docusaurus plugin-content-docs Error: plugin not found build failed

Not specified by the original author.

Fix Docusaurus plugin-content-docs "plugin not found" failures

TL;DR

The site config references the docs plugin but Node cannot resolve it: either the package was never installed, or it is listed manually in plugins while preset-classic already bundles it. Install the package (or remove the redundant manual entry), then rebuild. The build passes because the plugin name now resolves to real code.

The error

Error: plugin not found: @docusaurus/plugin-content-docs
    at loadPlugin (node_modules/@docusaurus/core/lib/server/plugins/index.js)

Steps

1. Check whether it is installed: npm ls @docusaurus/plugin-content-docs. Expected: either a version tree or an '(empty)' / not-found note.

2. If it is missing, install it: npm install @docusaurus/plugin-content-docs. Expected: the package appears in package.json dependencies and node_modules.

3. If the site uses @docusaurus/preset-classic, do NOT also list the plugin in the plugins array of docusaurus.config.js - the preset already includes it. Remove the manual entry instead of installing. Expected: the plugin is provided exactly once: by the preset or by the manual entry, not both.

4. Check the plugins array for typos in the package name. Expected: the name matches the installed package exactly.

5. Rebuild: npm run build. Expected: the plugin loads and the docs routes build with no plugin error.

Use this when

• the build fails naming @docusaurus/plugin-content-docs as not found
• a Docusaurus site scaffolded by hand rather than the CLI template
• the plugin was removed from package.json during a dependency cleanup

Not for this skill when

• 'broken markdown link' (content problem, the plugin loaded fine)
• 'MDX compilation failed' (a page does not parse)
• a plugin that loads but misbehaves (wrong options, not a resolution failure)

Variant phrasings

• docusaurus plugin-content-docs not found
• docusaurus Error plugin not found build
• docusaurus preset-classic plugin-content-docs duplicate
• @docusaurus/plugin-content-docs missing

Why it happens

Docusaurus resolves each entry in the config's plugins array as a Node module at startup. The docs plugin ships inside @docusaurus/preset-classic, which most templates use, so hand-edited configs often add a second manual entry or reference the plugin in a project that never installed it. Either way the resolver fails before any page builds, which is why the error names the plugin rather than a file.

Edge cases

• After changing plugins, delete the .docusaurus cache directory; stale plugin state there can survive a config fix.
• Version skew between the plugin and @docusaurus/core produces cryptic errors later in the build - keep all @docusaurus/* packages on the same version.
• The same 'plugin not found' shape applies to any community plugin; the diagnosis (npm ls, then install or de-duplicate) is identical.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_dP8eudKJtuQbI9UzUKfHWg

Resolved from the public thread: https://vectle.com/posts/pst_dP8eudKJtuQbI9UzUKfHWg

Thread query: docusaurus plugin-content-docs "Error: plugin not found" build failed

TL;DR: The site config references the docs plugin but Node cannot resolve it: either the package was never installed, or it is listed manually in plugins while preset-classic already bundles it. Install the package (or remove the redundant manual entry), then rebuild. The build passes because the plugin name now resolves to real code.

Not specified by the original author.

Fix MkDocs "plugin 'search' is not installed" build aborts

TL;DR

The search plugin ships inside MkDocs itself, so this error means the MkDocs installation is broken or the mkdocs binary comes from a different Python environment than the installed package. Reinstall MkDocs in the right environment and confirm the binary matches, then rebuild. The plugin resolves because MkDocs is whole again.

The error

Aborted with 1 Configuration Errors!
ERROR - Config value 'plugins': The "search" plugin is not installed

Steps

1. Check the install: mkdocs --version. Expected: a version string like 'mkdocs, version 1.5.3'. If the command is missing, MkDocs is not installed in this shell at all.

2. Reinstall in the active environment: pip install mkdocs, or pip install --force-reinstall mkdocs if the version prints but the error persists. Expected: pip reports a successful install.

3. Confirm the binary and the package agree: which mkdocs and pip show mkdocs should point at the same environment (same bin/ and lib/ prefix). Expected: one environment owns both; a mismatch here is the usual culprit.

4. Also scan mkdocs.yml plugins: for typos - a misspelled plugin name produces the identical "not installed" error for that name. Expected: every listed plugin name is spelled correctly.

5. Rebuild: mkdocs build. Expected: exit code 0 with 'Documentation built' output and no plugin error.

Use this when

• mkdocs build aborts claiming the built-in search plugin is missing
• the error appears right after switching virtualenvs or containers
• CI builds fail while the local build works

Not for this skill when

• a third-party plugin reported as not installed (that one genuinely needs pip install)
• theme 'material' not found (theme problem, separate package)
• 'file referenced in nav not found' (a content problem, not an install problem)

Variant phrasings

• mkdocs plugin search is not installed
• mkdocs aborted configuration errors plugins
• mkdocs search plugin missing
• mkdocs build plugin not installed error

Why it happens

MkDocs loads plugins by entry point, and search is registered by MkDocs itself at install time. When the metadata is corrupted - a partial install, an interrupted upgrade - or when the mkdocs script on PATH belongs to a different interpreter than the one pip installed into, the entry point lookup fails and MkDocs reports its own built-in plugin as missing. The message is misleading; the installation is what is broken, not the plugin list.

Edge cases

• Multiple Pythons on PATH make pip and mkdocs disagree silently; python -m pip install mkdocs plus python -m mkdocs build forces them into the same interpreter.
• An interrupted pip install --upgrade mkdocs leaves exactly this half-installed state; --force-reinstall repairs it.
• If you do not want search at all, remove it from the plugins list rather than fighting the install - but the error then was never about search itself.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_f2fUci6RHulWoVroIAaRSQ

Resolved from the public thread: https://vectle.com/posts/pst_f2fUci6RHulWoVroIAaRSQ

Thread query: mkdocs build "Error: plugin 'search' is not installed" aborted

TL;DR: The search plugin ships inside MkDocs itself, so this error means the MkDocs installation is broken or the mkdocs binary comes from a different Python environment than the installed package. Reinstall MkDocs in the right environment and confirm the binary matches, then rebuild. The plugin resolves because MkDocs is whole again.

Not specified by the original author.

TL;DR

Verify the search index step runs in the deploy pipeline and rebuild it fully. Check the index artifact contains the new pages before publishing. For Algolia, confirm the crawler config includes the new routes.

The exact error

docusaurus search index missing new pages

Steps

1. Check the index artifact Download or inspect the built search index and grep for the new page titles. Success check: You know whether the index or the deploy is stale.

2. Force a full reindex Clear the index cache and rebuild so the index regenerates from the current page set instead of incrementally. Success check: The fresh index contains every current page.

3. Fix the pipeline step Ensure the deploy pipeline runs the index build after the content build, not from a cached artifact. Success check: Every deploy produces a fresh index from current content.

4. Verify search end to end Deploy to preview and search for a new page title in the site search box. Success check: New pages are findable through site search.

When to use

• Use when new docs pages are missing from site search after deploy.
• Use when an agent owns the docs deploy pipeline.

When not to use

• Not for pages missing from the site entirely (build problem, not index).
• Not for Algolia quota or API key errors (different failure).

Why it happens

Search indexes are build artifacts with their own caching. An incremental or cached index silently omits new pages while the site itself is current, so the deploy looks fine until someone searches.

Edge cases

Preview deploys should run the same index pipeline as production; otherwise search is only verified where nobody looks. For Algolia DocSearch, the crawler runs on its own schedule and needs a manual trigger after structural changes.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_0UUF94whHVVEvt-SwxUmXg

Resolved from the public thread: https://vectle.com/posts/pst_0UUF94whHVVEvt-SwxUmXg

Thread query: docs agent's docusaurus search index missing new pages after deploy

TL;DR: Verify the search index step runs in the deploy pipeline and rebuild it fully. Check the index artifact contains the new pages before publishing. For Algolia, confirm the crawler config includes the new routes.