Class LocalEmbeddings

LocalEmbeddings provides local embedding generation using a runtime pipeline. Implements EmbeddingsInterface for compatibility with LangChain.

Implements

EmbeddingsInterface<number[] | Float32Array>

Index

Constructors

Methods

embedDocuments embedDocumentsRaw embedQuery

Constructors

constructor

new LocalEmbeddings(
runtime: LocalEmbeddingsRuntime,
hooks?: EmbeddingPipelineHooks,
): LocalEmbeddings
Create a LocalEmbeddings instance.
Parameters
- runtime: LocalEmbeddingsRuntime
  The runtime configuration for the embedding pipeline.
- hooks: EmbeddingPipelineHooks = {}
  Optional hooks for pipeline events.
Returns LocalEmbeddings
- Defined in pipeline/LocalEmbeddings.ts:35

Methods

embedDocuments

embedDocuments(
documents: string[],
): Promise<(Float32Array<ArrayBufferLike> | number[])[]>
Embed an array of documents.
Parameters
- documents: string[]
  Array of strings to embed.
Returns Promise<(Float32Array<ArrayBufferLike> | number[])[]>
Array of embeddings (number[] or Float32Array).
Implementation of EmbeddingsInterface.embedDocuments
- Defined in pipeline/LocalEmbeddings.ts:47

embedDocumentsRaw

embedDocumentsRaw(
documents: string[],
): Promise<{ buffer: ArrayBuffer; dims: number; rows: number }>
Return a packed ArrayBuffer with embeddings laid out row-major (rows * dims). This can be transferred or used to create typed-array views without additional allocations.
Parameters
- documents: string[]
  Array of strings to embed.
Returns Promise<{ buffer: ArrayBuffer; dims: number; rows: number }>
Object with buffer, rows, and dims.
- Defined in pipeline/LocalEmbeddings.ts:59

embedQuery

embedQuery(document: string): Promise<Float32Array<ArrayBufferLike> | number[]>
Embed a single query/document.
Parameters
- document: string
  String to embed.
Returns Promise<Float32Array<ArrayBufferLike> | number[]>
Embedding (number[] or Float32Array).
Implementation of EmbeddingsInterface.embedQuery
- Defined in pipeline/LocalEmbeddings.ts:75