Question: How to get score of ranked document?

[BernhardSchlegel]

Hey all, I'm using RAGatouille as a retriever for langchain like so

RAG = RAGPretrainedModel.from_index(".ragatouille/colbert/indexes/clr-docs")
retriever = RAG.as_langchain_retriever(k=10)
retrieval = RunnableParallel({
    "context": retriever,
    "question": RunnablePassthrough(),
})
rag_chain = (
    retrieval
    | RunnableParallel({
        "answer": ANSWER_PROMPT | llm,
        "docs": itemgetter("context")
    })
)

I've traced down as_langchain_retriever but I could not find any helpful arguments.

How do I get the score for each retrieved doc so I can, i.e., calculate the auPRC score?

Thanks, Bernhard.

[BernhardSchlegel]

Answer: Using the ContextualCompressionRetriever