Athas Boilerplate

AI-Assisted Docs Workflow

shared

How to keep docs AI-friendly for generation, retrieval, and assistant usage.

Current AI-friendly surfaces

  • llms.txt and llms-full.txt endpoints in docs app
  • per-page markdown route under /llms.mdx/docs/...
  • structured MDX pages with consistent headings and concise sections

Authoring rules for AI quality

  1. Prefer explicit section headers and short paragraphs.
  2. Keep command snippets complete and runnable.
  3. Use one canonical term for each concept to reduce ambiguity.
  4. Link related pages so assistants can traverse context quickly.
  5. Tag pages with audience frontmatter (internal, external-ready, shared).
  1. Draft in /internal first for operational accuracy.
  2. Validate with team usage and Q/A.
  3. Promote to /docs once language is external-ready.

Next AI improvements

  • add internal-specific LLM exports (/llms/internal.txt)
  • add optional frontmatter summary fields for retrieval ranking
  • define assistant prompt templates for docs update tasks

On this page