nooher/kasuku-lugha

Sovereign Swahili language primitives — pronunciation, morphology, read-aloud scoring. Zero-dependency, offline. By Laetoli (T) Limited. # @laetoli/lugha **Sovereign Swahili language primitives.** Zero-dependency, offline, deterministic building blocks extracted from the Kasuku platform so any product can embed Tanzania's language layer instead of a foreign cloud LLM. - **Pronunciation** — `phonemize` (syllables, IPA, ASCII respelling) + `transliterate` - **Morphology** — `conjugate` / `paradigm` (verb across persons & tenses) + noun classes - **Read-aloud scoring** — `scoreReading` (phoneme-aware, per word: good / close / miss) - **Editorial** — `audit` (clarity, style, repetition, show-don't-tell, Kiswahili register, 0–100 score) - **Poetry** — `analyzePoem` (arudhi: mizani + vina/rhyme scheme) - **Dictionary (Kamusi)** — `define` / `toSwahili` / `toEnglish` (1,414 meanings, WOLD CC-BY) No model, no network, no API keys. Standard Swahili is ~95% phonetically regular and agglutinatively regular, so these are encoded as rules. > The catalogue-bound features (book discovery, literary study guides) and the > neural voice / ASR live in the Kasuku app — they need its data and a browser. > This package is the pure, portable core. ## Install ```bash npm install github:nooher/kasuku-lugha ``` ## Use ```ts import { Lugha } from '@laetoli/lugha'; Lugha.phonemize('karibu'); // { word:'karibu', syllables:['ka','ri','bu'], ipa:'/kɑˈɾibu/', respell:'ka-REE-bu', stress:1 } Lugha.transliterate('asante sana'); // smooth form a generic TTS voice reads as Swahili Lugha.conjugate('soma', 'present'); // [{ pronoun:'mimi', form:'ninasoma', ... }, ...] (6 persons) Lugha.conjugate('soma', 'present', true);// negative: 'sisomi' ... Lugha.scoreReading('asante sana', heardText); // { overall, words:[{word,status,sim,syllables}] } ``` Named exports are also available: `phonemize`, `ttsText`, `paradigm`, `conjugate`, `scoreReading`, `NOUN_CLASSES`, `LUGHA_VERSION`, plus all the types. | Function | Returns | |---|---| | `phonemize(text)` | `{ syllables, ipa, respell, stress }` | | `transliterate(text)` / `tts …