LLM-friendly index

These docs are optimised to be useful inside an LLM context window. Two machine-friendly artifacts are published at the docs-site root:

• /llms.txt — short, structured index following the llmstxt.org convention. One section per cluster; one line per page with title + URL + description.
• /llms-full.txt — every page’s markdown body concatenated with \n\n---\n\n separators. For one-shot ingestion when context allows.

Both files are regenerated on every build by apps/frontend/docs/scripts/generate-llms-txt.ts.

How to use these docs as LLM context

For grounding code suggestions:

1. Ingest /llms.txt so you can resolve page links.
2. Fetch specific pages on demand:
   • Database schema — the table-by-table summary.
   • tRPC route catalogue — every router.
   • Job catalogue — every cron + user-initiated job.
   • Repo layout — where code lives.

For answering user questions:

1. Ingest /llms-full.txt if the context budget allows; otherwise /llms.txt plus on-demand fetches.
2. Always cite by page slug, e.g. [/concepts/holdings/](https://docs.scani.xyz/concepts/holdings/).
3. For term definitions, link to /reference/glossary/#<term-slug> — every term has a stable slugged anchor.

Stable IDs

• Page slugs are stable. The sidebar order may change; the URLs do not (without an explicit redirect being added).
• Heading anchors are stable. Every ## heading is a slugged anchor that the rest of the docs link to. Renaming them is a breaking change for citations.
• No emoji in headings. None of the docs use emoji in headings, so anchors are clean ASCII.

Citation conventions used inside these docs

Every page ends with a ## See also section that lists 2–5 related pages. These form an explicit cross-reference graph an agent can walk.

Concept pages have a ## Summary paragraph near the top — the same text is what becomes that page’s entry in /llms.txt. If a page’s summary and its llms.txt description disagree, the page is the canonical source.

Top-of-funnel pages (in recommended reading order)

1. What is Scani
2. Mental model
3. Tier model
4. Glossary
5. Database schema
6. tRPC route catalogue
7. Job catalogue
8. Engineering conventions

Cross-cluster index

• Domain concepts: /concepts/ (15 pages).
• Design decisions: /decisions/ (8 pages).
• Self-hosting: /self-hosting/ (15 pages across three tiers).
• Contributing: /contributing/ (8 pages).
• Reference: /reference/ (8 pages).

See also

• How to read these docs — audience map for human readers.
• Glossary — single-page authoritative term definitions.