Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

NVIDIA open-sources Auto Ontology for governed NL-to-SQL

AI summary

The supplied reports describe NVIDIA’s open-source Auto Ontology as a semantic catalog and business-ontology system for natural-language SQL generation over structured data. It ingests database metadata to draft terms, column mappings, relationships and reusable SQL attributes, which users can revise and certify before query agents use them to retrieve context, generate and check SQL, and execute queries. The project exposes Web, authenticated REST API, CLI and MCP interfaces. The reports caution that generated definitions require business-owner approval, final SQL is still model-generated, and successful execution does not guarantee a correct business answer. They also note validation, retry and fallback paths, meaning returned SQL should not automatically be treated as a successfully executed or authoritative result.

In-depth insight

Auto Ontology places reviewed business definitions, column mappings, join paths and reusable SQL expressions into an observable agentic query workflow instead of relying on database schemas alone. This can reduce repeated interpretation of business metrics, but the reports also emphasize that its value depends on terminology quality, human certification and careful evaluation of failure paths.

日期:2026-11-05 GitHub Trending 日榜第 1 位;榜单采集时间:16:00:25(北京时间)。

“本月收入是多少?”看起来只需要一个 SUM。接到真实业务数据库,问题很快就变了:订单金额还是实收金额,取消订单算不算,时间按创建还是支付,金额存的是元还是分?SQL 能运行,与回答了正确的问题,是两件事。

NVIDIA 的 Auto Ontology 把一部分工作放到提问之前。它读取数据目录,生成映射到物理列的业务术语、关系和 SQL 属性,让人修订这些定义,再让查询 Agent 从这份语义上下文中选材料、写 SQL、检查并执行。值得看的是这些材料怎样进入执行链,以及检查失败以后系统究竟怎么做。

📋 项目概览

项目内容
主仓库NVIDIA/auto-ontology
定位面向结构化数据问答的语义目录、业务本体与 NL-to-SQL 系统
核心实现Python 后端、LangGraph 查询编排、Next.js 前端、Postgres 与 pgvector
后端清单版本1.1.0,Python 要求为 3.12 至 3.13
对外入口Web、经过认证的 REST API、单独发行的 MCP 服务
部署方式Docker Compose 或 Kubernetes Helm
许可证Apache-2.0;第三方组件另见仓库的许可说明
快照热度5,898 Stars、461 Forks、46 个 open issues

#NVIDIA#Auto Ontology#PostgreSQL#pgvector#LangGraph#Next.js#FastAPI#NVIDIA NIM

Original · 18 min readRead the original
Group chat

No comments yet. Start the conversation.