llms-full.txt

llms-full.txt is a convention for one large file holding a whole documentation set as plain text, so a model can take the lot in a single fetch. It is not part of the llms.txt proposal.

The proposal defines /llms.txt as an index: a title, a summary, and lists of links. The "full" variant came afterwards, from documentation platforms that wanted one URL to hand a model. The FastHTML project shipped llms-ctx-full.txt, other sites ship llms-full.txt, and the naming was never settled because no specification covers it.

What Domduck records

Presence of /llms.txt only. The full variant is not probed, and that is a deliberate omission rather than a gap: an unnamed convention with at least two spellings cannot be measured consistently across a corpus, and a measurement that misses half its subject is worse than no measurement.

Domduck also does not serve one. /llms.txt is generated from the documentation registry and every URL in it is asserted by a test. A concatenation of pages a model can fetch in one hop each is worth building when there are more of them.

Why it matters

Two files with similar names do different jobs, and the difference is size rather than purpose. If you publish one, publish /llms.txt first, because it is the one with a written specification and the one anything else builds on.

If you publish the full variant, keep it generated. A hand-maintained concatenation of your documentation is stale within a release, and a model that reads a stale copy answers questions about a product you no longer ship.