Algolia Recommend tool

Use the Algolia Recommend tool to provide personalized recommendations based on trends and user behavior.

This is a beta feature according to Algolia’s Terms of Service (“Beta Services”).

Your agent can use Algolia’s recommendation models to suggest trending items, complementary products, or visually similar alternatives.

For example, for the query “What’s trending in running shoes?”, the agent retrieves trending products from your Recommend API instead of guessing or using outdated training data.

Key capabilities

• Trending items: show globally trending products or trending within specific categories
• Trending facets: discover popular categories, brands, or attributes
• Bought together: suggest items frequently purchased with a given product (for example, during checkout or product views)
• Related products: find alternatives or similar items (useful for out-of-stock products)
• Looking similar: recommend visually similar products based on image analysis
• Configurable parameters: apply a filter, set a result limit, or customize a recommendation

Use a Search API key (never an Admin API key). For more information, see Algolia Recommend tool security.

Configure the Recommend tool

From the dashboard

From the Agent Studio agent edit view in the Algolia dashboard:

1. Click Add tool > Other tools
2. Configure the Recommend tool using the JSON Schema (see API tab for example)
3. Click Add tool

With the API

JSON tool definition:

JSON

{
  "type": "algolia_recommend",
  "name": "product_recommendations",
  "allowedConfigs": [
    {
      "modelName": "trending-items",
      "index": "products",
      "description": "Returns trending products globally or by category"
    },
    {
      "modelName": "bought-together",
      "index": "products",
      "description": "Returns items frequently bought with a given product"
    },
    {
      "modelName": "related-products",
      "index": "products",
      "description": "Returns alternative or similar products"
    }
  ],
  "predefinedRecommendParameters": {
    "queryParameters": {
      "filters": "inStock:true",
      "attributesToRetrieve": ["title", "price", "image", "rating"]
    }
  }
}

Required fields

• type: must be "algolia_recommend"
• allowedConfigs: array of recommendation model configurations (minimum 1)
  • modelName: one of trending-items, trending-facets, bought-together, related-products, or looking-similar
  • index: Algolia index name (1-100 characters)
  • description: describe when to use this recommendation model

Optional fields

• name: custom name for this tool instance (defaults to "algolia_recommend", 3-32 characters)
• predefinedRecommendParameters: global parameters applied to all recommendation requests
  • queryParameters: filters or search parameters
    • filters: apply filters (for example, "inStock:true AND isPublished:true")
    • facets: facets to return in the response
    • attributesToRetrieve: limit returned attributes
  • threshold: confidence threshold (0-100, defaults to 0)
  • maxRecommendations: number of results (defaults to 10)

Runtime parameters

The agent can dynamically adjust these at runtime:

• Item-based models (related-products, bought-together, looking-similar): requires objectID
• Trend-based models (trending-items, trending-facets): can filter by facetName and facetValue
• maxRecommendations: number of recommendations to fetch
• threshold: confidence level for recommendation quality

For complete configuration details, see the Agent Studio API reference.

See also

• Tools overview
• Algolia Recommend
• Credentials