Semantic Search with Voyage AI Embeddings

Set up Meilisearch with Voyage AI’s v3.5 embedding models for semantic search.

Voyage AI provides high-quality embedding models optimized for search and retrieval. This guide shows you how to configure Meilisearch with Voyage AI embeddings using the REST embedder.

Requirements

• A Meilisearch project
• A Voyage AI account with an API key

Available models

Voyage AI offers the following embedding models:

Model	Use case
voyage-3.5-lite	Fast, cost-effective general-purpose embeddings
voyage-3.5	Balanced quality and performance
voyage-3-large	Highest quality, larger model

The older voyage-2 family (voyage-2, voyage-large-2, voyage-large-2-instruct, voyage-multilingual-2) is still supported but Voyage recommends upgrading to the 3.5 series for better performance.

Configure the embedder

Update your index settings with the Voyage AI embedder configuration:

{
  "voyage": {
    "source": "rest",
    "apiKey": "<VOYAGE_API_KEY>",
    "documentTemplate": "A product named '{{doc.name}}': {{doc.description}}",
    "url": "https://api.voyageai.com/v1/embeddings",
    "request": {
      "model": "voyage-3.5-lite",
      "input": ["{{text}}", "{{..}}"]
    },
    "response": {
      "data": [
        { "embedding": "{{embedding}}" },
        "{{..}}"
      ]
    }
  }
}

Send this configuration to Meilisearch:

curl \
  -X PATCH 'MEILISEARCH_URL/indexes/INDEX_NAME/settings' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer MEILISEARCH_KEY' \
  --data-binary '{
    "embedders": {
      "voyage": {
        "source": "rest",
        "apiKey": "<VOYAGE_API_KEY>",
        "documentTemplate": "A product named '\''{{doc.name}}'\'': {{doc.description}}",
        "url": "https://api.voyageai.com/v1/embeddings",
        "request": {
          "model": "voyage-3.5-lite",
          "input": ["{{text}}", "{{..}}"]
        },
        "response": {
          "data": [
            { "embedding": "{{embedding}}" },
            "{{..}}"
          ]
        }
      }
    }
  }'

Replace <VOYAGE_API_KEY> with your actual Voyage AI API key. Adjust model depending on your quality and cost requirements.

Meilisearch handles batching and rate limiting automatically. Monitor the tasks queue to track indexing progress.

Test the search

{
  "q": "comfortable shoes for walking",
  "hybrid": {
    "semanticRatio": 0.5,
    "embedder": "voyage"
  }
}

Next steps

• Document template best practices to optimize which fields are embedded
• Custom hybrid ranking to tune the balance between keyword and semantic results
• Embedder settings reference for all configuration options