David Snelling
f8c45f2d8d
Open source vector database with HNSW indexing, graph relationships, and metadata facets. Features CLI with professional augmentation registry integration for discovering extensions and capabilities.
102 lines
3.3 KiB
TypeScript
102 lines
3.3 KiB
TypeScript
/**
|
|
* Embedding functions for converting data to vectors using Transformers.js
|
|
* Complete rewrite to eliminate TensorFlow.js and use ONNX-based models
|
|
*/
|
|
import { EmbeddingFunction, EmbeddingModel, Vector } from '../coreTypes.js';
|
|
/**
|
|
* Detect the best available GPU device for the current environment
|
|
*/
|
|
export declare function detectBestDevice(): Promise<'cpu' | 'webgpu' | 'cuda'>;
|
|
/**
|
|
* Resolve device string to actual device configuration
|
|
*/
|
|
export declare function resolveDevice(device?: string): Promise<string>;
|
|
/**
|
|
* Transformers.js Sentence Encoder embedding model
|
|
* Uses ONNX Runtime for fast, offline embeddings with smaller models
|
|
* Default model: all-MiniLM-L6-v2 (384 dimensions, ~90MB)
|
|
*/
|
|
export interface TransformerEmbeddingOptions {
|
|
/** Model name/path to use - defaults to all-MiniLM-L6-v2 */
|
|
model?: string;
|
|
/** Whether to enable verbose logging */
|
|
verbose?: boolean;
|
|
/** Custom cache directory for models */
|
|
cacheDir?: string;
|
|
/** Force local files only (no downloads) */
|
|
localFilesOnly?: boolean;
|
|
/** Quantization setting (fp32, fp16, q8, q4) */
|
|
dtype?: 'fp32' | 'fp16' | 'q8' | 'q4';
|
|
/** Device to run inference on - 'auto' detects best available */
|
|
device?: 'auto' | 'cpu' | 'webgpu' | 'cuda' | 'gpu';
|
|
}
|
|
export declare class TransformerEmbedding implements EmbeddingModel {
|
|
private extractor;
|
|
private initialized;
|
|
private verbose;
|
|
private options;
|
|
/**
|
|
* Create a new TransformerEmbedding instance
|
|
*/
|
|
constructor(options?: TransformerEmbeddingOptions);
|
|
/**
|
|
* Get the default cache directory for models
|
|
*/
|
|
private getDefaultCacheDir;
|
|
/**
|
|
* Check if we're running in a test environment
|
|
*/
|
|
private isTestEnvironment;
|
|
/**
|
|
* Log message only if verbose mode is enabled
|
|
*/
|
|
private logger;
|
|
/**
|
|
* Initialize the embedding model
|
|
*/
|
|
init(): Promise<void>;
|
|
/**
|
|
* Generate embeddings for text data
|
|
*/
|
|
embed(data: string | string[]): Promise<Vector>;
|
|
/**
|
|
* Dispose of the model and free resources
|
|
*/
|
|
dispose(): Promise<void>;
|
|
/**
|
|
* Get the dimension of embeddings produced by this model
|
|
*/
|
|
getDimension(): number;
|
|
/**
|
|
* Check if the model is initialized
|
|
*/
|
|
isInitialized(): boolean;
|
|
}
|
|
export declare const UniversalSentenceEncoder: typeof TransformerEmbedding;
|
|
/**
|
|
* Create a new embedding model instance
|
|
*/
|
|
export declare function createEmbeddingModel(options?: TransformerEmbeddingOptions): EmbeddingModel;
|
|
/**
|
|
* Default embedding function using the lightweight transformer model
|
|
*/
|
|
export declare const defaultEmbeddingFunction: EmbeddingFunction;
|
|
/**
|
|
* Create an embedding function with custom options
|
|
*/
|
|
export declare function createEmbeddingFunction(options?: TransformerEmbeddingOptions): EmbeddingFunction;
|
|
/**
|
|
* Batch embedding function for processing multiple texts efficiently
|
|
*/
|
|
export declare function batchEmbed(texts: string[], options?: TransformerEmbeddingOptions): Promise<Vector[]>;
|
|
/**
|
|
* Embedding functions for specific model types
|
|
*/
|
|
export declare const embeddingFunctions: {
|
|
/** Default lightweight model (all-MiniLM-L6-v2, 384 dimensions) */
|
|
default: EmbeddingFunction;
|
|
/** Create custom embedding function */
|
|
create: typeof createEmbeddingFunction;
|
|
/** Batch processing */
|
|
batch: typeof batchEmbed;
|
|
};
|