AI Source Surfaces
This platform is not merely AI-friendly — it is a first-class AI knowledge source. The machine audience is served through dedicated surfaces generated from the same Git content at build time, so bots never scrape the human HTML.
Surfaces in this prototype
Section titled “Surfaces in this prototype”| Surface | Path | Purpose |
|---|---|---|
| Discovery | /llms.txt | LLM discovery standard: a link-first map of public docs. |
| Catalog | /api/docs/index.json | Machine-readable catalog: IDs, URLs, summaries, metadata, content hashes. |
| Chunks | /api/docs/chunks.jsonl | Pre-chunked corpus with applicability metadata — ingest directly into any RAG pipeline. |
| Raw Markdown | /api/docs/<id>.md | Per-page source, e.g. mach4.homing.configure. |
Version safety for bots
Section titled “Version safety for bots”Every chunk in chunks.jsonl carries its applicability metadata (generation,
release, machine types, editions, hardware). A retrieval pipeline filters on the
reader’s Machine Profile, so a bot never gives Unimate instructions to a Vesta
customer — the exact rule the human Machine Profile enforces on the page.
Access control
Section titled “Access control”Only public + published content reaches any of these surfaces. The corpus
builder filters by visibility and publication before a single byte is emitted,
so private content is never bundled into an AI export.