Quarry: what it does and how to install it

Quarry copies documentation from many Git repositories into one shared documentation repository. Each source repository contributes a folder. Every contributor gets its own local clone and SQLite index for searching pages, following dependency edges, and checking documented contract fields.

Quarry does not generate prose or inspect source code to infer dependencies. It imports the files already present in a configured documentation folder. Capstone can produce that folder; hand-maintained documentation works when it follows the formats described in Writing pages and interfaces.

Four places, four responsibilities#

An import pins a source commit. The index records a separate docs-repository commit. Those hashes answer different questions: which source code a folder documents, and which combined snapshot a query used.

The usual lifecycle is:

source docs at a committed HEAD
  -> quarry add / quarry update
  -> shared docs repository
  -> quarry sync in another source repository
  -> local SQLite index
  -> quarry docs ... / quarry check

quarry check has an extra input: the producer's current working-tree documentation. This makes it useful before merging a change, while ordinary queries answer from imported docs.

Install from the verified source revision#

As of 9 September 2026, the GitHub repository has no published releases or v0.* tags, and the README's latest/download/quarry-installer.sh route returns HTTP 404. Use the source build below. The manifest's version 0.2.0 identifies the code; it does not mean a v0.2.0 binary release has been published.

With Rust/Cargo and Git installed:

git clone https://github.com/GentBajko/quarry.git
cd quarry
git checkout 78c71fab5f5818662dc02d5867246cec963567a9
cargo install --path . --locked
quarry --version

Cargo normally puts the installed executable in its configured bin directory, commonly ~/.cargo/bin; ensure that directory is on your PATH. To build without installing:

cargo build --release --locked
./target/release/quarry --version

The manifest declares Rust edition 2024 and rust-version = "1.85". This guide's local source build used Rust/Cargo 1.91.1 and Git 2.43.0; it does not establish compatibility with every older compiler named by the manifest. The lockfile makes the dependency selection reproducible. The guide's recorded build used cargo build --offline --locked with dependencies already cached.

Recorded output, from that binary:

$ quarry --version
quarry 0.2.0

The README specifies Git 2.30 or newer on PATH. Quarry invokes the installed git executable, using its credentials, transport, proxy configuration, and commit identity. It does not implement its own authentication prompt or token store.

Release routes documented by the repository#

The README lists shell/PowerShell installers, platform archives, and cargo binstall quarry. These are repository-documented release routes, not verified currently working installation paths. No published GitHub assets existed at the verification date, and a package-registry/binstall route was not established by this guide. Do not use the following installer commands as a working quickstart today.

Shell installer URL and its intended invocation, once a release exists:

curl -LsSf https://github.com/GentBajko/quarry/releases/latest/download/quarry-installer.sh | sh

PowerShell equivalent, once the corresponding asset exists:

irm https://github.com/GentBajko/quarry/releases/latest/download/quarry-installer.ps1 | iex

The reusable CI template names a tag-specific URL ending in /releases/download/v0.2.0/quarry-installer.sh, which likewise requires a release that has not been published at this verification date. See CI and automation for the required source-build adaptation.

The release configuration names these intended targets:

This is a configured build matrix, not a list of currently downloadable artifacts. It includes GNU targets, so “a single binary” should not be read as “every build is fully statically linked.” SQLite is bundled into Quarry through rusqlite; normal use needs neither an external SQLite server nor its command-line program. There is no background daemon or model service. The release configuration disables an automatic updater.

Start in the right directory#

Run commands anywhere inside the source Git worktree. Quarry finds the root with git rev-parse --show-toplevel and uses the root's .quarry/ folder. The error says to run in a repo's root, but nested working directories resolve correctly.

Bare quarry, bare quarry docs, --help, and --version can display help without initialization. Other commands need a Git repository. Contribution commands also need a usable origin URL and a committed source HEAD.

For a new project that only needs to read other repos' docs, initialize Git first, then link to the docs repository. An origin, source commit, source docs folder, and quarry add are unnecessary for read-only use:

mkdir new-project
cd new-project
git init
quarry init --url git@github.com:acme/docs-quarry.git
quarry docs list

Continue with First import for contribution, or Reading and graph queries for retrieval.

Source basis#

Verified against Quarry 0.2.0, source revision 78c71fa. Cargo.toml dist-workspace.toml src/context.rs src/gitcmd.rs src/cli.rs README.md