Skip to main content

graniteEmbeddingClient

Workspace API

Workspace API / services/granite/graniteEmbeddingClient

services/granite/graniteEmbeddingClient

Classes​

EmbeddingInputTooLongError​

Defined in: services/granite/graniteEmbeddingClient.ts:45

Error thrown when embedding input exceeds token limit

Extends​

  • Error

Constructors​

Constructor​

new EmbeddingInputTooLongError(message, estimatedTokens, maxTokens): EmbeddingInputTooLongError

Defined in: services/granite/graniteEmbeddingClient.ts:46

Parameters​
message​

string

estimatedTokens​

number

maxTokens​

number

Returns​

EmbeddingInputTooLongError

Overrides​

Error.constructor

Properties​

estimatedTokens​

readonly estimatedTokens: number

Defined in: services/granite/graniteEmbeddingClient.ts:48

maxTokens​

readonly maxTokens: number

Defined in: services/granite/graniteEmbeddingClient.ts:49

EmbeddingDimensionError​

Defined in: services/granite/graniteEmbeddingClient.ts:59

Error thrown when embedding dimensions don't match expected

Extends​

  • Error

Constructors​

Constructor​

new EmbeddingDimensionError(expected, actual): EmbeddingDimensionError

Defined in: services/granite/graniteEmbeddingClient.ts:60

Parameters​
expected​

number

actual​

number

Returns​

EmbeddingDimensionError

Overrides​

Error.constructor

Properties​

expected​

readonly expected: number

Defined in: services/granite/graniteEmbeddingClient.ts:61

actual​

readonly actual: number

Defined in: services/granite/graniteEmbeddingClient.ts:62

GraniteEmbeddingClient​

Defined in: services/granite/graniteEmbeddingClient.ts:113

Client for Granite Embedding model via vLLM OpenAI-compatible API

Constructors​

Constructor​

new GraniteEmbeddingClient(): GraniteEmbeddingClient

Defined in: services/granite/graniteEmbeddingClient.ts:120

Returns​

GraniteEmbeddingClient

Methods​

generateEmbedding()​

generateEmbedding(text): Promise<number[]>

Defined in: services/granite/graniteEmbeddingClient.ts:171

Generate embedding for a single text

Parameters​
text​

string

Returns​

Promise<number[]>

generateEmbeddings()​

generateEmbeddings(texts, batchSize): Promise<number[][]>

Defined in: services/granite/graniteEmbeddingClient.ts:180

Generate embeddings for multiple texts in batch

Parameters​
texts​

string[]

batchSize​

number = 32

Returns​

Promise<number[][]>

Throws​

EmbeddingInputTooLongError if any input exceeds 512 tokens

generateQueryEmbedding()​

generateQueryEmbedding(query): Promise<number[]>

Defined in: services/granite/graniteEmbeddingClient.ts:335

Generate query embedding with optional query prefix Some embedding models perform better with "query: " prefix for retrieval

Parameters​
query​

string

Returns​

Promise<number[]>

Variables​

GRANITE_EMBEDDING_MODEL_ID​

const GRANITE_EMBEDDING_MODEL_ID: string = EMBEDDING_CONFIG.model

Defined in: services/granite/graniteEmbeddingClient.ts:73

GRANITE_EMBEDDING_DIMENSIONS​

const GRANITE_EMBEDDING_DIMENSIONS: number = EMBEDDING_CONFIG.dimensions

Defined in: services/granite/graniteEmbeddingClient.ts:74

GRANITE_EMBEDDING_MAX_TOKENS​

const GRANITE_EMBEDDING_MAX_TOKENS: number = EMBEDDING_CONFIG.maxTokens

Defined in: services/granite/graniteEmbeddingClient.ts:75

Functions​

getGraniteEmbeddingClient()​

getGraniteEmbeddingClient(): GraniteEmbeddingClient

Defined in: services/granite/graniteEmbeddingClient.ts:347

Returns​

GraniteEmbeddingClient