Hey HN, maker here. I'm a solo dev who builds side projects and this one started because I thought the gap between "I need a chart" and actually having one was still way too wide for most people. Journalists, teachers, marketers - they shouldn't need to find a dataset, clean it, and learn a charting tool just to get a simple visual.

The interesting technical bit is the two-step flow. When you type a query, it first uses Gemini with Google Search grounding to find and validate the data, then a second call generates the ECharts config. I went with this approach instead of a single prompt because it made the data sourcing much more reliable. Caching common queries also keeps costs down. The screenshot feature uses Gemini's vision capabilities to extract the data from chart images and recreate them. Useful for pulling editable versions out of PDFs and reports. Stack: Next.js, TypeScript, Supabase, Vercel, Gemini API, ECharts. Some queries worth trying:

"S&P 500 performance over the last 10 years" "global CO2 emissions by country" "UK house prices since 2000" "programming language popularity"

Would love feedback on where the NL parsing breaks for you.