# NVIDIA/auto\-ontology

NVIDIA Auto\-Ontology: Building Semantic Reasoning Layers for Agents

- HackerByte page: <https://hackerbyte.net/en/projects/nvidia-auto-ontology-building-semantic-reasoning-layers-for-agents-1222953295>
- GitHub: <https://github.com/NVIDIA/auto-ontology>
- Programming language: Python

Auto Ontology is a Python framework for building semantic, graph\-based reasoning layers over data. It enables deterministic query generation, better context retrieval, and more explainable AI for LLM and agent workflows.

## AI-generated summary

Auto\-Ontology is a Python framework designed to create semantic, graph\-based reasoning layers over enterprise data. It aims to improve the reliability of AI agents by providing them with a deterministic understanding of data meaning, relationships, and provenance. This context layer helps agents avoid incorrect answers by grounding their queries and actions in certified business definitions.

## Key points

- Generates an ontology from database metadata, mapping business terms to columns and defining relationships.
- Enables agents to query data with deterministic context, reducing incorrect answers compared to schema\-in\-context\-window approaches.
- Provides a web UI for reviewing, editing, and certifying semantic definitions and SQL expressions.
- Includes an answerability check to determine if a question can be answered by the data before querying.
- Exposes context to agents via an MCP server, agent skills, and a REST API, supporting LLM and agent workflows.