Documentation and GitHub Pages¶
The site uses MkDocs and the free, open-source Material theme. It requires no paid theme, analytics service or external font service. Python documentation dependencies are separate from the audio package.
Local preview¶
From the repository root:
Open http://127.0.0.1:8000. If the research API is already using that port:
Build the deployable static site and validate links:
Output goes to site/, which is ignored by Git. Edit Markdown in docs/, navigation in mkdocs.yml
and the small visual overrides in docs/stylesheets/extra.css.
Publish¶
- In the repository’s Settings → Pages → Build and deployment, select GitHub Actions as the source.
- Push the documentation changes to the default branch (
masterin this checkout). - The Documentation workflow validates and builds the site, uploads the Pages artifact and deploys
it using the
github-pagesenvironment. Pull requests build without deploying. - The expected URL is https://pawel-kaczmarek.github.io/The-A-Files/.
The workflow can also be started manually from the Actions tab. Publication requires repository access
and an enabled Pages environment. If the owner or repository name changes, update site_url, repo_url
and the README documentation link before publishing.
See GitHub’s custom-workflow instructions and Material’s publication guide.
Maintaining scientific descriptions¶
The method, attack and metric pages and the catalogue lists in the README are generated from the card
of each class (see adding components). Edit the card next to the implementation, never
the generated sections between <!-- catalogue:... --> markers, then regenerate:
python -m taf.catalogue_docs # rewrite the generated sections
python -m taf.catalogue_docs --check # exit 1 when they are out of date
tests/test_catalogue.py runs the check, so CI fails when a card changed without regenerating.
The prose around the generated sections is written by hand. For each addition, document the mechanism,
identifier, limitations and source. Distinguish paper algorithms, local adaptations and wrappers over
released models. Cite a paper’s section or experimental conditions when quoting results; do not imply
those results were reproduced locally.