Features overview
Docsense adds local documentation search to Docusaurus: a navbar SearchBar, a build-time index, and in-browser queries against a static JSON file you own.
This page lists what ships today, then the planned work on the public roadmap.
Available now
Search from the navbar
Docsense registers a SearchBar in the Docusaurus navbar. Readers open it from the trigger control or with:
- macOS: ⌘K
- Windows and Linux: Ctrl+K
You can change the trigger label, placeholder, shortcut, and empty-state copy. See Customize the search bar.
Build-time indexing
Docsense does not crawl your published site. On docusaurus start and docusaurus build, the plugin scans your Markdown and MDX, extracts searchable text (including many MDX string attributes), and writes static/docsense/search-index.json.
The index deploys with the rest of your static assets, so search freshness matches the build you shipped.
Section-level results
Results can deep-link to a heading. Each hit includes breadcrumbs and a short snippet so readers can judge relevance before they click.
Filters and page limits
The modal can filter by page type (How-to, Concept, Reference, and related types inferred from paths or frontmatter).
The default page limit is 8. Readers change it with Pages − / +. The control stops at 3 on the low end and 20 on the high end; those buttons disable at the edges. See Use filters and page limits.
Typo tolerance and stemming
Queries use exact, prefix, and fuzzy matching, plus English stemming so forms such as installing and install connect. Details live in Ranking and typos and MDX extraction and stemming.
Highlight on the destination page
After a reader opens a result, Docsense can mark matching terms on the destination page so the jump from search to content stays oriented. See Highlight terms on the page.
Custom look and feel
Theme the SearchBar with CSS variables (colors, radius, modal chrome, and mark styles). This site’s deep green and gold theme is documented with a worked example in Style the search UI.
Feature roadmap
The following capabilities are planned. They are not available in the current release.
| Capability | Direction |
|---|---|
| Sharded indexes | Split large indexes so the browser can load only the fragments a query needs. Intended for documentation sites whose single JSON index becomes too large to download up front. |
| Broader language support | Expand tokenization and stemming beyond the current English-first path, including CJK and additional locales. |
| Optional Ask AI integration | Support partner- or self-hosted answer experiences on top of the existing search index, without requiring a Docsense-hosted model service. |
Sharded indexes
Today Docsense ships one static JSON index per site. That model is simple, inspectable, and a strong fit for typical documentation sets.
Sharding is the roadmap path for very large corpora: a new index layout, a loader that fetches fragments on demand, and caching so repeat queries stay fast. Until that ships, prefer Docsense for sites where a single index download is acceptable. For multi-site HTML at extreme scale, Pagefind remains a common alternative. See Compare open-source Docusaurus search.
Next steps
- Install Docsense - Add the plugin and rebuild once.
- How Docsense works - Build-time index and in-browser search flow.
- Use filters and page limits - Page-type chips and the 3 to 20 page control.
- Compare open-source Docusaurus search - How Docsense relates to other search options.
- FAQ - Common product questions.