How GenAI Data Chat works

GenAI Data Chat never lets the model invent numbers. It only explains tables your data already produced.

You upload a CSV in the browser. Column types are inferred automatically. When you ask a question, the API sends the schema and sample rows to Groq or OpenAI.

The model returns a structured JSON query plan — filters, group-bys, and aggregates. A small TypeScript engine runs that plan in memory. A second LLM call turns the result rows into a short natural-language answer.

Built with Next.js, React, TypeScript, OpenAI, and deployed on Vercel. Open it from the GenAI Data Chat icon on the home page.