Index

class vecworks.Index

Data class specifying how a Retriever may access an index associated with a vector store.

Fundamentals

__init__(name: str, distance: DISTANCES, vectorizer: Vectorizer | None = None, density: DENSITY | None = None, max_distance: float | None = None, top_k: int | None = None, bind: str | None = None, return_distance_as: str | None = None)

Initializes the dataclass.

Parameters

name

Name of the vector index to access.

distance

Distance function, specified using DISTANCES, to calculate the similarity of the (vectorized) input passed to the retriever with the indexed vectors.

vectorizer

Vectorizer used to vectorize input passed to the retriever. If no vectorizer is specified, the input will be kept as-is.

density

The density of the vectors stored in the index, as classified using DENSITY.

max_distance

Maximum distance under which indices are still considered for selection. This value depends on the distance function applied. Refer to DISTANCES for guidance.

top_k

Maximum number of hits to retrieve from the vector index. If neither threshold nor top_k are set, this parameter is automatically set to 10.

bind

Name of the argument passed to the Retriever, which contents should be vectorized and compared with the index. If no name is passed, it is set to the contents of name.

return_distance_as

Name of the return variable holding the calculate distance. If no name is specified, the distance is not included in the output.