add_dense_and_sparse

Function

Adds a document index to a vector database. Specifically, this API first performs embedding on a document chunk to obtain dense vectors and sparse vectors, and then stores the vectors into a vector database.

Prototype

def add_dense_and_sparse(ids, dense_embeddings, sparse_embeddings, docs, metadatas, **kwargs)

Parameters

Parameter

Data Type

Required/Optional

Description

ids

List[int]

Required

Index ID list of vectors to be added. Its length range is [0, 10 million).

dense_embeddings

ndarray

Required

NumPy array object.

sparse_embeddings

List[Dict[int, float]]

Required

Sparse vector object.

docs

List[str]

Optional

Document to which vectors will be added.

metadatas

List[dict]

Optional

Metadata of the document to which vectors will be added.

kwargs

Dict

Optional

Keyword parameter, which can only be document_id, to specify the ID of the document to which vectors are to be added. Other keyword parameters are invalid.

The shape of dense_embeddings must be 2D. The number of vectors contained in dense_embeddings and sparse_embeddings must be equal to the length of ids. The number of documents contained in docs must be equal to the length of ids. The total number of vectors added at a time must be less than 10 million.