Class NodeEmbedder

IEmbedder — code embedding abstraction.

All platforms (Node/Web) implement this interface. Consumers depend only on IEmbedder, never on platform specifics.

Implements

IEmbedder

Index

Properties

dimensions maxSequenceLength modelInfo

Methods

create createFromPackage embed embedBatch dispose

Properties

`Readonly`dimensions

dimensions: number

Embedding dimension (nomic-embed-code: 768)

`Readonly`maxSequenceLength

maxSequenceLength: number

Maximum sequence length (nomic-embed-code: 512)

`Readonly`modelInfo

modelInfo: ModelInfo

Model identifier

Methods

`Static`create

create(options: NodeEmbedderOptions): Promise<NodeEmbedder>
Create from a model file. The tokenizer.json is expected alongside the model.
Parameters
- options: NodeEmbedderOptions
Returns Promise<NodeEmbedder>
- Defined in packages/embed-code-node/src/onnx-embedder.ts:57

`Static`createFromPackage

createFromPackage(): Promise<NodeEmbedder>
Create from the embedded model in this package's models/ directory.

Returns Promise<NodeEmbedder>
- Defined in packages/embed-code-node/src/onnx-embedder.ts:86

embed

embed(text: string): Promise<Float32Array<ArrayBufferLike>>
Embed a single text → 768d Float32Array
Parameters
- text: string
Returns Promise<Float32Array<ArrayBufferLike>>
Implementation of IEmbedder.embed
- Defined in packages/embed-code-node/src/onnx-embedder.ts:92

embedBatch

embedBatch(
texts: string[],
options?: BatchOptions,
): Promise<Float32Array<ArrayBufferLike>[]>
Embed a batch with concurrency control + progress + backpressure
Parameters
- texts: string[]
- Optionaloptions: BatchOptions
Returns Promise<Float32Array<ArrayBufferLike>[]>
Implementation of IEmbedder.embedBatch
- Defined in packages/embed-code-node/src/onnx-embedder.ts:112

dispose

dispose(): Promise<void>
Release ONNX session / buffers

Returns Promise<void>
Implementation of IEmbedder.dispose
- Defined in packages/embed-code-node/src/onnx-embedder.ts:124