MMR Search in Chroma not working, typo suspected

[jppaolim]

System Info

Langchain v0.0.171 Mac OS

Who can help?

@jeffchuber

Information

• [ ] The official example notebooks/scripts
• [X] My own modified scripts

Related Components

• [ ] LLMs/Chat Models
• [ ] Embedding Models
• [ ] Prompts / Prompt Templates / Prompt Selectors
• [ ] Output Parsers
• [ ] Document Loaders
• [X] Vector Stores / Retrievers
• [ ] Memory
• [ ] Agents / Agent Executors
• [ ] Tools / Toolkits
• [ ] Chains
• [ ] Callbacks/Tracing
• [ ] Async

Reproduction

If I initialise a chroma database and then retriever

db = Chroma.from_documents(texts, embeddings_function(), 
                    metadatas=[{"source": str(i)} for i in range(len(texts))], persist_directory=PERSIST_DIRECTORY)    

querybase = db.as_retriever(search_type="mmr", search_kwargs={"k":3, "lambda_mult":1})

retrieved files are then identical whether I pass 0.1 or 0.9 as lambda_mult parameter.

Expected behavior

I expect different file.

Digging into the code there is a typo I think in langchain.vectorstores.chroma last line should be lambda_mult and not lambda_mul :

As this is my first time, not sure how to properly suggest or test :)

def max_marginal_relevance_search(
        self,
        query: str,
        k: int = 4,
        fetch_k: int = 20,
        lambda_mult: float = 0.5,
        filter: Optional[Dict[str, str]] = None,
        **kwargs: Any,
    ) -> List[Document]:
        """Return docs selected using the maximal marginal relevance.
        Maximal marginal relevance optimizes for similarity to query AND diversity
        among selected documents.
        Args:
            query: Text to look up documents similar to.
            k: Number of Documents to return. Defaults to 4.
            fetch_k: Number of Documents to fetch to pass to MMR algorithm.
            lambda_mult: Number between 0 and 1 that determines the degree
                        of diversity among the results with 0 corresponding
                        to maximum diversity and 1 to minimum diversity.
                        Defaults to 0.5.
            filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None.
        Returns:
            List of Documents selected by maximal marginal relevance.
        """
        if self._embedding_function is None:
            raise ValueError(
                "For MMR search, you must specify an embedding function on" "creation."
            )

        embedding = self._embedding_function.embed_query(query)
        docs = self.max_marginal_relevance_search_by_vector(
            embedding, k, fetch_k, lambda_mul=lambda_mult, filter=filter
        )
        return docs

[jeffchuber]

@hwchase17 🤔 any thoughts here? I didn't write this but happy to help.

[Satyam-79]

Chroma.from_documents() takes Document object as a parameter not text and metadata separately. you have to use Chroma.from_texts() according to your use-case because your are providing texts and metadata separately.

[jppaolim]

In the end, I decided to use something else to get around ... but I still think the last line of this python code is suspect with a typo :)