Class OracleEmbeddingStore

java.lang.Object
dev.langchain4j.store.embedding.oracle.OracleEmbeddingStore
All Implemented Interfaces:
EmbeddingStore<TextSegment>

public final class OracleEmbeddingStore extends Object implements EmbeddingStore<TextSegment>

An EmbeddingStore which uses AI Vector Search capabilities of Oracle Database. This embedding store supports metadata filtering and removal

Instances of this store are created by configuring a builder:


 EmbeddingStore<TextSegment> example(DataSource dataSource) {
   return OracleEmbeddingStore.builder()
     .dataSource(dataSource)
     .embeddingTable("example")
     .build();
 }
 

It is recommended to configure a DataSource which pools connections, such as the Universal Connection Pool (UCP) or Hikari. A connection pool will avoid the latency of repeatedly creating new database connections.

This embedding store requires a EmbeddingTable to be configured with OracleEmbeddingStore.Builder.embeddingTable(String). If the table does not already exist, it can be created by passing a CreateOption to OracleEmbeddingStore.Builder.embeddingTable(String, CreateOption) or to EmbeddingTable.Builder.createOption(CreateOption).

By default no index is created on the EmbeddingTable. The OracleEmbeddingStore.Builder.index(Index...) allows to create indexes on the embedding and metadata columns of the embedding table. Two builders allow to configure an Index: IVFIndexBuilder to configure an IVF index on the embedding column and JSONIndexBuilder to configure function-based indexes on keys of the metadata column.

Instances of this embedding store are safe for use by multiple threads.