CONTRIBUTING.md

Contributing a dataset

Datasets are added by pull request.

1. Fork the repo and create a branch.

2. Create datasets/<slug>.md with this frontmatter:

   ---
   title: '<dataset name>'
   desc: '<one-sentence description>'
   thumbnail: /thumbnails/<slug>.png
   publication: <https://...>     # optional
   github: <https://...>          # optional
   domain:
     - <domain tag from src/data/tags.js>
   modality:
     - <modality tag from src/data/tags.js>
   tasks:
     - <task tag from src/data/tags.js>
   ---
   
   <full markdown description here>

3. Add the thumbnail to public/thumbnails/<slug>.png (16:9, PNG or JPG — e.g. 640×360).

4. Pick tags from the curated list in src/data/tags.js. Tags are split across three facets — domain, modality, and tasks — and each frontmatter field only accepts tags from its own facet. At least one tag is required across the three. The schema rejects anything not in the curated list; if your dataset truly needs a new tag, add an entry to src/data/tags.js in the same PR.

5. Run npm test locally — the schema validator runs as part of the test suite, so you'll see immediately if anything is wrong.

6. Open a PR. CI will re-run validation; once it's green, a maintainer will review and merge.