从链中先前步骤访问LangChain LCEL变量

Question

我正在尝试跳上 LCEL 和 Langserve 列车，但我在理解访问管道字典中设置的变量所涉及的一些“魔法”时遇到了一些困难。

这些变量似乎可以从提示模板中解析。我想在自定义函数等中检索这些值，但我不清楚如何直接访问它们。采取以下人为的示例，其目的是返回源文档以及响应中的答案：

class ChatResponse(BaseModel):
    answer: str
    sources: List[Document]

store = FAISS.from_texts(
    ["harrison worked at kensho"], embedding=OpenAIEmbeddings()
)
retriever = store.as_retriever()
template = """Answer the question based only on the following context:
{context}

Question: {question}
"""
prompt = ChatPromptTemplate.from_template(template)
llm = ChatOpenAI()

def format_response(answer):
    sources = [] # TODO lookup source documents (key: 'context')
    return ChatResponse(answer=answer, sources=sources)

retrieval_chain = (
        {"context": retriever, "question": RunnablePassthrough()}
        | prompt
        | llm
        | StrOutputParser()
        | RunnableLambda(format_response)
)
app = FastAPI()
add_routes(app, retrieval_chain, path="/chat", input_type=str, output_type=ChatResponse)

在

format_response

中，我留下了

TODO

来查找源文档。我想从管道的

context

键检索源文档。我如何访问从链的第一步设置的这个密钥？

Answer 1

来自文档https://python.langchain.com/docs/use_cases/question_answering/sources/

from langchain_core.runnables import RunnableParallel

def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)

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)

然后这个：

rag_chain_with_source.invoke("where did harrison work ?")

返回：

{'context': [Document(page_content='harrison worked at kensho')],
 'question': 'where did harrison work ?',
 'answer': 'Harrison worked at Kensho.'}