The [ ARKIV ] docs site already exposes per-page LLM exports — Open in ChatGPT, Copy Markdown, and .md mirrors. These are useful when a developer navigates to a page directly, but they are invisible to the ingestion pipelines that power AI coding assistants and builder tools.
LLM crawlers follow the llmstxt.org convention: they fetch one root-level index per domain at /llms.txt before deciding what to ingest. Without it, our per-page exports are never discovered, and [ ARKIV ] is absent from the tool suggestions developers receive when building on-chain applications.
Add two files to the root of the docs site:
/llms.txt — curated index of canonical pages, one entry per line in the format # Title\n> description\n- [Page](url), following the llmstxt.org spec. Covers the pages that matter most to a builder: quickstart, entity model, expiration behaviour, query API, SDK reference, and authentication./llms-full.txt — concatenated full markdown of all documentation pages, generated at build time. This is what crawlers pull when they want full content rather than the index.
Both files should be regenerated on each docs build and served as plain text from the site root.
The [ ARKIV ] docs site already exposes per-page LLM exports — Open in ChatGPT, Copy Markdown, and
.mdmirrors. These are useful when a developer navigates to a page directly, but they are invisible to the ingestion pipelines that power AI coding assistants and builder tools.LLM crawlers follow the llmstxt.org convention: they fetch one root-level index per domain at
/llms.txtbefore deciding what to ingest. Without it, our per-page exports are never discovered, and [ ARKIV ] is absent from the tool suggestions developers receive when building on-chain applications.Add two files to the root of the docs site:
/llms.txt— curated index of canonical pages, one entry per line in the format# Title\n> description\n- [Page](url), following the llmstxt.org spec. Covers the pages that matter most to a builder: quickstart, entity model, expiration behaviour, query API, SDK reference, and authentication./llms-full.txt— concatenated full markdown of all documentation pages, generated at build time. This is what crawlers pull when they want full content rather than the index.Both files should be regenerated on each docs build and served as plain text from the site root.