Data license

The code of Vocab Bloom Hub is released under the MIT license. The dictionary data — the entries, meanings, examples, translations and inflected forms stored in the database, exported by Export dictionary, served by the public /api/v1 and published as the HuggingFace dataset Fristail27/vocab-bloom-hub-en — is licensed separately, under the

Creative Commons Attribution 4.0 International (CC BY 4.0), SPDX identifier CC-BY-4.0.

Copyright (c) 2026 Aleksei Ryzhov.

You are free to copy, redistribute and adapt the data for any purpose, including commercially, as long as you credit the source, link to the license and indicate whether you changed anything. A sufficient attribution:

Vocabulary data from the Vocab Bloom Hub English dataset (CC BY 4.0), https://huggingface.co/datasets/Fristail27/vocab-bloom-hub-en

The same terms travel with the data everywhere it goes:

• manifest.json of every export carries license ("CC-BY-4.0") and attribution;
• GET /api/v1/meta returns license, license_url and attribution;
• the admin Export dictionary page shows them next to the download;
• the HuggingFace dataset card declares license: cc-by-4.0 and ships LICENSE and NOTICE.

The constant behind all of them is apps/server/core/constants/data_license.ts.

Why CC BY 4.0

MIT is a software license: it is ambiguous for a database of facts and says nothing about attribution of derived datasets. CC BY 4.0 is the standard choice for open lexical data — free for any use, machine-readable, and it asks for exactly the one thing the project needs, credit that leads users back to the source. Share-alike (CC BY-SA) would have kept derived datasets open but discourages embedding the dictionary in products, which is what the public API is for; CC0 would have dropped the attribution the project relies on to be found.

Provenance

Most of the data is generated by language models and is not human-verified; every record says which model produced it. What that means for using the data, the known limitations and how to report errors are in docs/data.md. Pass the notice on: GET /api/v1/meta returns it as notice, the dataset card and the word pages of the website show it.

The models are chosen so that this license can be honoured. A provider whose terms only bind the customer — DeepSeek's Open Platform terms assign the output to the user and expressly allow derivative products and training other models — is compatible with CC BY 4.0; a provider whose terms make the customer prevent third parties from training on the output (xAI's API terms for Grok, as of August 2026) is not, because CC BY 4.0 grants every recipient exactly that right. The published data is generated with models of the first kind; the check is repeated before a new model joins the pipeline. The terms of DeepSeek also ask that the DeepSeek name is used only factually, without logos or any suggestion of endorsement.

Contributions to the data

An entry edited in the admin UI, a correction sent through Report a mistake on a word page and applied by the owner, and a dataset revision contributed through the repository are licensed under the same CC BY 4.0 — the word pages say so next to the form, and CONTRIBUTING.md records it for pull requests.

Your own instance

A self-hosted instance ships the same license values in its exports and its /api/v1/meta. If you add data of your own under different terms, change DATA_LICENSE in apps/server/core/constants/data_license.ts and this file together, and make sure the terms of the data you started from are respected — CC BY 4.0 allows relicensing derived works, but the attribution to the original stays.