HOME·SCIENCE·WHAT IS CORPUS HUB?
SCIENCE · CODE AND CORPORA

What is Corpus Hub?

AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-11-13 · UPDATED · 2026-11-14
SHORT ANSWER · UPDATED 2026-11-14

What is Corpus Hub? It is XERJ's registry of signed, licence-reviewed index packs. An agent can ask a public corpus and get file:line citations, without cloning the tree. The catalogue generated on 2026-10-18 listed 14 corpora. On a recorded 2026-10-19 run, the first snappy answer took 2,356 ms.

Key number: The catalogue generated on 2026-10-18 listed 14 corpora. On a recorded 2026-10-19 run, the first snappy answer took 2,356 ms including the download; later answers took 30 to 40 ms. (as of 2026-10-19 · source: /answers/ask-questions-about-a-github-repo-without-cloning)

TL;DR Corpus Hub is XERJ's registry of signed, licence-reviewed index packs. An agent can ask a public corpus and get file:line citations, without cloning the tree. The catalogue generated on 2026-10-18 listed 14 corpora.

Try it: from registry to an answer

How a signed pack is used

  1. 1. Pin and review

    A corpus is a source pinned to a commit, with its licence reviewed, and packed as an index.

  2. 2. Sign the catalogue

    The client verifies a signed catalogue. The catalogue generated on 2026-10-18 listed 14 corpora.

  3. 3. Pull once

    xerj ask pulls the signed pack (for example the 10.8 MB snappy pack) and answers with file:line. Nothing is cloned.

  4. 4. Later questions stay local

    On the recorded 2026-10-19 run, the first snappy answer took 2,356 ms including the download; later answers took 30 to 40 ms.

Pack sizes, the 14-corpus catalogue date, and the 2,356 ms / 30 to 40 ms timings are from the published GitHub-repo ask answer (v1.0.0-rc.97, 2026-10-19). This walkthrough does not add a new measurement.
PASTE THIS TO YOUR AI AGENT — IT DOES THE REST Act as a coding agent. Read https://xerj.org/llms.txt, install XERJ, start a local node, index one folder I point you to, and run one search so I can see a cited passage.

How it works

A corpus here is a source pinned to a commit, with its licence reviewed, and packed as an index. A signed catalogue lists those packs. The client verifies the catalogue, pulls a pack, and asks it. The answer comes back with file:line. The source tree is not cloned onto the machine that asked.

Later questions can stay on the local copy of the pack, so they do not pay the download again.

How XERJ uses it

xerj ask pulls a signed pack, for example the 10.8 MB snappy pack, and answers with file:line. The catalogue generated on 2026-10-18 listed 14 corpora. On the recorded 2026-10-19 run, published in the GitHub-repo ask answer, the first snappy answer took 2,356 ms including the download; later answers took 30 to 40 ms.

This walkthrough does not add a new measurement. Issue #1800 is open for any-repo ask. The product page also describes Corpus Hub packs: product.

History and research

Corpus Hub is a XERJ product name. This page rests on the published ask answer and the product docs, not on an external paper.

Guides

Start with the Corpus Hub docs. Then read The Corpus Hub, Corpus Hub v2 and ask questions about a GitHub repo without cloning.

Related terms

Syntax-aware search and hybrid search.

FAQ

What is Corpus Hub?

It is XERJ's registry of signed, licence-reviewed index packs. An agent pulls a pack and asks it, with file:line citations, without cloning the source tree.

What has XERJ published about it?

The catalogue generated on 2026-10-18 listed 14 corpora. On the recorded 2026-10-19 run, the first snappy answer took 2,356 ms including the download; later answers took 30 to 40 ms.

Does the client trust the pack blindly?

No. The client verifies a signed catalogue. A corpus is a source pinned to a commit, with its licence reviewed, and packed as an index.

Where is the how-to?

The practical guide is the GitHub-repo ask answer, which is also the source of the timings above. Issue #1800 is open for any-repo ask.

Evidence

Related