In-Memory H2 Vector Spring Boot Starter

The in-memory-h2vector-spring-boot-starter is a Spring Boot starter that provides support for vector fields and vector similarity functions using an in-memory H2 database. All vector operations are performed with java vector-api, leveraging efficient SIMD operations for vector similarity calculations. This starter simplifies the integration of vector-based operations into your Spring Boot applications.

Features

• In-memory vector database : Provides an in-memory H2 'extended' database for storing and querying vector fields.
• Simple integration: Entities as usual - easily add vector fields to your entities using JPA.
• Vector Similarity Functions: Provides pre-configured functions for vector similarity operations. currently supports cosine, and l2 similarity.

Requirements

• Jdk 22+, with JVM arg --add-modules=jdk.incubator.vector added.

Installation

To include this starter in your Spring Boot project, add the following dependency to your pom.xml:

<dependency>
    <groupId>io.github.benayat</groupId>
    <artifactId>in-memory-h2vector-spring-boot-starter</artifactId>
    <version>0.0.4</version>
</dependency>

Usage example

To define a vector field, you need to annotate a float[] field with @LOB to store the vector as a byte array, and @Convert to use a convert from float[] to byte[] and vice-versa.

@Getter
@NoArgsConstructor
@AllArgsConstructor
@ToString
@Builder
@Entity
public class TextRecord {
    @Id @GeneratedValue(strategy = GenerationType.AUTO)
    private long id;
    @Lob
    @Convert(converter = FloatArrayToByteArrayConverter.class)
    private float[] vector;
    private String text;
}

and then you can use any JPA repository to query the vector field for k-nearest neighbors using the cosine similarity function, with a native query.

public interface TextRecordRepository extends JpaRepository<TextRecord, Long> {
    @Query(value = "SELECT * FROM text_record record ORDER BY cosine_similarity(record.vector, :vector) DESC LIMIT :k", nativeQuery = true)
    List<TextRecord> findTopKByVectorSimilarity(@Param("vector") byte[] vector, @Param("k") int k);
}

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any changes.

Author

Benaya Trabelsi.

contact

email: [email protected]