4.4.1.4 Creating a Database Target

Prerequisite: A Database connection.

To create a Database target:

  1. On the Catalog page, click Create New Item.
  2. Hover the mouse over Target and select Database Table from the submenu.
  3. On the Type Properties screen, enter the following details:
    • Name: Enter a unique name for the target. This is a mandatory field.
    • Display Name: Enter a display name for the target. If left blank, the Name field value is copied.
    • Description
    • Tags
    • Target Type: The selected target is displayed.
  4. Click Next.
  5. On the Target Details screen, enter the following details:

    • Connection: Select a database connection from the drop-down list.

  6. Click Next.
  7. On the Shape screen, enter the following details:
    • Table Name: Select a database table from the drop-down list.
      If you want to create a Vector Database target, select a table with vector columns. The Vector Data screen is displayed. Enter the following details:
      • Vector Field: From the drop-down, select the model you want to use to embed a vector field.

        Note:

        Oracle AI Database supports Open Neural Network Exchange (ONNX) models for generating vector embeddings directly within the database. As an example, you can utilize the downloadable, augmented version of Hugging Face’s all-MiniLM-L12-v2 model in ONNX format. This model can be loaded directly into the database via DBMS_VECTOR.LOAD_ONNX_MODEL or DBMS_DATA_MINING.IMPORT_ONNX_MODEL. All the models you have loaded will be displayed in the dropdown.

        Note:

        Oracle AI Database supports supports Open Neural Network Exchange (ONNX) models for generating vector embeddings directly within the database. As an example, you can utilize the downloadable, augmented version of Hugging Face’s all-MiniLM-L12-v2 model in ONNX format. This model can be loaded directly into the database via DBMS_VECTOR.LOAD_ONNX_MODEL or DBMS_DATA_MINING.IMPORT_ONNX_MODEL. All the models you have loaded, will be displayed in the dropdown.

        For example, to load a FINBERT model, run the following commands:
        -- Load ONNX model for embedding creation, run as DMUSER
EXECUTE DBMS_VECTOR.DROP_ONNX_MODEL(model_name => 'finbert', force => true);
EXECUTE DBMS_VECTOR.LOAD_ONNX_MODEL('DM_DUMP','FINBERT.onnx','finbert');

        For more information on vector embeddings, see Generate Vector Embedding.

        Note:

        Vector column will be displayed as a TEXT column, when using this Database target in a pipeline.
      • Click Save.
  8. Click Save.