Chapter 5 Model Management in the Database

Vectorization is performed using pre-trained models called embedding models. By managing models in a database, you can integrate model deployment (loading) with database operations and achieve fine-grained access control at the model level. When managing models in a database, you can use ONNX format models that integrate a series of processes for vectorization into a single file. Additionally, to perform vectorization (inference by the model), a inference server called Triton Inference Server is used.

For the formats of models that can be handled by Fujitsu Enterprise Postgres, refer to the "5.2.6.1 Database-side Model Specifications". Also, for requirements regarding the inference server, refer to "5.2.1 Setting Up Inference Server".