Load an Apache Parquet file — or pick a public dataset — and the charts build themselves: an instant auto-generated dashboard, with a full custom builder (chart type, X/Y fields, series, aggregation) one click away. Everything renders locally in your browser and nothing is uploaded.
Drop a .parquet file — or pick a dataset — and the page classifies every column, then builds an instant dashboard: up to eight auto-detected charts (a time series when a date or year column exists, category bars, distributions, a scatter for correlated measures, and more) plus dataset stats. Identifier-like and mostly-null columns are skipped automatically.
Click any panel — or switch to the custom chart — to take over: choose the chart type (bar, line, area, scatter, histogram, pie, heatmap, or box plot), map columns to the X axis, Y axis, and an optional series, and pick an aggregation (count, sum, average, min, or max). The chart redraws instantly as you change the configuration.
No file handy? The picker features curated open datasets — from the Hugging Face Hub and Our World in Data — that load straight from their source into your browser only when you click, plus eight built-in synthetic samples, each tuned to show off a chart type. None of the synthetic data is real.
Download the finished chart as an SVG vector file or a PNG image. Both are produced locally in the browser — there is no server-side rendering. The chart configuration is mirrored into the URL fragment, so a link reproduces the same chart.
The Parquet file is decoded in your browser via WebAssembly and every aggregation and chart is computed locally — nothing is uploaded. The URL fragment never reaches a server.