// the find
lux-org/lux
Automatically visualize your pandas dataframe via a single print! 📊 💡
Lux hooks into pandas so that just printing a dataframe in Jupyter triggers a set of automatically generated charts instead of nothing. It's for people doing exploratory data analysis who want to skim a dataset's shape without writing matplotlib/seaborn boilerplate for every column pair.
The API is genuinely clever: no new verbs to learn, you keep writing normal pandas and the widget just shows up. The Enhance/Filter/Generalize split for next-step recommendations is a real information architecture, not a random grid of plots — it gives you a reason to look at each chart. You can export any generated Vis to actual Altair, Matplotlib, or Vega-Lite code, so you're not stuck inside the widget once you find something worth keeping. It also has a SQL executor path alongside the Pandas one, so it's not strictly limited to data that fits in memory.
No commits since March 2024 — this is an unmaintained project at this point, and a library this dependent on Jupyter widget internals rots fast as notebook/lab versions move on. Setup is brittle by its own admission: separate nbextension/labextension install steps, version-pinned to specific jupyterlab-manager releases, and the README states it's 'only been tested with Chrome.' The 'intelligent' part of the recommendations is a fixed interestingness heuristic with no real way to tune or explain why a chart was ranked above another, which matters once you stop trusting it blindly on wider dataframes. It's also Jupyter-notebook-only — there's no story for scripts, pipelines, or anything outside that one environment.