LLMs and AI tools

Documentation is written for people, but it is increasingly read by models: a chat assistant answering a question about your project, an agent writing code against your API. ZenithDocs publishes a Markdown view of the site for them, generated at build time.

An index of the site, following the llms.txt convention: the site title, its description, then one Markdown link per page, grouped the way the sidebar is.

# ZenithDocs

> A static, lightweight and themeable documentation engine built on Astro and Lumos.

## Writing content

- [Markdown](/writing/markdown/): Standard Markdown, GitHub Flavored Markdown and a few extensions.

It is small on purpose: a model reads it first, then fetches only the pages it needs.

Every page, in reading order, separated by ---. Useful to paste a whole documentation site into a context window in one go.

Each page is also served as Markdown, at the same path with a .md extension: /installation/ is also available at /installation.md. Every page links to its Markdown version, and an untranslated page to the one of the default language:

<link rel="alternate" type="text/markdown" href="/installation.md" />

The title, the description, the source URL and the body of the page:

# Installation

> Add ZenithDocs to a new or existing Astro project.

Source: https://docs.example.com/installation/

ZenithDocs needs Astro 7 and Node.js 22.18 or later.

MDX import and export lines are removed, since they mean nothing outside a bundler. Component tags are kept: they carry content, and models read them fine.

zenith({
  title: 'My Docs',
  llms: false,
});