedding_function=embeddings )
应该改为:
docsearch = Chroma(persist_directory=persist_directory, embedding_function=embeddings )