AIDB 1.0.7 release notes v7
Released: 10 Dec 2024
AIDB 1.0.7 is the GA release of AIDB. Since the technical preview, a number of enhancements have been made.
| Type | Description |
|---|---|
| Enhancement | Integrated model runtime: The technical preview had a separate model runtime. The model runtime is now integrated into the product. |
| Enhancement | Support for external models with OpenAI API: The technical preview only supported internal models. External models can now be used via the OpenAI API. |
| Enhancement | Secure management of API credentials: API credentials are now securely managed via the Postgres user mapping feature. The technical preview had no secure credential management. |
| Enhancement | Model registry for configuring internal and external models: A model registry allows configuring and switching between internal and external models. This is a new concept over the technical preview which had only retrievers. |
| Enhancement | PGFS allows working with external data from S3 object stores and file systems: The technical preview could only work with S3 in retrievers. The new volume concept allows working with external data in general. |
| Enhancement | Low-level primitives for models: Low-level primitives for models (encode, transform) have been added for both image and text data, complementing the existing high-level functions. |
| Enhancement | Enhanced retriever concept for data retrieval: Retriever calls can now return rich table results with data such as vector distance. The flexible interface allows retrieving just the key or the source data, enabling integration with custom workflows and pipelines. |
| Enhancement | Support for image and text data: Images can be retrieved from external storage or Postgres tables, including from bytea columns. Images can also be retrieved ad-hoc in a query. |
| Enhancement | Support for building fully custom AI workflows and pipelines: Direct SQL access to models (encode, transform, and more) enables batch execution. Flexible interfaces in high-level composite functions (such as retrieve) and direct access to low-level functions (such as embed_single_row) allow users to build custom workflows. |