Benchmarks¶

Page-weight receipts for the themes Void competes with. The same four Markdown pages (home, a nested guide with a code block, and a reference table) are built with void, material, and readthedocs; the table below is the byte count shipped by each build. Smaller is faster — and this page is regenerated by CI on every push to main.

Measured with MkDocs 1.6.1 at commit f5516ac.

Methodology¶

• Each project uses the same minimal content and a search plugin; only the theme differs (void additionally registers the void plugin, the same way a real Void site does).
• “Total shipped (raw)” sums every file emitted under site/ (HTML, CSS, JS, fonts, and static assets). “Over the wire (gzip)” is the same set compressed with gzip to approximate transfer size.
• Themes that cannot be built on a machine (for example material when mkdocs-material is not installed) are shown as “—”; the CI workflow installs all three, so the table is always complete upstream.
• No network requests happen during measurement — CDN-only assets are counted by what the theme itself ships to the browser.

Lighthouse¶

Lighthouse runner is enforced in CI, not measured here: .github/workflows/performance.yml asserts a minimum score of 95 in every category (performance, accessibility, best-practices, SEO) against the live docs site. A run that scores below the bar fails the build. Latest runs: Actions → Performance.

Regenerating¶

python tools/emit_benchmarks.py

The Benchmarks CI workflow runs the same command on every push to main and opens a pull request when the numbers change. To regenerate manually, add mkdocs-material to the environment (pip install mkdocs-material) so the material column is populated.