Skip to content

Use Zensical for the Markdown manual

On 2026-10-27 the owner preferred Zensical's modern Material-style experience and delegated the final choice. After independent comparison, the project chose Zensical over Sphinx + MyST + Furo. Both met the bounded acceptance; the decision reflects the preferred navigation and presentation, not a claim that Sphinx cannot provide dark mode or render Markdown.

The comparison PR #14 retains the implementation, review and hosted evidence. Its immutable comparison report records methods, versions, adaptations and limitations. The dual builders are not carried into production; this repository maintains one generator.

Both candidates rendered the same eight existing Markdown documents, installed Click command tree, five scenario outcomes and provenance. All 24 local comparison cases passed. A fresh-output local build with dependencies cached measured Sphinx at 0.998 seconds / 1,040,683 bytes / 69 files, and Zensical at 0.420 seconds / 923,548 bytes / 28 files. Locked tool dependency closures were 37 and 12 respectively, including platform alternatives and shared dependencies, excluding application/browser tools. These are single-run observations, not a benchmark or quality score. Shared staging and package preparation were excluded.

mkdocs-click worked directly as a Markdown extension; no command-generation adapter was necessary. Relative build/source paths avoided a Zensical path invariant error. Furo needed accessible Pygments palettes after a sampled dark code contrast failure. Both provided system preference, manual theme controls, mobile navigation and working interactive search. Neither result establishes a full accessibility audit or actual Android-device compatibility.

Production keeps canonical repository Markdown as the only guide prose source, installed wheels as command truth, and packaged-container help and feature outcomes as observable evidence. The link rewriter handles the explicitly listed pages and current inline-link syntax; additions need corresponding checks. Sphinx and the comparison harness are removed from normal dependencies and CI.

The existing main-only Pages publication, freshness check inside the deployment lock, scoped permissions and 14-day review artifacts remain. Artifact expiration does not remove the hosted manual. Independent adoption review and passing CI precede merge; navigation/search must be verified through a normal browser at the canonical manual after deployment.