975 names for 65 sky objects, in 57 languages. What speakers of the Indian languages call a star, an asterism or a constellation — each name quoted from the dictionary or ethnography it comes from, with the page cited.

Every label reads the Indian name first, then the English one. There is no single local name — that is what this database is about — so the label follows your filters rather than picking a language by fiat. Choose a language and every label becomes that language's own word for the object. With no language chosen the label is the pan-Indian form the object carries in the Sanskrit database, which is the layer most of these languages share in adapted shape; switch borrowed off and every label turns vernacular instead, which is the most useful thing that control does. Where a label is one language's word rather than the shared form, the tooltip and the panel name that language, so that nobody's word is passed off as everyone's. The Indian names button turns the whole behaviour off.

The backdrop is the real sky to magnitude 4 — 519 naked-eye stars, drawn larger and brighter the brighter they are, so that a marker can be seen to sit on the star it names. The scale is gentle rather than photometric: at true flux ratio Sirius would swallow its neighbours. Positions are Hipparcos, from the same snapshot the constellation figures in docs/constellation-lines/ are keyed to.

Marker area is the number of languages that name the object, recomputed as you filter. Colour is what kind of thing it is, which does not change when you filter. The Milky Way is drawn as the galactic band rather than a point, because that is where it actually is.

Register is the point of the database. Most Indian languages took the 27 Sanskrit nakshatra names and adapted them; a table of those adaptations would be large and nearly uninformative. The tags separate that borrowed layer from names each language built itself — 754 of 975 names are not borrowed. Filter to tribal alone to see a sky organised on entirely different principles.

Sources are public-domain lexicography and ethnography, chiefly the Digital Dictionaries of South Asia and archive.org — Molesworth 1857, Platts 1884, Kittel, Brown 1852, Gundert 1872, the Madras Tamil Lexicon, Praharaj, Maffei 1883, Hoffmann's Encyclopaedia Mundarica, Bodding, Rivers 1906 on the Toda, Man 1883 on the Andamans. Every quotation was re-fetched and matched against its source. Where a source is still in copyright the finding is paraphrased and no quotation is shown.

Nothing was back-transliterated. Where a source printed only a romanization, no script is given — much 19th-century Indic type survives in these scans only as unreadable OCR, and several entries were read off page images instead.

Full database, with every caveat and the negative findings: docs/star-names-local/ in the repository. The separate Sanskrit-text database is docs/star-names/.