create_retrieval_chain() — langchain Function Reference

Architecture documentation for the create_retrieval_chain() function in retrieval.py from the langchain codebase.

Function python AgentOrchestration ClassicChains

Entity Profile

AgentOrchestration→ ClassicChains→ create_retrieval_chain() — langchain Function Reference

Dependency Diagram

graph TD
  f4c87bdc_ef55_4e94_60c4_b5ba7464abce["create_retrieval_chain()"]
  a9b15678_260f_ff5a_60e8_7e75bfd1bbc7["retrieval.py"]
  f4c87bdc_ef55_4e94_60c4_b5ba7464abce -->|defined in| a9b15678_260f_ff5a_60e8_7e75bfd1bbc7
  style f4c87bdc_ef55_4e94_60c4_b5ba7464abce fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

libs/langchain/langchain_classic/chains/retrieval.py lines 12–68

def create_retrieval_chain(
    retriever: BaseRetriever | Runnable[dict, RetrieverOutput],
    combine_docs_chain: Runnable[dict[str, Any], str],
) -> Runnable:
    """Create retrieval chain that retrieves documents and then passes them on.

    Args:
        retriever: Retriever-like object that returns list of documents. Should
            either be a subclass of BaseRetriever or a Runnable that returns
            a list of documents. If a subclass of BaseRetriever, then it
            is expected that an `input` key be passed in - this is what
            is will be used to pass into the retriever. If this is NOT a
            subclass of BaseRetriever, then all the inputs will be passed
            into this runnable, meaning that runnable should take a dictionary
            as input.
        combine_docs_chain: Runnable that takes inputs and produces a string output.
            The inputs to this will be any original inputs to this chain, a new
            context key with the retrieved documents, and chat_history (if not present
            in the inputs) with a value of `[]` (to easily enable conversational
            retrieval.

    Returns:
        An LCEL Runnable. The Runnable return is a dictionary containing at the very
        least a `context` and `answer` key.

    Example:
        ```python
        # pip install -U langchain langchain-openai

        from langchain_openai import ChatOpenAI
        from langchain_classic.chains.combine_documents import (
            create_stuff_documents_chain,
        )
        from langchain_classic.chains import create_retrieval_chain
        from langchain_classic import hub

        retrieval_qa_chat_prompt = hub.pull("langchain-ai/retrieval-qa-chat")
        model = ChatOpenAI()
        retriever = ...
        combine_docs_chain = create_stuff_documents_chain(
            model, retrieval_qa_chat_prompt
        )
        retrieval_chain = create_retrieval_chain(retriever, combine_docs_chain)

        retrieval_chain.invoke({"input": "..."})
        ```
    """
    if not isinstance(retriever, BaseRetriever):
        retrieval_docs: Runnable[dict, RetrieverOutput] = retriever
    else:
        retrieval_docs = (lambda x: x["input"]) | retriever

    return (
        RunnablePassthrough.assign(
            context=retrieval_docs.with_config(run_name="retrieve_documents"),
        ).assign(answer=combine_docs_chain)
    ).with_config(run_name="retrieval_chain")