dhruva

The Gita online is mostly two things: a scan of a public-domain translation from 1909, or an ad-farm that wants a newsletter signup before the second verse. Neither is readable, and neither explains anything — you get the words and none of the argument they sit inside.

There is also a format problem. A text written to be read slowly is the worst possible fit for infinite scroll, and a reading app that needs a network and an account is one you can’t take to a bus or a hospital waiting room — which is when people actually reach for this book.

It’s a static site: 1,339 pages prerendered at build time, with React islands only where interaction genuinely earns one — the reader, search, the journal. The corpus lives as JSON and the search index is built during the build, held under 400 KB gzipped by a check that exits nonzero rather than warns, so search works offline from the first visit.

The palette is keyed to the time of day — dawn, noon, dusk, night — and a reader can override theme, contrast, motion, line length, emphasis and how much commentary shows. That’s eight palette configurations across every route, which is a lot of ways for contrast to quietly fail, so an audit runs axe over 13 routes in all of them and every faint ink value carries its measured ratio in a comment beside it.

Reading progress, journal reflections and the words you’ve met are stored in IndexedDB and never sent anywhere. The colophon’s claim is deliberately literal: nothing you read or write here leaves your device.

Offline took the most care. The service worker precaches the 18 chapter pages — each carrying its whole chapter, so the entire book reads offline — and deliberately not the 701 verse permalinks, which would mean a 9 MB download before anyone read a word. Those cache as they’re visited.