Nvidia – open-source auto-ontology

NVIDIA has developed an open-source tool called Auto Ontology, which builds a layer of contextual understanding for enterprise data. This layer reconciles definitions across systems and is exposed to various agent harnesses through MCP, CLI, and API. The tool complements existing semantic models owned by customers.

Auto Ontology is distributed under the Apache License 2.0 and does not accept external contributions at the moment. It uses NVIDIA NIM endpoints for inference, either from the hosted NVIDIA API catalog or self-hosted NIMs. The tool can be deployed on a Kubernetes cluster, with instructions available in the DEPLOYMENT.md file.

To set up the tool, create an environment file (.env) from the template and fill in the necessary values such as Postgres credentials, DEFAULT_MODELS_API_KEY, and CONNECTION_STRINGS. The tool supports SSO for authentication and provides an API gateway for conversational requests. The web UI manages the UUID automatically, and browser and service callers should use the Next.js gateway for proper authorization.