Embedding Model

Spring AI provides EmbeddingModel to embed text. EmbeddingModel extends from Model. See Embedding Model for the underlying concept and how to choose a model; this page covers the Spring AI API for calling one.

note

EmbeddingModel is rarely used directly, but used indirectly with Vector Store.

EmbeddingModel

The model request is EmbeddingRequest. This request contains a list of Strings as the input and an EmbeddingOptions as the options.

EmbeddingRequest

public class EmbeddingRequest implements ModelRequest<List<String>> {
  private final List<String> inputs;

  @Nullable
  private final EmbeddingOptions options;
}

The model response is EmbeddingResponse. This response contains a list of Embeddings as the result.

EmbeddingResponse

public class EmbeddingResponse implements ModelResponse<Embedding> {
  private final List<Embedding> embeddings;

  private final EmbeddingResponseMetadata metadata;
}

An Embedding contains an array of float as the result.

Embedding

public class Embedding implements ModelResult<float[]> {
  
  private final float[] embedding;

  private final Integer index;

  private final EmbeddingResultMetadata metadata;

}    

The call method of EmbeddingModel sends an EmbeddingRequest to an embedding model and returns an EmbeddingResponse.

EmbeddingModel

public interface EmbeddingModel 
    extends Model<EmbeddingRequest, EmbeddingResponse> {
    @Override
    EmbeddingResponse call(EmbeddingRequest request);
}

EmbeddingModel also has some shortcut methods.

• float[] embed(String text)
• float[] embed(Document document)
• List<float[]> embed(List<String> texts)
• List<float[]> embed(List<Document> documents, EmbeddingOptions options, BatchingStrategy batchingStrategy)
• EmbeddingResponse embedForResponse(List<String> texts)

The dimensions method of EmbeddingModel returns the number of dimensions of the embedded vectors.

EmbeddingOptions

EmbeddingOptions represents the options for embedding models.

• getModel returns the model used for embedding.
• getDimensions returns the requested dimensions.

EmbeddingOptions is created using EmbeddingOptionsBuilder.

var options = EmbeddingOptionsBuilder.builder()
    .withModel("text-embedding-3-large")
    .withDimensions(1024)
    .build();

BatchingStrategy

BatchingStrategy is the strategy used for batching multiple documents.

public interface BatchingStrategy {
  
  List<List<Document>> batch(List<Document> documents);
}    

TokenCountBatchingStrategy is a built-in implementation of BatchingStrategy. This strategy batches documents based on their token counts, ensuring that each batch does not exceed the calculated maximum input token count.