NVIDIA/auto-ontology
- Stars
- 10,343
- Language
- Python
- License
- Apache-2.0
- Weekly
- rank 1 · +9,044 stars gained
Repository metrics were observed on this edition date and may have changed.
Key Details Auto Ontology is a library and web app that builds business terminology, SQL properties, and table relationships from database metadata and executes natural-language questions as SQL. Its distinguishing features are a human review and certification stage for definitions and a workflow that displays SQL and processing steps alongside answers. Use Cases Connect a sample sales database, then ask about sales for a particular product, region, or period and inspect the SQL and processing steps together. Result rows can be downloaded as CSV or viewed as charts. Costs and Limitations Apache-2.0 code and a self-deployment configuration are provided. It uses the hosted NVIDIA NIM API or self-hosted NIM and requires an API key. Hosted API pricing and self-hosted NIM terms need to be checked; server, database, and self-hosted inference infrastructure operating costs are separate. Adoption Considerations The quick start requires Docker Compose v2, uv, and an NVIDIA API key. Database connections, passwords, authentication settings, and terminology certification require configuration or work, and external contributions are currently not accepted. The latest supplied release is a backup tag registered on 2026-09-01 KST, so the process for managing formal releases needs to be checked. Why It Stands Out It ranked first on weekly Trending, with an observed increase of 9,044 stars during that period. The described workflow goes beyond query generation to certify metric definitions and distinguish questions that cannot be answered from the data.
View on GitHub