Open, curated vocabulary for language learning: German, English and Italian, CEFR-leveled from A1 to C2, with definitions, word forms, examples, idiomatic expressions and links between related words.
Browse the content live. The explorer reads the same files consumers download.
Good open vocabulary data is hard to find: sources disagree, levels are incoherent, and links between words are missing. Generating entries on demand costs money on every call and gives a different answer each time.
Lexicon writes each entry once, reviews it, and ships it as static, versioned JSON files:
- Same answer every time. Every change is a reviewed git diff.
- Leveled and linked. CEFR levels and word links are decided once.
- Cheap to serve. Consumers download the files and use them offline.
Download the distribution from the
latest release:
root_manifest.json, one index per language, and a manifest.json plus
content.json per pack. Every file carries contract_version, so a consumer
can reject a version it does not understand.
That distribution is the only contract. Read CONTRACT.md and
docs/guides/CONSUMER_GUIDE.md.
Requirements: Node 22 or later and pnpm.
pnpm test # run the tooling tests
pnpm run lexicon # open the authoring console
pnpm run release # rebuild the packs and dist/lexicon_distribution/
New entries start as needs_review. Only reviewed entries reach a pack, and
the git diff is the final check. Grammar such as German noun forms or
separable verbs (auf|stehen) lives in language plugins, so adding a
language does not change the core.
Publishing a release is a maintainer task: see docs/guides/RELEASING.md.
CONTRACT.md: the distribution contractdocs/README.md: documentation mapdocs/guides/CONSUMER_GUIDE.md: using the distributiondocs/guides/PACK_AUTHORING.md: writing entriesdocs/reference/WORKFLOW_COMMANDS.md: all commandsCONTRIBUTING.md: how to contribute
Code (tools/) is Apache-2.0. Content, packs and docs are CC BY 4.0 unless
noted otherwise. See LICENSE, docs/CONTENT_LICENSE.md, NOTICE and
ATTRIBUTION.md.