openxet docs
openxet is the open-source server and client for the Xet protocol, Hugging Face's chunk-deduplicated storage for large, versioned files. The official Hugging Face clients work against it unmodified. openxet is an independent project, not affiliated with or endorsed by Hugging Face.
- Run it: start with the storage-only quickstart, then configure the server.
- Script it: every command takes
--json, and the exit codes are stable. - Integrate with it: the CAS routes are the API stock clients call.
- What changed: the changelog.
- Why it is built this way: the design, the architecture decisions and the research notes, under Project.
For agents
Every page is also published as plain Markdown at the same path with .md
in place of .html, for example
reference/cli.md.
/llms.txt lists them all.
Editing these docs
This directory is the only place the docs are written; the site at
https://openxet.com/docs/ is built from it on every merge to main.
SUMMARY.md is the navigation, and every page under docs/ belongs in it
except the implementation plans in superpowers/plans/ and the ADR template.
Links are written relative to the file, as GitHub shows them; the build
points links to files outside the book at GitHub.
Preview while you edit, with mdBook
0.5.4 (cargo install mdbook --version 0.5.4 --locked) and python3, from
the repository root:
mdbook serve site --open # live preview of the docs at http://localhost:3000
scripts/build-site.sh # the whole site as deployed, into target/site/public
scripts/build-site.sh fails on an mdBook warning, on a page missing from
SUMMARY.md and on a link to a file that does not exist. CI's docs-site
job runs it, then checks every link and anchor in the built site and in
docs/ offline with lychee 0.24.2:
lychee --offline --include-fragments --root-dir "$PWD/target/site/public" \
--exclude-path docs/SUMMARY.md target/site/public docs