AIDB 4.0.0 release notes v7
Released: 05 May 2025
AIDB 4.0.0 includes the following enhancements and bug fixes:
| Type | Description |
|---|---|
| Enhancement | Renamed retrievers to knowledge bases: The retriever pipeline is now the knowledge base pipeline. Functions aidb.create_retriever_for_table() and aidb.create_retriever_for_volume() are renamed to aidb.create_knowledge_base_for_table() and aidb.create_knowledge_base_for_volume(). The aidb.delete_retriever function is now aidb.delete_knowledge_base(). The aidb.RetrieverSourceDataFormat enum is now aidb.PipelineDataFormat. Aliases aidb.kbs and aidb.preps are added. Old names are deprecated but still available. |
| Enhancement | Renamed processing mode to auto processing: The processing_mode setting in pipeline creation, config, and overview functions has been renamed to auto_processing for consistency with the feature name. |
| Enhancement | Added ability to write preparer pipeline output to a volume: Preparer pipelines now support AIDB volumes (for example, an S3 bucket using PGFS) as an output destination. All four source/destination combinations (table-to-table, table-to-volume, volume-to-table, volume-to-volume) are now supported. |
| Enhancement | Added arm64 support for Debian 12 and RHEL 9: The AI Accelerator now supports arm64 architecture on Debian 12 and RHEL 9. |
| Enhancement | aidb.rerank_text() now returns a rich table: The function now returns its results in a table — one row per text — including columns for the raw logit_score, the text itself, and the index from the input. This makes it easier to use in larger SQL queries, for example by joining with other data or sorting by logit_score. |
| Enhancement | Improved debugging for AI models accessed over HTTP APIs: AIDB now logs the HTTP status code, error response content, and request content when encountering errors with external models. This applies to NIM models and OpenAI API-compatible models. |
| Enhancement | Added TLS configuration support for models using HTTP APIs: Models connecting to external HTTP endpoints can now be configured with TLS settings via the tls_config object. Options include disabling certificate verification using insecure_skip_verify for development, or specifying a custom certificate authority using ca_path. |
| Enhancement | Added PerformOcr operation to preparer: Preparer pipelines can now utilize a PerformOcr operation that leverages the PaddleOCR model with the NVIDIA NIM Image OCR API. |