NVIDIA opens source auto-ontology library to fix agent data reliability
The new Apache 2.0 library builds semantic layers from enterprise metadata, letting agents verify metrics before querying databases.
Key points
- NVIDIA released Auto Ontology as an open-source library under the Apache License 2.0.
- The tool generates knowledge graphs from metadata to reconcile inconsistent data definitions across systems.
- It exposes context to agents via MCP, CLI, and API, requiring NVIDIA NIM endpoints for inference.
NVIDIA published Auto Ontology as an open-source library, according to a GitHub repository trending on Hacker News.reported The project aims to solve reliability issues in agentic analytics by providing a context layer that tells agents where data lives and what it means.This tool helps AI agents understand what data means before they use it.The library exposes a semantic catalog via MCP, CLI, and API, using NIM for inference.
What changed
The release introduces a library that builds semantic layers from existing enterprise metadata. It reconciles definitions across different systems and exposes this unified context to any agent harness. The software is distributed under the Apache License 2.0, though the project is not currently accepting external contributions.
How it works
Auto Ontology discovers enterprise ontologies from metadata and connects existing semantic models. It grounds agents by providing trusted business meaning for metrics and entities. The system validates generated queries against certified definitions and returns provenance with every answer. It serves context through Model Context Protocol, command-line interface, and API endpoints.
What the evidence means
The library complements semantic models customers already own rather than replacing them. It uses NVIDIA NIM endpoints for inference, which can be hosted via the NVIDIA API catalog or self-hosted. The defaults include the nvidia/llama-nemotron-rerank-vl-1b-v2 model for reranking tasks.reported
Cost and market position
The source notes that reliability gates agentic analytics, stating that a revenue metric answered faster is worth nothing if the number is wrong. Auto Ontology is distributed under the Apache License 2.0 and complements the semantic models customers already own. The library requires environment variables for model endpoints and API keys, tying usage to NVIDIA's infrastructure.
Why it matters
- Product teams can use the library to provide agents with trusted business meaning for metrics.
- The Apache 2.0 license allows integration with existing semantic models, as stated in the repository.
- Deployment flexibility includes Helm charts for Kubernetes or Docker Compose for local stacks.
- The system includes a "Learn" feature to fold certified corrections and usage back into the ontology.
Topics: NVIDIA, open source, AI agents, data ontology, NIM
Updated
Then / now / next
Agents could retrieve documents but lacked reliable context for metric verification.
NVIDIA provides an open library to build and expose semantic layers from metadata.
Watch for updates on external contribution acceptance or new supported inference endpoints.
Story timeline
- Then Agents could retrieve documents but lacked reliable context for metric verification.
- 03 Nov 2026 Nvidia – open-source auto-ontology
- Now NVIDIA provides an open library to build and expose semantic layers from metadata.
- Next Watch for updates on external contribution acceptance or new supported inference endpoints.
The strongest doubt
Counterpoint from the sources
The project is not accepting external contributions, which may limit community-driven improvements or bug fixes.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
Does Auto Ontology accept external code contributions?
No, the project is currently not accepting external contributions according to the repository.
What license governs the Auto Ontology library?
The library is distributed under the Apache License 2.0.
Which model is used for reranking by default?
The default reranking model is nvidia/llama-nemotron-rerank-vl-1b-v2.
Connected stories
Reading receipt
- NVIDIA released Auto Ontology as an open-source library under the Apache License 2.0.
- The tool generates knowledge graphs from metadata to reconcile inconsistent data definitions across systems.
- It exposes context to agents via MCP, CLI, and API, requiring NVIDIA NIM endpoints for inference.
