NVIDIA/auto-ontology
Auto-Ontology is a Python framework by NVIDIA/auto-ontology for building graph-based reasoning layers that help agents query enterprise data with fewer hallucinations. It targets developers building reliable LLM and agent workflows.
4.2k · Python
It operates as a service you run next to your databases to compile metadata into an ontology before queries happen. Deploying an extra service alongside your databases means more infrastructure to manage, but you get deterministic query generation instead of blind vector searches.