Embedding models

Mastra's model router supports embedding models using the same provider/model string format as language models. This provides a unified interface for both chat and embedding models with TypeScript autocomplete support.

Quickstart

import { ModelRouterEmbeddingModel } from '@mastra/core/llm'
import { embedMany } from 'ai'

// Generate embeddings
const { embeddings } = await embedMany({
  model: new ModelRouterEmbeddingModel('openai/text-embedding-3-small'),
  values: ['Hello world', 'Semantic search is powerful'],
})

Supported models

OpenAI

• text-embedding-3-small - 1536 dimensions, 8191 max tokens
• text-embedding-3-large - 3072 dimensions, 8191 max tokens
• text-embedding-ada-002 - 1536 dimensions, 8191 max tokens

const embedder = new ModelRouterEmbeddingModel('openai/text-embedding-3-small')

Google

• gemini-embedding-001 - 768 dimensions, 2048 max tokens

const embedder = new ModelRouterEmbeddingModel('google/gemini-embedding-001')

VoyageAI

VoyageAI provides specialized embedding models optimized for retrieval tasks. These models are available as standalone packages:

npm

npm install @mastra/voyageai

pnpm

pnpm add @mastra/voyageai

Yarn

yarn add @mastra/voyageai

Bun

bun add @mastra/voyageai

Available models:

• voyage-4-large - 1024 dimensions (default), supports 256-2048 dimensions, best general-purpose and multilingual retrieval quality (120k max tokens per batch)
• voyage-4 - 1024 dimensions (default), supports 256-2048 dimensions, optimized for general-purpose and multilingual retrieval (320k max tokens per batch)
• voyage-4-lite - 1024 dimensions (default), supports 256-2048 dimensions, optimized for latency and cost (1M max tokens per batch)
• voyage-code-3 - 1024 dimensions (default), supports 256-2048 dimensions, optimized for code retrieval
• voyage-finance-2 - 1024 dimensions, optimized for finance retrieval and RAG
• voyage-law-2 - 1024 dimensions, optimized for legal retrieval and RAG (16k context)
• voyage-3-large - 1024 dimensions (default), supports 256-2048 dimensions (previous generation)
• voyage-3.5 - 1024 dimensions (default), supports 256-2048 dimensions (previous generation)
• voyage-3.5-lite - 1024 dimensions (default), supports 256-2048 dimensions, optimized for latency and cost (previous generation)
• voyage-multimodal-3.5 - 1024 dimensions, supports text + images

import { voyage, voyageEmbedding } from '@mastra/voyageai'

// Use default model (voyage-3.5)
const { embeddings } = await voyage.doEmbed({
  values: ['Hello world'],
})

// Use specific model (voyage-3-large)
const largeEmbeddings = await voyage.large.doEmbed({
  values: ['More complex content'],
})

// Custom configuration
const customModel = voyageEmbedding({
  model: 'voyage-3.5',
  inputType: 'query', // or 'document'
  outputDimension: 512, // 256, 512, 1024, or 2048
  baseUrl: 'https://ai.mongodb.com/v1', // Optional: custom endpoint (e.g. MongoDB-hosted Voyage)
})

const { embeddings: customEmbeddings } = await customModel.doEmbed({
  values: ['Custom configuration example'],
})

VoyageAI with MongoDB:

VoyageAI works seamlessly with MongoDB Atlas Vector Search:

import { voyage } from '@mastra/voyageai'
import { MongoDBVector } from '@mastra/mongodb'

const mongoVector = new MongoDBVector({
  id: 'mongodb-vector',
  uri: process.env.MONGODB_URI,
  dbName: process.env.MONGODB_DB_NAME,
})

// Create index matching VoyageAI dimensions
await mongoVector.createIndex({
  indexName: 'documents',
  dimension: 1024, // voyage-3.5 default
})

// Generate and store embeddings
const { embeddings } = await voyage.doEmbed({
  values: chunks.map(chunk => chunk.text),
})

await mongoVector.upsert({
  indexName: 'documents',
  vectors: embeddings,
  metadata: chunks.map(chunk => ({ text: chunk.text })),
})

Multimodal embeddings (text + images):

import { voyage } from '@mastra/voyageai'

const { embeddings } = await voyage.multimodal.doEmbed({
  values: [
    {
      content: [
        { type: 'text', text: 'Product description' },
        { type: 'image_url', image_url: 'https://example.com/image.jpg' },
      ],
    },
  ],
})

For more details, see the MongoDB + VoyageAI integration guide.

Authentication

The model router automatically detects API keys from environment variables:

• OpenAI: OPENAI_API_KEY
• Google: GOOGLE_API_KEY (falls back to GOOGLE_GENERATIVE_AI_API_KEY)
• VoyageAI: VOYAGE_API_KEY

# .env
OPENAI_API_KEY=sk-...
GOOGLE_API_KEY=...
VOYAGE_API_KEY=pa-...

Custom Providers

You can use any OpenAI-compatible embedding endpoint with a custom URL:

import { ModelRouterEmbeddingModel } from '@mastra/core/llm'

const embedder = new ModelRouterEmbeddingModel({
  providerId: 'ollama',
  modelId: 'nomic-embed-text',
  url: 'http://localhost:11434/v1',
  apiKey: 'not-needed', // Some providers don't require API keys
})

Usage with Memory

The embedding model router integrates seamlessly with Mastra's memory system:

import { Memory } from '@mastra/memory'
import { Agent } from '@mastra/core/agent'
import { ModelRouterEmbeddingModel } from '@mastra/core/llm'

const agent = new Agent({
  id: 'my-agent',
  name: 'my-agent',
  instructions: 'You are a helpful assistant',
  model: 'openai/gpt-5.1',
  memory: new Memory({
    embedder: new ModelRouterEmbeddingModel('openai/text-embedding-3-small'),
  }),
})

note

The embedder field accepts:

• EmbeddingModelId (string with autocomplete)
• EmbeddingModel<string> (AI SDK v1)
• EmbeddingModelV2<string> (AI SDK v2)

Usage with RAG

Use embedding models for document chunking and retrieval:

import { ModelRouterEmbeddingModel } from '@mastra/core/llm'
import { embedMany } from 'ai'

// Embed document chunks
const { embeddings } = await embedMany({
  model: new ModelRouterEmbeddingModel('openai/text-embedding-3-small'),
  values: chunks.map(chunk => chunk.text),
})

// Store embeddings in your vector database
await vectorStore.upsert(
  chunks.map((chunk, i) => ({
    id: chunk.id,
    vector: embeddings[i],
    metadata: chunk.metadata,
  })),
)

TypeScript Support

The model router provides full TypeScript autocomplete for embedding model IDs:

import type { EmbeddingModelId } from '@mastra/core'

// Type-safe embedding model selection
const modelId: EmbeddingModelId = 'openai/text-embedding-3-small'
//                                  ^ Autocomplete shows all supported models

const embedder = new ModelRouterEmbeddingModel(modelId)

Error handling

The model router validates provider and model IDs at construction time:

try {
  const embedder = new ModelRouterEmbeddingModel('invalid/model')
} catch (error) {
  console.error(error.message)
  // "Unknown provider: invalid. Available providers: openai, google"
}

Missing API keys are also caught early:

try {
  const embedder = new ModelRouterEmbeddingModel('openai/text-embedding-3-small')
  // Throws if OPENAI_API_KEY is not set
} catch (error) {
  console.error(error.message)
  // "API key not found for provider openai. Set OPENAI_API_KEY environment variable."
}

Next Steps

• Memory & Semantic Recall: Use embeddings for agent memory
• RAG & Chunking: Build retrieval-augmented generation systems
• Vector Databases: Store and query embeddings