Skip to content

Robots

llms.txt, raw Markdown per page, and provenance metadata for agents.

AI generated, pending review Updated 10 days ago · Flare 0.3

This page is for agents reading the docs. Every page records how it was written and ships a machine-readable form, and the indexes below regenerate on each build so links do not drift.

https://flare-ts.dev/llms.txt is a compact map of the docs: project name, description, and links to the most important pages. Fetch it first when you need orientation before pulling full pages.

https://flare-ts.dev/llms-full.txt concatenates the docs corpus into one file for a single fetch. It is large; prefer llms.txt for navigation and pull individual pages for depth.

https://flare-ts.dev/llms-small.txt is the same corpus minified for a smaller context budget. Reach for it when llms-full.txt will not fit but you still want the whole corpus in one fetch.

Append .md to any docs URL to get that page as Markdown with YAML frontmatter (title, description, diataxis, provenance when present):

https://flare-ts.dev/docs/introduction.md
https://flare-ts.dev/docs/http/routing.md

The body matches the authored source, fenced code blocks included. Use this when you need the exact text the site was built from, not an HTML scrape.

Authored pages carry provenance frontmatter:

provenance:
status: ai-generated # or human-reviewed after sign-off
model: <model id>
generatedAt: "2026-07-07"
reviewedBy: null
reviewedHash: <sha256> # stamped at sign-off; review fails the lint if the body drifts from it

The badge on the page reflects this metadata. AI authorship is disclosed. Grounding is a corpus property, not a page property: web/grounding.json pins the one framework version every page’s examples were verified against, and the badge renders that pin.

  • Examples: every TypeScript block on an authored page compiles against the published @flare-ts packages in CI.
  • Generated API reference: TypeDoc output composed from source, not hand-edited prose. Treat it as the symbol authority.
  • Precedence: when a prose claim and the API reference disagree, the published types win.