nla-lib.org

The NLA-lib website, deployed to GitHub Pages by .github/workflows/deploy.yml.

• Homepage (index.md) and Design page (design/index.md, how NLA-lib defines things): Markdown converted by python3 md2html.py index.md design/index.md (needs pandoc). Commit the .md and the .html. A page's menu is the menu: list in its front matter; a page with mathematics loads KaTeX.
• Explorer (explore/, served at nla-lib.org/explore/): the static viewer of the NLAlib library, built from NLAlib's extracted data at every deploy. See explore/README.md. It was the repository NLA-lib/explore until 2026-11-06; its history is kept here.
• Shared by both halves, one copy each: css/site.css (fonts, links, the reading column, the title and the menu, so the header does not move between pages), font/ and vendor/katex/. The explorer's build copies them into its output.

Every menu must list the same pages; explore/tests/test_menu.py checks it. Run the explorer's tests with cd explore && python3 -m unittest discover -s tests.

Explorer preview

The Explorer is authored in Markdown files, with YAML metadata and generated sections. It opens with Algorithms, Theory, and the full Library reference. Tasks and method families are discovery pages; mathematical algorithms, implementations, and theory topics have canonical identities. See the architecture.

Install the page-rendering dependencies with python3 -m pip install -r explore/requirements.txt. The regular build turns the generic extraction into library records and Markdown pages, without the retired program/target pipeline. Matching published core sources automatically fill the CG implementation pages. Use --core-src <NLAlib> to preview a newer working core; unfinished pages elsewhere remain placeholders.

Build the Explorer into _site/explore using explore/build.py and matching published data, then add the homepage and shared assets as in the deployment workflow. Run python3 preview.py from this repository root (with --port 8081 for the current demo).