RagLeapOn 2026-10-12 the maintainer agreed by email to include the work in antonyrag/ragleap-core in the XERJ Corpus Hub. The Hub now lists a candidate corpus, ragleap--ragleap-core, and a contributor page for @antonyrag. The facts below come from those pages as they stood on 2026-10-21.
What the XERJ Corpus Hub is
According to its own description, it is a shared library of pre-indexed corpora that AI agents search locally and cite by file and line. A corpus is a pinned source, here the whole repository at one commit. The catalogue and the search packs are signed, and the xerj command line refuses anything that does not verify. Queries run on the user's own machine.
What is listed, and its status
- One selected commit,
7d3e27596b(a commit in the ragleap-graph Java port, 9 files and 822 lines). The corpus itself is the whole repository at the pinned commit. Licence: MIT. - Status: selected, in licence review. It is not in the Hub registry yet, so
xerj corpus addandxerj askcannot fetch it. It becomes installable when its card says live, and this post will be updated then.
What this is not
A candidate listing is not an audit, a certification or an endorsement by XERJ, and we have not measured how well AI agents retrieve this corpus. The Hub's links still use the old address antonyrag/ragleap-core, which redirects to ragleap/ragleap-core after the repository moved to the ragleap organisation. The Hub says the work is included with the author's permission, under the repository's own licence and with attribution, and that it can be removed on request.