Sourced from docusaurus-plugin-llms's releases.

v0.6.0

This release ships a large batch of correctness fixes across content cleaning, partial resolution, output assembly, multi-version handling, and URL construction, plus a fully rewritten documentation set published at https://rachfop.github.io/docusaurus-plugin-llms.

Fixed

• <TabItem label="A > B"> leaked tag fragments: the TabItem label emitter cut its tag match at the first >, so a quoted label containing > dumped the rest of the tag into the output as prose. The matcher is now attribute-aware, like the PascalCase stripper.
• Multi-line descriptions rendered as broken blockquotes: a frontmatter description spanning lines quoted only its first line, leaving the rest as prose under the heading. Every line is now blockquoted.
• $ sequences corrupted partial content: partial files were spliced with a string replacement, so shell/perl patterns like echo $1, kill $$, or $& in a partial's code samples were silently rewritten (attr values substituted, tags re-inserted, document fragments duplicated). The splice now ignores replacement patterns.
• CRLF line endings corrupted output: files authored or checked out with Windows line endings had their fenced code blocks unmasked (imports and HTML inside code samples were stripped) and their descriptions mis-extracted (a heading-only file got the heading as description; a heading-less file got the entire document). readFile now normalizes CRLF/CR to LF.
• Root slug (slug: "/") crossed version subtrees: in multi-version mode, a section-root page linked to the current version's route (e.g. /docs) instead of its own version's (/stable/docs). Root-slug resolution now honors version and section scoping.
• siteConfig.url with a trailing slash produced //baseUrl in every generated URL, mismatching sitemap.xml.
• routeBasePath with a leading/trailing slash (e.g. '/docs') silently disabled route scoping for that section, falling every doc back to heuristic URLs. Slashes are now normalized.
• Blog detection misclassified sibling sections: any directory whose name merely started with the blog dir (e.g. a blog-api docs section next to blog) was treated as the blog for fallback URL construction.
• Code-fence samples containing import statements were rewritten: a fenced sample of an import line was resolved as a real partial import and spliced the partial's content into the code sample (or dropped the line when the target did not exist). Partial-import resolution now runs on fence-masked content.
• llms-full.txt flattened heading hierarchy: each document's inner ## Foo stayed at ##, colliding with the ## {title} document header, and the document's own title H1 appeared again under its header. Inner headings are now demoted one level under the document header, and the duplicate title heading is removed.
• Individual .md files duplicated their title and intro: the file header emitted # {title} plus the description blockquote, then the body repeated the identical H1 and (for pages whose description was auto-extracted from the first paragraph) the identical paragraph. Both duplicates are now stripped from the body.

... (truncated)

Sourced from docusaurus-plugin-llms's changelog.

[0.6.0] - 2026-08-31

Fixed

• <TabItem label="A > B"> leaked tag fragments: the TabItem label emitter cut its tag match at the first >, so a quoted label containing > dumped the rest of the tag into the output as prose. The matcher is now attribute-aware, like the PascalCase stripper.
• Multi-line descriptions rendered as broken blockquotes: a frontmatter description spanning lines quoted only its first line, leaving the rest as prose under the heading. Every line is now blockquoted.
• $ sequences corrupted partial content: partial files were spliced with a string replacement, so shell/perl patterns like echo $1, kill $$, or $& in a partial's code samples were silently rewritten (attr values substituted, tags re-inserted, document fragments duplicated). The splice now ignores replacement patterns.
• CRLF line endings corrupted output: files authored or checked out with Windows line endings had their fenced code blocks unmasked (imports and HTML inside code samples were stripped) and their descriptions mis-extracted (a heading-only file got the heading as description; a heading-less file got the entire document). readFile now normalizes CRLF/CR to LF.
• Root slug (slug: "/") crossed version subtrees: in multi-version mode, a section-root page linked to the current version's route (e.g. /docs) instead of its own version's (/stable/docs). Root-slug resolution now honors version and section scoping.
• siteConfig.url with a trailing slash produced //baseUrl in every generated URL, mismatching sitemap.xml.
• routeBasePath with a leading/trailing slash (e.g. '/docs') silently disabled route scoping for that section, falling every doc back to heuristic URLs. Slashes are now normalized.
• Blog detection misclassified sibling sections: any directory whose name merely started with the blog dir (e.g. a blog-api docs section next to blog) was treated as the blog for fallback URL construction.
• Code-fence samples containing import statements were rewritten: a fenced sample of an import line was resolved as a real partial import and spliced the partial's content into the code sample (or dropped the line when the target did not exist). Partial-import resolution now runs on fence-masked content.
• llms-full.txt flattened heading hierarchy: each document's inner ## Foo stayed at ##, colliding with the ## {title} document header, and the document's own title H1 appeared again under its header. Inner headings are now demoted one level under the document header, and the duplicate title heading is removed.
• Individual .md files duplicated their title and intro: the file header emitted # {title} plus the description blockquote, then the body repeated the identical H1 and (for pages whose description was auto-extracted from the first paragraph) the identical paragraph. Both duplicates are now stripped from the body.
• Triple blank lines in llms-full.txt: stripping a body's duplicate title heading left its surrounding blank lines behind. Gaps now collapse to

... (truncated)

• fa38f9b Clean up docs and add README badges
• e387927 Fix duplicated content and flattened headings in generated output
• fdef2c8 Deploy docs site on plugin source changes
• 48488b2 Mask code fences during partial-import resolution
• 4d8910b Apply documentation style guide across README and docs
• e9fca71 Add dogfooded docs site on GitHub Pages
• 0631b10 fix: TabItem attribute edge case, multi-line descriptions, and a friendlier p...
• c50648c fix: CVE-2026-9277 security vulnerability (#70)
• 4edcf06 fix: harden URL resolution, CRLF handling, and partial splicing (deep-dive au...
• bc14c92 fix: preserve prop content in JSX stripping and inline imported MDX page bodi...
• Additional commits viewable in compare view

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)