If you run a blog or a content site, you have probably hit this: articles migrated from a newsletter, a forum or an old CMS carry no Markdown headings at all. The old outline component sees no h2 or h3 and simply tells you “no table of contents”. In a three-thousand-word piece the reader never knows where they are.
SmartToc’s core is a five-level fallback chain: if the article has proper headings, use the heading tree; if there are no headings but build-time AI data exists, use semantic sections; if the author used plain short sentences as mini-headings, detect them as pseudo-headings; if there is nothing at all, fall back to paragraph navigation. Every level tries to give the reader the best navigation available.
For existing articles with no heading structure, the AI clusters adjacent paragraphs by meaning — paragraphs about the same thing become one chapter — and then writes a short precise title for each chapter. The outline data lives in a separate JSON file; the component only renders it into the sidebar, and not a single character of the article changes. Any LLM that can emit JSON can generate that file, and it is cached once at build time.
The component is open source on GitHub under the MIT licence, published as silentxx-smart-toc. Installation is one npm command, wrap the body in an article tag, and everything else is automatic. The docs are bilingual with an architecture diagram and a FAQ — five minutes from quick start to running.
This is the first SilentXx open-source release, but it will not be the last. The options trading system, the AI automation workflows, the content production pipeline — tools that have proven themselves here will be released over time. If this component is useful to you, a star or an issue telling us your use case is very welcome.

