Skip to main content

Elasticsearch

https://www.elastic.co/

Maven Dependency​

<dependency>
    <groupId>dev.langchain4j</groupId>
    <artifactId>langchain4j-elasticsearch</artifactId>
    <version>0.36.2</version>
</dependency>

APIs​

The ElasticsearchEmbeddingStore comes with 2 implementations:

  • Using approximate kNN queries with the ElasticsearchConfigurationKnn configuration class (default).
  • Using scriptScore queries with the ElasticsearchConfigurationScript configuration class. Note that this implementation is using cosine similarity.

Common options​

To create the ElasticsearchEmbeddingStore instance, you need to provide an Elasticsearch RestClient:

String apiKey = "VnVhQ2ZHY0JDZGJrU...";
RestClient restClient = RestClient
    .builder(HttpHost.create("https://localhost:9200"))
    .setDefaultHeaders(new Header[]{
        new BasicHeader("Authorization", "ApiKey " + apiKey)
    })
    .build();

Note:

See the Elasticsearch documentation on how to create a RestClient instance.

Then you can create the embedding store. It uses the approximate kNN query implementation by default.

ElasticsearchEmbeddingStore store = ElasticsearchEmbeddingStore.builder()
        .restClient(restClient)
        .build();

ElasticsearchConfigurationKnn configuration (default)​

The previous is equivalent to:

ElasticsearchEmbeddingStore store = ElasticsearchEmbeddingStore.builder()
        .configuration(ElasticsearchConfigurationKnn.builder().build())
        .restClient(restClient)
        .build();

ElasticsearchConfigurationScript configuration​

If you want to use the previous, but slower behavior, you can use the ElasticsearchConfigurationScript configuration class:

ElasticsearchEmbeddingStore store = ElasticsearchEmbeddingStore.builder()
        .configuration(ElasticsearchConfigurationScript.builder().build())
        .restClient(restClient)
        .build();

Examples​