The data pipeline

make data needs uv. It fetches the lexicon bundle pinned in packages/data/VERSION, verifies its checksum, exports fixtures and tables from zimankit, and compiles a lexicon for each variety of about 2 MB gzipped.

The result is reproducible: packages/data/lexicon.lock.json records it, and CI fails if a rebuild differs.

scripts/content/import-zimankit-tutorials.ts imports the lessons from a zimankit checkout into packages/content.