Show HN: Open-source AI data generator (now hosted) (www.metabase.com)

An open-source AI data generator — now hosted — lets you turn a few dropdowns into realistic, business-ready datasets in minutes for demos, tests, and dashboards without touching real customer data. The app previews a schema and 10 sample rows after you pick an industry and parameters, then exports full datasets as CSV or SQL or launches a dockerized Metabase instance on demand so you can explore the data visually. The UX uses Next.js, Tailwind CSS, and ShadCN UI; the repo is public so you can star, fork, or contribute. Technically, the generator uses a two-stage pipeline: an LLM (OpenAI by default, or local LiteLLM) composes detailed data specifications for your chosen business domain, and those specs drive local data synthesis (via Faker and local tooling) to produce unlimited, plausible records. This keeps sensitive production data off-cloud when using LiteLLM locally, while the Metabase-in-Docker option makes it trivial to spin up/tear down analytics for quick validation. For AI/ML practitioners and product teams, that means faster, safer prototyping, reliable testing with realistic schemas, and reproducible synthetic datasets you can inspect before exporting or connecting to downstream workflows.