Where llms.txt is a table of contents, llms-full.txt is the book. It concatenates every canonical page as Markdown into a single file so a retrieval system can ingest the entire site without traversing it page-by-page. This matters for models that cap crawl depth.
The file can be large (hundreds of KB), so it should be generated at build time, cached at the CDN edge, and regenerated only when content changes. Surfient regenerates it hourly by default and exposes an explicit ETag so retrievers can avoid redundant fetches.