Voyage Embeddings

The Voyage embeddings provider converts text into vectors using Voyage AI’s embedding models. Voyage offers state-of-the-art embeddings with specialized models for code, finance, and legal domains.

Installation

npm install @youcraft/recall-embeddings-voyage

Usage

import { createMemory } from '@youcraft/recall'
import { voyageEmbeddings } from '@youcraft/recall-embeddings-voyage'

const embeddings = voyageEmbeddings({
  apiKey: process.env.VOYAGE_API_KEY!,
})

const memory = createMemory({ db, embeddings, extractor })

Configuration

Option	Type	Default	Description
apiKey	string	required	Your Voyage API key
model	string	"voyage-3-lite"	Embedding model to use
inputType	string	"document"	Input type for embeddings

Models

Model	Dimensions	Description
voyage-3-large	1024	Highest quality general-purpose embeddings
voyage-3	1024	High quality general-purpose embeddings
voyage-3-lite	512	Fast, efficient general-purpose embeddings
voyage-code-3	1024	Optimized for code retrieval
voyage-finance-2	1024	Optimized for finance domain
voyage-law-2	1024	Optimized for legal domain
voyage-multilingual-2	1024	Multilingual support

Examples

High quality embeddings

const embeddings = voyageEmbeddings({
  apiKey: process.env.VOYAGE_API_KEY!,
  model: 'voyage-3-large',
})

Code retrieval

const embeddings = voyageEmbeddings({
  apiKey: process.env.VOYAGE_API_KEY!,
  model: 'voyage-code-3',
})

Domain-specific models

// For finance applications
const embeddings = voyageEmbeddings({
  apiKey: process.env.VOYAGE_API_KEY!,
  model: 'voyage-finance-2',
})

// For legal applications
const embeddings = voyageEmbeddings({
  apiKey: process.env.VOYAGE_API_KEY!,
  model: 'voyage-law-2',
})

Input Types

Voyage models support different input types optimized for specific use cases:

Type	Use Case
document	Embedding documents to be searched (default)
query	Embedding search queries

For memory storage and retrieval, document is typically the best choice.

Batch Processing

The provider automatically handles Voyage’s batch size limit (128 texts per request):

// Handled automatically - splits into batches of 128
await embeddings.embedBatch(texts) // Works with any size array