【AI大模型应用开发】【LangChain系列】实战案例4：再战RAG问答，提取在线网页数据，并返回生成答案的来源

loader = WebBaseLoader(
    web_paths=("https://lilianweng.github.io/posts/2023-06-23-agent/",),
    bs_kwargs=dict(
        parse_only=bs4.SoupStrainer(
            class_=("post-content", "post-title", "post-header")
        )
    ),
)
docs = loader.load()

class WebBaseLoader(BaseLoader):
    """Load HTML pages using `urllib` and parse them with `BeautifulSoup'."""
    def __init__(
        self,
        web_path: Union[str, Sequence[str]] = "",
        header_template: Optional[dict] = None,
        verify_ssl: bool = True,
        proxies: Optional[dict] = None,
        continue_on_failure: bool = False,
        autoset_encoding: bool = True,
        encoding: Optional[str] = None,
        web_paths: Sequence[str] = (),
        requests_per_second: int = 2,
        default_parser: str = "html.parser",
        requests_kwargs: Optional[Dict[str, Any]] = None,
        raise_for_status: bool = False,
        bs_get_text_kwargs: Optional[Dict[str, Any]] = None,
        bs_kwargs: Optional[Dict[str, Any]] = None,
        session: Any = None,
    ) -> None:
        """Initialize loader.
        Args:
            web_paths: Web paths to load from.
            requests_per_second: Max number of concurrent requests to make.
            default_parser: Default parser to use for BeautifulSoup.
            requests_kwargs: kwargs for requests
            raise_for_status: Raise an exception if http status code denotes an error.
            bs_get_text_kwargs: kwargs for beatifulsoup4 get_text
            bs_kwargs: kwargs for beatifulsoup4 web page parsing
        """

text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
splits = text_splitter.split_documents(docs)

vectorstore = Chroma.from_documents(documents=splits, embedding=OpenAIEmbeddings())

retriever = vectorstore.as_retriever()

prompt = hub.pull("rlm/rag-prompt")
llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0)
def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)
rag_chain = (
    {"context": retriever | format_docs, "question": RunnablePassthrough()}
    | prompt
    | llm
    | StrOutputParser()
)

result = rag_chain.invoke("What is Task Decomposition?")
print(result)

import bs4
from langchain import hub
from langchain_community.document_loaders import WebBaseLoader
from langchain_community.vectorstores import Chroma
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter

from langchain_core.runnables import RunnableParallel
rag_chain_from_docs = (
    RunnablePassthrough.assign(context=(lambda x: format_docs(x["context"])))
    | prompt
    | llm
    | StrOutputParser()
)
rag_chain_with_source = RunnableParallel(
    {"context": retriever, "question": RunnablePassthrough()}
).assign(answer=rag_chain_from_docs)

result = rag_chain_with_source.invoke("What is Task Decomposition")
print(result)

rag_chain_with_source = RunnableParallel(
    {"context": retriever, "question": RunnablePassthrough()}
).assign(answer=rag_chain_from_docs)