Skip to main content

Class: HashEmbedder

Deterministic embedder: hashed bag of words, L2-normalised. Texts that share words get similar vectors, so retrieval ranks sensibly in tests.

Constructors​

Constructor​

new HashEmbedder(options?): HashEmbedder;

Parameters​

ParameterTypeDescription
options?{ dimensions?: number; }-
options.dimensions?numberVector length

Returns​

HashEmbedder

Properties​

PropertyType
dimensionsnumber
metadataobject
metadata.namestring
metadata.typestring
metadata.versionstring

Methods​

embed()​

embed(text, options?): Promise<number[]>;

Parameters​

ParameterTypeDescription
textstring-
options?{ signal?: AbortSignal; }-
options.signal?AbortSignal-

Returns​

Promise<number[]>


embedBatch()​

embedBatch(texts, options?): Promise<number[][]>;

Parameters​

ParameterTypeDescription
textsstring[]-
options?{ signal?: AbortSignal; }-
options.signal?AbortSignal-

Returns​

Promise<number[][]>


embedQuery()​

embedQuery(text, options?): Promise<number[]>;

Same as embed().

Parameters​

ParameterTypeDescription
textstring-
options?{ signal?: AbortSignal; }-
options.signal?AbortSignal-

Returns​

Promise<number[]>


getCacheKey()​

getCacheKey(role): string;

Cache identity for createRagPipeline({ cache }).

Parameters​

ParameterTypeDescription
rolestring-

Returns​

string