Congeegator & Lexicoff

Two Wiktionary-powered language apps:

• Congeegator — verb conjugation tables with downloadable languages for offline use.
• Lexicoff — an offline dictionary with definitions, inflected forms, pronunciation and etymology.

How they work

Both apps use SvelteKit, Svelte and TypeScript, run as installable progressive web apps, and are hosted on Cloudflare Workers. A shared Python pipeline processes English Wiktionary data from kaikki.org, with generated datasets distributed through Cloudflare R2.

Data delivery has two purposes: small letter-based chunks for efficient server-rendered SEO pages, and complete language bundles for client apps. Congeegator's Cloudflare Worker fetches the relevant letter chunk to render a page for visitors and crawlers. Congeegator automatically syncs the selected language as a full bundle, stored in IndexedDB using Dexie. Its browser page loader can also fetch an individual letter chunk when a verb is unavailable locally.

Lexicoff lets users manually download one Zstandard-compressed SQLite database per language. SQLite runs in the browser through WebAssembly, reading downloaded files from OPFS, the browser's private filesystem. FTS5 indexes support prefix, inflection, definition and phonetic searches, with a trigram index for fuzzy matching. Separate background workers handle downloads and queries. Once a language is installed, lookups work offline.

Repository

apps/congeegator/   Conjugation app
apps/lexicoff/      Offline dictionary
pipeline/          Shared data extraction and database generation

Each app has its own npm dependencies and independent language catalogue.

Development

Install dependencies and start either app from its directory:

cd apps/lexicoff # or apps/congeegator
npm ci
npm run dev

Local dictionary/conjugation data must be generated separately. From the repository root, use Pixi to generate a development subset:

pixi run generate-dev

The configured Pixi environment currently targets Linux. Development apps serve generated files from their r2_data/ directory through /dev-r2-mock.

For full datasets, run pixi run generate. To generate just Lexicoff:

pixi run python -m pipeline.generate --app lexicoff

Checks

From either app directory:

npm run check
npm run lint
npm test
npm run build:staging

From the repository root:

pixi run test
pixi run python -m pipeline.generate --check-manifest-metadata

See AGENTS.md for architecture details, source-pinning instructions and deployment behavior.