Auto Ontology

Overview

Auto Ontology is an open Python library that builds a context layer from enterprise metadata, reconciles definitions across systems, and exposes the resulting semantic, graph-based reasoning layer to agent harnesses through MCP, CLI, and API. It complements semantic models customers already own and supports discovery, grounding, validation, exposure, and learning.

Key Features

• Natural-language querying of structured data translates questions to SQL and executes them against connected databases, powered by NVIDIA NIM.
• Semantic catalog and ontology layer over relational sources, stored in Postgres alongside the catalog it describes.
• Authentication and RBAC using Better Auth.
• Web UI for chat, analysis, data catalog, analytics, and settings.
• Flexible deployment through a Helm chart for Kubernetes or Docker Compose for a local stack.
• Context exposure to any agent harness through MCP, CLI, and API.

Use Cases

• Grounding LLM and agent workflows with trusted business meaning for metrics, entities, and relationships.
• Reconciling inconsistent definitions across systems so agents use shared, certified context.
• Validating generated queries against certified definitions and lineage, with provenance returned with every answer.
• Running natural-language analytics over connected databases such as Databricks, Postgres, Snowflake, or DuckDB.

Getting Started

• Clone the repository and create a .env file from .env.example, then fill in required values such as Postgres credentials, DEFAULT_MODELS_API_KEY, and CONNECTION_STRINGS.
• Set AUTH_SECRET and APP_URL, which are required.
• Build the images and start the local stack with Docker Compose.
• Open the UI at localhost:3000; the backend API runs on port 3001 and pgAdmin on port 5050.

Deployment & Requirements

• Local deployment requires Docker with Docker Compose v2, or Kubernetes using the provided Helm chart.
• An NVIDIA API key is required for NVIDIA NIM chat and ingestion.
• Connection details are required for source databases to query, such as Databricks, Postgres, Snowflake, or DuckDB.
• Postgres with pgvector provides app metadata, the catalog and ontology, and the vector store.

Before You Adopt

• License: Apache-2.0. Review its terms before using, modifying, or distributing the project.
• Vault is off by default; a partial configuration is ignored with a warning and falls back to plaintext storage for connection credentials.
• The FastAPI port is a trusted internal service boundary and must not be exposed publicly because its identity header is not independently verified by FastAPI.
• Auto Ontology uses NVIDIA NIM endpoints for inference, which is an external dependency.
• The project states it is currently not accepting external contributions.