maximal_marginal_relevance() — langchain Function Reference

Architecture documentation for the maximal_marginal_relevance() function in utils.py from the langchain codebase.

Function python DataProcessing DocumentLoaders calls 1

Entity Profile

DataProcessing→ DocumentLoaders→ maximal_marginal_relevance() — langchain Function Reference

Dependency Diagram

graph TD
  0c88bb05_2797_2c68_7ac3_726357557644["maximal_marginal_relevance()"]
  530fd015_66ee_ef3b_a35b_3710e1b1764c["utils.py"]
  0c88bb05_2797_2c68_7ac3_726357557644 -->|defined in| 530fd015_66ee_ef3b_a35b_3710e1b1764c
  4c281b40_1396_20c6_d4c7_0be61771cba1["_cosine_similarity()"]
  0c88bb05_2797_2c68_7ac3_726357557644 -->|calls| 4c281b40_1396_20c6_d4c7_0be61771cba1
  style 0c88bb05_2797_2c68_7ac3_726357557644 fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

libs/core/langchain_core/vectorstores/utils.py lines 106–157

def maximal_marginal_relevance(
    query_embedding: np.ndarray,
    embedding_list: list,
    lambda_mult: float = 0.5,
    k: int = 4,
) -> list[int]:
    """Calculate maximal marginal relevance.

    Args:
        query_embedding: The query embedding.
        embedding_list: A list of embeddings.
        lambda_mult: The lambda parameter for MMR.
        k: The number of embeddings to return.

    Returns:
        A list of indices of the embeddings to return.

    Raises:
        ImportError: If numpy is not installed.
    """
    if not _HAS_NUMPY:
        msg = (
            "maximal_marginal_relevance requires numpy to be installed. "
            "Please install numpy with `pip install numpy`."
        )
        raise ImportError(msg)

    if min(k, len(embedding_list)) <= 0:
        return []
    if query_embedding.ndim == 1:
        query_embedding = np.expand_dims(query_embedding, axis=0)
    similarity_to_query = _cosine_similarity(query_embedding, embedding_list)[0]
    most_similar = int(np.argmax(similarity_to_query))
    idxs = [most_similar]
    selected = np.array([embedding_list[most_similar]])
    while len(idxs) < min(k, len(embedding_list)):
        best_score = -np.inf
        idx_to_add = -1
        similarity_to_selected = _cosine_similarity(embedding_list, selected)
        for i, query_score in enumerate(similarity_to_query):
            if i in idxs:
                continue
            redundant_score = max(similarity_to_selected[i])
            equation_score = (
                lambda_mult * query_score - (1 - lambda_mult) * redundant_score
            )
            if equation_score > best_score:
                best_score = equation_score
                idx_to_add = i
        idxs.append(idx_to_add)
        selected = np.append(selected, [embedding_list[idx_to_add]], axis=0)
    return idxs