Search

Search is enabled by default with the built-in compact provider. Small and medium sites need no additional configuration.

How compact matching works

compact lowercases contiguous Latin letters and digits into word tokens. For contiguous Han text, it emits each character and every adjacent two-character token. For example, 组件文档 yields 组, 件, 文, 档, 组件, 件文, and 文档. Queries use the same tokenizer.

The index includes titles, descriptions, tags, the full searchable body text, and bounded code samples. Ranking favors complete phrases and tokens in titles, followed by descriptions, tags, and route paths; a body-only mention can still find a page without outranking a stronger title match. Results display the page description rather than a snippet from the matching body position. The UI highlights complete queries and useful matching tokens in titles and descriptions. This keeps the index small and browser-side dependency-free, but it does not provide fuzzy matching or match-position snippets. Use Pagefind when a large site needs more mature tokenization and chunked loading.

Use Pagefind

For larger sites, install the official plugin package and Pagefind:

pnpm

pnpm add -D @canofold/plugins pagefind

npm

npm install --save-dev @canofold/plugins pagefind

yarn

yarn add --dev @canofold/plugins pagefind

Then configure the search provider:

canofold.config.ts

import { pagefind } from '@canofold/plugins'import { defineConfig } from 'canofold'export default defineConfig({  search: { provider: pagefind() }})

Pagefind indexes final HTML and filters results by the active locale and version.

Exclude content

Set search: false in page frontmatter to exclude one page:

yaml

---search: false---

Use search: { enabled: false } to disable search for the entire site. Private sites should protect search indexes with the same access policy as page content.