Search

Every open-docs site ships with full-text search, powered by Pagefind. The index is built from your pages at deploy time and served as static files — there's no search server, and it keeps working offline. Open it from the box in the top bar, or with ⌘K / Ctrl K.

What gets indexed

Only the body of each page is indexed. The chrome around it is left out on purpose, so it doesn't pollute results:

• the sidebar, top bar, and footer;
• the prev / next links (they just repeat the neighbouring pages' titles);
• the on-page table of contents;
• the "Rendering diagram…" placeholder for Mermaid blocks.

How results are ranked

open-docs tunes Pagefind for documentation out of the box — there's nothing to configure:

• Headings outrank body text. A term in a page title or heading counts far more than the same word buried in a paragraph, so the most relevant page surfaces first (titles weigh most, then ##, then ###).
• Exact matches win. Matches closer to what you typed are favoured over fuzzy or partial ones — documentation searches are usually precise.
• Long pages aren't penalised. A thorough page won't lose to a stub that happens to mention the term once.

If you fork open-docs, these knobs live in src/lib/components/search.svelte.

Pushing a passage higher

When the heading weights aren't enough — a key definition sits in the middle of a long page, say — wrap it in a {% boost %} block to weight it more heavily in the index. It renders unchanged; it only affects search.

{% boost weight=8 %}
The container builds the site at start, so one image serves any docset.
{% /boost %}

See the Markdoc tags reference for the weight scale.