Website quality and discoverability¶
The documentation workflow retains MkDocs Material as a strict compatibility and rollback baseline and qualifies the selected Zensical Modern build before it can be published. It keeps the Python package release, scientific validation evidence and documentation-only experimental integrations clearly separated. No site audit alters a published estimator or raw participant data.
Automated quality controls¶
| Test | Input | What it establishes | What it does not establish |
|---|---|---|---|
| Static source contracts | Markdown, local links, SVG references and JavaScript | Required pages and known dynamic links exist in the repository | That every browser can reach every external URL |
| Strict dual-builder build | Python docs, public API, deterministic gallery | The MkDocs baseline and selected Zensical Modern site build strictly | That a successful plot validates statistical inference |
| Public metadata audit | Built HTML, CFF, sitemap and robots | Correct publication version, canonical URLs and citation identity for representative pages | Search-engine indexing or rich-result eligibility |
| Playwright desktop/mobile | Built site on the real GitHub Pages subpath | Images, interactions, route loading and browser screenshots | Full cross-browser and device coverage |
| Locally installed axe-core | Representative interactive pages | Tagged WCAG 2.0–2.2 accessibility issue report | Manual accessibility or WCAG certification |
| Zensical Modern qualification | Complete rendered route inventory, native search, 16 desktop/mobile and theme views, axe-core | Tested route parity and browser behavior in audited configurations | Complete assistive-technology or WCAG certification |
The isolated accessibility runner installs axe-core only in CI and injects it into a locally served copy of the generated documentation. It is not loaded for actual website visitors. Its results identify concrete failures and issues requiring manual review; automation alone cannot demonstrate complete WCAG compliance.
Citation and discovery information¶
Software citation is based on the repository's CITATION.cff. The homepage exposes a static SoftwareSourceCode JSON-LD description aligned with version 1.2.0, the MIT license and the official GitHub repository. A robots.txt file points to the documentation builder's generated sitemap.
These additions make source identification and crawling clearer. They do not guarantee ranking or enhanced search results. No DOI is present in the current CFF, and the page does not assert one.
Accessibility issues discovered and addressed¶
The first optional axe-core run flagged an unnamed Material loading progressbar, top-tab text just below the 4.5:1 contrast threshold in the light indigo theme, and documentation links distinguishable only by color. A shared minimal CSS/JS refinement now adds a progressbar accessible name, more legible navigation/footer text and underlined inline documentation links. These are specific remediations, not a declaration of complete accessibility.
The browser report retains all automated findings with their node selectors and severity. The optional Chromium workflow now fails its accessibility step for any serious or critical WCAG-tagged findings in the sampled routes. It is non-deploying and remains separate from the guarded documentation publication gate. Zero findings on these routes still requires human keyboard and screen-reader testing.
Regression governance¶
The site-quality and strict documentation gates must succeed before merging a website PR. Browser and alternate-generator previews provide independent audit artifacts. The selected Zensical Modern build is gated on URL and fragment parity, native search interaction, desktop/mobile rendering and high-impact automated accessibility findings. Human screen-reader, keyboard and mobile assistive-technology review remains open in issue #499; automated CI does not certify WCAG conformance.
Visual explorer · First analysis · Citation and research reporting · Zensical preview