langchain4j-oracle/README.md
This module implements EmbeddingStore using Oracle Database.
<dependency>
<groupId>dev.langchain4j</groupId>
<artificatId>langchain4j-oracle</artificatId>
<version>0.1.0</version>
</dependency>
Instances of this store can be created by configuring a builder. The builder requires that a DataSource and an embedding table be provided. The distance between two vectors is calculated using cosine similarity which measures the cosine of the angle between two vectors.
It is recommended to configure a DataSource with pools connections, such as the Universal Connection Pool or Hikari. A connection pool will avoid the latency of repeatedly creating new database connections.
If an embedding table already exists in your database provide the table name.
EmbeddingStore embeddingStore = OracleEmbeddingStore.builder()
.dataSource(myDataSource)
.embeddingTable("my_embedding_table")
.build();
If the table does not already exist, it can be created by passing a CreateOption to the builder.
EmbeddingStore embeddingStore = OracleEmbeddingStore.builder()
.dataSource(myDataSource)
.embeddingTable("my_embedding_table", CreateOption.CREATE_IF_NOT_EXISTS)
.build();
By default the embedding table will have the following columns:
| Name | Type | Description |
|---|---|---|
| id | VARCHAR(36) | Primary key. Used to store UUID strings which are generated when the embedding store |
| embedding | VECTOR(*, FLOAT32) | Stores the embedding |
| text | CLOB | Stores the text segment |
| metadata | JSON | Stores the metadata |
If the columns of your existing table do not match the predefined column names or you would like to use different column names, you can use a EmbeddingTable builder to configure your embedding table.
OracleEmbeddingStore embeddingStore =
OracleEmbeddingStore.builder()
.dataSource(myDataSource)
.embeddingTable(EmbeddingTable.builder()
.createOption(CREATE_OR_REPLACE) // use NONE if the table already exists
.name("my_embedding_table")
.idColumn("id_column_name")
.embeddingColumn("embedding_column_name")
.textColumn("text_column_name")
.metadataColumn("metadata_column_name")
.build())
.build();
The builder allows you to create indexes on the embedding and metadata columns of the EmbeddingTable by providing an instance of the Index class. Two builders allow you to create instances of the Index class: IVFIndexBuilder and JSONIndexBuilder.
IVFIndexBuilder allows you to configure an IVF (Inverted File Flat) index on the embedding column of the EmbeddingTable.
OracleEmbeddingStore embeddingStore =
OracleEmbeddingStore.builder()
.dataSource(myDataSource)
.embeddingTable(EmbeddingTable.builder()
.createOption(CreateOption.CREATE_OR_REPLACE) // use NONE if the table already exists
.name("my_embedding_table")
.idColumn("id_column_name")
.embeddingColumn("embedding_column_name")
.textColumn("text_column_name")
.metadataColumn("metadata_column_name")
.build())
.index(Index.ivfIndexBuilder().createOption(CreateOption.CREATE_OR_REPLACE).build())
.build();
JSONIndexBuilder allows you to configure a function-based index on keys of the metadata column of the EmbeddingTable.
OracleEmbeddingStore.builder()
.dataSource(myDataSource)
.embeddingTable(EmbeddingTable.builder()
.createOption(CreateOption.CREATE_OR_REPLACE) // use NONE if the table already exists
.name("my_embedding_table")
.idColumn("id_column_name")
.embeddingColumn("embedding_column_name")
.textColumn("text_column_name")
.metadataColumn("metadata_column_name")
.build())
.index(Index.jsonIndexBuilder()
.createOption(CreateOption.CREATE_OR_REPLACE)
.key("name", String.class, JSONIndexBuilder.Order.ASC)
.key("year", Integer.class, JSONIndexBuilder.Order.DESC)
.build())
.build();
For more information about Oracle AI Vector Search refer to the documentation.
This module also provides OracleChatMemoryStore, a simple persistent implementation of ChatMemoryStore.
Create a table:
CREATE TABLE chat_memory (
memory_id VARCHAR2(255) PRIMARY KEY,
content CLOB NOT NULL
);
Use it in chat memory:
ChatMemoryStore store = OracleChatMemoryStore.builder()
.dataSource(myDataSource)
.tableName("chat_memory")
.build();
ChatMemory chatMemory = MessageWindowChatMemory.builder()
.id("conversation-1")
.maxMessages(10)
.chatMemoryStore(store)
.build();
OracleChatMemoryStore stores one row per memory id, with all messages serialized as JSON in the content column.
By default, integration tests will run a docker image of Oracle Database using TestContainers. Alternatively, the tests can connect to an Oracle Database if the following environment variables are configured:
jdbc:oracle:thin@example:1521/serviceName