tinyfish-io/bigset — reverse-engineered prompt

Build me a web app where I can type a plain English request for a dataset, like companies hiring engineers or current GPU prices, and the app figures out the table columns, researches the live web, checks sources, removes duplicates, and gives me a clean spreadsheet style table.

I want users to sign in, create datasets, watch progress while the research agents work, browse the finished rows, and download the results as CSV or XLSX. Let me choose how often a dataset refreshes, such as every 30 minutes, every few hours, daily, or weekly, so the data stays up to date without me rerunning it manually.

Use TinyFish for web search and page fetching, OpenRouter for the AI calls, and Clerk for accounts. It should run locally with Docker and a simple dev command, with a frontend, backend, database, and any worker needed for scheduled refreshes. Keep the first version practical and expect some rough edges. Look up current docs online if you need to.