@alfavit/engine

Pure, dependency-free TypeScript engine that converts Uzbek text from Cyrillic or the 1995 Latin alphabet into the 2026 Latin alphabet (sh→ş, ch→ç, gʻ→ğ, oʻ→ö). It powers alfavit.uz, the Telegram bot, the Chrome extension, the macOS app and the public API.

npm i @alfavit/engine

import { transliterate, detectScript } from '@alfavit/engine'

const r = transliterate('Ўзбекистон шаҳри')
r.text   // 'Özbekiston şahri'
r.flags  // [{ start: 3, end: 4, chosen: 'e', alternatives: ['ye'], reason: 'cyrillic-e-position' }]

detectScript("o'zbek") // 'old-latin'

API

• transliterate(text, options?) → { text, segments, flags }
  • options.source: 'auto' | 'cyrillic' | 'old-latin' (default 'auto'; the script is detected per run of text, so mixed input works)
  • options.onAmbiguity: 'first' | 'flag'
  • segments: { source, output, script, start, end }[] — one per run of text; foreign runs pass through unchanged
  • flags: { start, end, chosen, alternatives, reason }[] — positions where more than one spelling is possible. Reasons include cyrillic-e-position (е → e or ye depending on position) and cyrillic-ts (ц → s, with c as the alternative)
• detectScript(text) → 'cyrillic' | 'old-latin' | 'foreign'
• version

Design notes

Old-Latin input is normalized first: every apostrophe variant used for oʻ/gʻ (ʻ ʼ ' ‘ ’ `) is recognized. Conversion is deterministic and offline — no network, no locale data, no dependencies. ESM only.

Run the tests with pnpm test.

License

MIT © 2026 Abdumajid Rashidov