llms.txt

Overview

llms.txt is an auto-generated documentation endpoint that provides structured content for AI assistants. The index routes are /llms.txt (docs index), /llms.txt/docs, /llms.txt/blog, and /llms-full.txt. Per-page markdown is served through the .md and .txt aliases under /docs/* and /blog/*.

How It Works

The endpoint automatically scans documentation and published blog content, generating markdown files that include:

• Complete documentation pages
• Published blog articles
• Structured navigation
• Cross-references between pages

Access Points for AI Assistants

• llms.txt - Index for complete documentation for ZeroStarter
• llms.txt/blog - Index for latest articles and updates about ZeroStarter
• llms-full.txt - Complete documentation in one page for AI assistants
• Per-page .md and .txt aliases for individual docs and blog pages, e.g. /docs/manage/llms-txt.md or /blog/<slug>.txt

Only the per-page .md/.txt aliases under /docs/* and /blog/* are rewritten (see rewrites in web/next/next.config.ts). The top-level /docs.md, /docs.txt, /blog.md, and /blog.txt forms are not routes and return 404; use /llms.txt/docs and /llms.txt/blog for the index views.

Implementation

Two route handlers back these endpoints, and both are force-static with revalidate = 60:

• web/next/src/app/(llms.txt)/llms.txt/[[...slug]]/route.ts serves /llms.txt and the /llms.txt/docs and /llms.txt/blog index views, plus every per-page alias rewritten from /docs/* and /blog/*. It uses docsSource and blogSource from Fumadocs, sorts docs by meta.json order, and lists only published blog posts via getPublishedBlogPosts(). Console docs are deliberately excluded (privacy boundary).
• web/next/src/app/(llms.txt)/llms-full.txt/route.ts serves /llms-full.txt: every docs and blog page concatenated into one file, prefixed with site.llmsFullPreamble from packages/config/src/site.ts.

Page bodies come from getLLMText in web/next/src/lib/llms.ts, which:

• Calls page.data.getText("processed") and falls back to getText("raw") if the processed output is unavailable. The processed output exists because each collection in web/next/source.config.ts sets includeProcessedMarkdown: true (see the Documentation page for more on this flag).
• Runs decodeHTML from the entities package to undo Fumadocs' HTML-escaping (e.g. * back to *), so the LLM text is clean.
• Prefixes each page with a # [title](url) heading linking back to the live page.

The per-page docs handler also appends a footer pointing back to /llms.txt so an assistant can find the rest of the navigation.