[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"readme:auto-ontology":3},"\u003Ch1>Auto Ontology\u003C\u002Fh1>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fbeaverbench.github.io\u002F#leaderboard\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBEAVER_leaderboard-%231-76b900\" alt=\"#1 on the BEAVER leaderboard\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fbird-interact.github.io\u002F\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FBIRD--Interact_leaderboard-%232-76b900\" alt=\"#2 on the BIRD-Interact leaderboard\" \u002F>\u003C\u002Fa>\u003C\u002Fp>\n\u003Cp>Agents can retrieve documents and query databases, but reliability gates agentic analytics: a revenue metric answered faster is worth nothing if the number is wrong. Agents need context to answer accurately: what each business concept means, where it lives, how it connects, how to calculate it, where it comes from, and whether to trust it. An ontology holds that context: a shared, governed map from the concepts people ask about to the data behind them. NVIDIA Auto Ontology is an open library that builds that ontology automatically from enterprise metadata, reconciles definitions across systems, and exposes it to any agent harness through MCP, CLI, and API. It complements the semantic models customers already own, and \u003Ca href=\"https:\u002F\u002Fossie.apache.org\u002F\" rel=\"nofollow ugc noopener\">Apache Ossie\u003C\u002Fa> keeps it portable.\u003C\u002Fp>\n\u003Cp>It enables developers to:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>Discover:\u003C\u002Fstrong> generate enterprise ontologies from metadata; connect existing ontologies and semantic models; build AI-ready knowledge graphs.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Ground:\u003C\u002Fstrong> give agents trusted business meaning for metrics, entities, and relationships; reconcile inconsistent definitions across systems.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Validate:\u003C\u002Fstrong> test a draft with real questions before experts certify it; check generated queries against the catalog and the definitions retrieval surfaced; return the SQL and steps with every answer.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Expose:\u003C\u002Fstrong> serve context to any agent harness through MCP, CLI, and API.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Learn:\u003C\u002Fstrong> fold certified corrections and usage back into the ontology, so shared context improves with use.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>An ontology matters most when:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>questions are varied, open-ended, and ambiguous;\u003C\u002Fli>\n\u003Cli>data spans many disparate sources with overlapping or conflicting\ndefinitions;\u003C\u002Fli>\n\u003Cli>meanings must be reused and governed across runs, teams, and agents;\u003C\u002Fli>\n\u003Cli>the model or its context budget can't hold the full documentation;\u003C\u002Fli>\n\u003Cli>users ask in business terms and need routing to the right source and grain.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>With one well-modeled source and a fixed set of questions, direct SQL may be\nenough.\u003C\u002Fp>\n\u003Ch2>Quickstart\u003C\u002Fh2>\n\u003Cp>This runs Auto Ontology with Docker Compose against Wide World Importers,\nMicrosoft's sample sales database. You need Docker with Compose v2,\n\u003Ca href=\"https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002F\" rel=\"nofollow ugc noopener\">uv\u003C\u002Fa>, and an NVIDIA API key (\u003Ccode>nvapi-...\u003C\u002Fcode>) from\n\u003Ca href=\"https:\u002F\u002Fbuild.nvidia.com\" rel=\"nofollow ugc noopener\">build.nvidia.com\u003C\u002Fa>.\u003C\u002Fp>\n\u003Col>\n\u003Cli>\u003Cp>Clone the repository and create \u003Ccode>.env\u003C\u002Fcode>:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-bash\">git clone https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fauto-ontology.git &amp;&amp; cd auto-ontology\ncp .env.example .env\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>Then set these values in \u003Ccode>.env\u003C\u002Fcode>:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-dotenv\">DEFAULT_MODELS_API_KEY=nvapi-...\nAUTH_SECRET=&lt;a long random string, e.g. from openssl rand -base64 32&gt;\nAPP_URL=http:\u002F\u002Flocalhost:3000\nPOSTGRES_PASSWORD=&lt;a new password&gt;\nCONNECTION_STRINGS=postgresql:\u002F\u002Fpostgres:&lt;the same password&gt;@postgres:5432\u002Fwideworldimporters\nAUTO_ONTOLOGY_ADMIN_PASSWORD=&lt;a new password&gt;\nPGADMIN_PASSWORD=&lt;a new password&gt;\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>Compose publishes Postgres, pgAdmin, and the web app on the host, so replace\neach default password before you start it. \u003Ccode>.env.example\u003C\u002Fcode> also sets an API\nkey per model role: replace each \u003Ccode>nvapi-your-api-key-here\u003C\u002Fcode> with your key, or\ndelete those lines so they fall back to \u003Ccode>DEFAULT_MODELS_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n\u003C\u002Fli>\n\u003Cli>\u003Cp>Start Postgres and load the sample database. The seed script lives in\nauto-ontology-eval; pass it the Postgres password you set:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-bash\">docker compose up -d --wait postgres\ngit clone https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fauto-ontology-eval.git ..\u002Fauto-ontology-eval\n(cd ..\u002Fauto-ontology-eval &amp;&amp; uv sync &amp;&amp; cp .env.example .env &amp;&amp; \\\n  POSTGRES_PASSWORD='&lt;the same password&gt;' \\\n  uv run python scripts\u002Fse\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003C\u002Fli>\n\u003C\u002Fol>\n",1795123396686]