Models and Inference

Oracle Autonomous AI Vector Database supports integrated embedding and reranking using in-database ONNX models to simplify pipelines and reduce external dependencies.

Oracle provides a number of ONNX embedding models that are available out-of-the box. They can be loaded into the database as needed from the Vector Database Console Models page. The following models are currently available:

Loading these models into the database is a one-time activity. For details, see Load a Model Using the Vector Database Console.

You also have the option to bring your own ONNX models. You can load any ONNX model by hosting it in Oracle Cloud Infrastructure Object Storage and calling the load_model API. After a model is loaded, the Autonomous AI Vector Database provides in-database inference for both embedding and reranking based on the loaded ONNX model. See Load Your Own ONNX Model for more information about the syntax and semantics of the load model function.

For examples of embedding and reranking, see Embed Data and Rerank Results.