Loading

Elastic Inference Service for self-managed clusters

Elastic Inference Service (EIS) is available with zero setup on Elastic Cloud Hosted and Serverless deployments. To use EIS with other deployment types, you can use Cloud Connect. Cloud Connect enables you to use Elastic Cloud services in your self-managed cluster without having to install and maintain their infrastructure yourself.

You can use EIS to enable features such as:

For a full list of EIS-powered features, refer to AI features powered by EIS.

Before you can use EIS with your self-managed cluster, ensure you meet the following requirements:

  1. Open Cloud Connect

    In your self-managed Kibana instance, navigate to the Cloud Connect page using the search bar.

    Screenshot showing Cloud Connect page

  2. Get your Cloud Connect API key

    Sign up or log in to Elastic Cloud and get the Cloud Connect API key:

    • If you don’t have an account yet, click Sign up and follow the prompts to create your account and start a free Cloud Trial.
    • If you already have an Elastic Cloud account, click Log in.

  3. Connect your cluster

    Copy the Cloud Connect API key, paste it into your self-managed cluster's Cloud Connect page, then click Connect.

  4. Enable Elastic Inference Service

    On the Cloud connected services page, click Connect for Elastic Inference Service.

    Screenshot showing Cloud Connect and EIS

After you connect Elastic Inference Service through Cloud Connect, Elasticsearch automatically creates multiple inference endpoints for search and chat use cases, along with corresponding Kibana AI connectors. Supported Kibana features now use these connectors automatically.

In this example, you create an index with a semantic_text field, index a document, then run a query that returns a semantically related match.

In Dev Tools, run the following requests:

  1. Create an index with a semantic_text field 

    				PUT /semantic-search-eis
					{
  "mappings": {
    "properties": {
      "text": {
        "type": "semantic_text"
      }
    }
  }
}
    1. Because you already enabled EIS, the semantic_text field type uses EIS through the default inference endpoint (.elser-2-elastic). To learn more, refer to semantic_text.

  2. Index a document 

    				POST /semantic-search-eis/_doc
					{
  "text": "Aberdeen Football Club"
}

  3. Run a search query 

    				GET /semantic-search-eis/_search
					{
  "query": {
    "match": {
      "text": "soccer"
    }
  }
}

The response should include the indexed document:

{
  "took": 161,
  "timed_out": false,
  "_shards": {
    "total": 1,
    "successful": 1,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 1,
      "relation": "eq"
    },
    "max_score": 4.729913,
    "hits": [
      {
        "_index": "semantic-search-eis",
        "_id": "oyH935sBG2FaZ-zOMrer",
        "_score": 4.729913,
        "_source": {
          "text": "Aberdeen Football Club"
        }
      }
    ]
  }
}

For information about EIS regions and request routing, refer to Region and hosting.

EIS is billed per million tokens. For details on pricing and usage tracking, refer to Pricing and Monitor your token usage.