| Name | Type | Description |
|---|---|---|
embeddings* | list[list[float]] | List of embeddings vectors. |
k | int | Default: 4Number of neighbors to be retrieved. |
filter_ | list[Namespace] | dict | None | Default: None |
numeric_filter | list[NumericNamespace] | None | Default: None |
sparse_embeddings | list[dict[str, list[int] | list[float]]] | None | Default: None |
rrf_ranking_alpha | float | Default: 1 |
Finds the k closes neighbors of each instance of embeddings.
For v1: list of Namespace objects. For v2: dict.
List of NumericNamespace objects for filtering (v1 only).
List of Sparse embedding dictionaries which represents an
embedding as a list of indices and as a list of sparse values:
ie. [{"values": [0.7, 0.5], "indices": [10, 20]}]
Reciprocal Ranking Fusion weight, float between 0 and
1.0
Weights Dense Search VS Sparse Search, as an example:
rrf_ranking_alpha=1: Only Denserrf_ranking_alpha=0: Only Sparserrf_ranking_alpha=0.7: 0.7 weighting for dense and 0.3 for
sparse