Installing the CLI
The data-dict command line tool validates a data-dict.yaml file against the specification, against a dataset’s metadata, and against the data itself; it can also draft, render, export, and translate dictionaries. See validation for what each level checks.
Every release ships prebuilt binaries, so you don’t need a Rust toolchain to install it.
Install script
On macOS and Linux:
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/tidyverse/data-dict/releases/latest/download/data-dict-cli-installer.sh | shOn Windows, in PowerShell:
powershell -ExecutionPolicy Bypass -c "irm https://github.com/tidyverse/data-dict/releases/latest/download/data-dict-cli-installer.ps1 | iex"The script downloads the binary for your platform, verifies its checksum, and puts data-dict in ~/.cargo/bin (%USERPROFILE%\.cargo\bin on Windows), adding that directory to your PATH if it isn’t already there. Set DATA_DICT_CLI_INSTALL_DIR to install somewhere else, and DATA_DICT_CLI_NO_MODIFY_PATH=1 to leave your PATH alone.
To install a specific version, replace latest/download with download/v0.0.1 (or whichever tag you want).
Check that it worked:
data-dict --versionInstall with uv or pipx
data-dict is on PyPI as data-dict-yaml, so the Python tool installers can manage it for you:
uv tool install data-dict-yamlpipx install data-dict-yamlEither way the command you get is data-dict. To run it once without installing anything:
uvx --from data-dict-yaml data-dict validate-spec data-dict.yamlLater, uv tool upgrade data-dict-yaml moves to the newest release, and uv tool install data-dict-yaml==0.0.2 pins an older one.
The wheels carry the same prebuilt binary as the install script, so there is nothing to compile and no Rust toolchain needed. They cover macOS (Apple silicon and Intel), Linux (x86-64 and ARM64), and Windows (x86-64). There is no wheel for Windows on ARM64: use the install script above, which fetches the x86-64 binary.
Download a binary
If you’d rather not pipe a script into a shell, grab the archive for your platform from the releases page, unpack it, and move the data-dict binary onto your PATH. Each archive has a matching .sha256 file, and every release has a combined sha256.sum.
The binaries aren’t code-signed, so macOS quarantines archives downloaded through a browser. If Gatekeeper refuses to run data-dict, clear the flag with xattr -d com.apple.quarantine data-dict, or use the install script above, which isn’t affected.
Prebuilt binaries are available for:
| Platform | Target |
|---|---|
| macOS (Apple silicon) | aarch64-apple-darwin |
| macOS (Intel) | x86_64-apple-darwin |
| Linux (ARM64) | aarch64-unknown-linux-musl |
| Linux (x86-64) | x86_64-unknown-linux-musl |
| Windows (x86-64) | x86_64-pc-windows-msvc |
The Linux binaries are statically linked against musl, so they have no libc dependency and run on any distribution, glibc or musl.
From R
The datadict R package (in r/ in the repository) installs the binary into R’s user cache directory and calls it for you:
pak::pak("tidyverse/data-dict/r")
datadict::dd_install()
datadict::dd_validate_data("path/to/project")dd_install() downloads the release archive for your platform, checks it against the published .sha256, and unpacks data-dict into tools::R_user_dir("datadict", "cache"). dd_validate_data() runs validate-data, writes the HTML report, and opens it in your browser.
Build from source
On any other platform, build it yourself with Cargo:
cargo install --git https://github.com/tidyverse/data-dict data-dict-cliThe prebuilt binaries bundle the data-dict.yaml language server, which the editor integrations use. A source install leaves it out unless you ask for it:
cargo install --git https://github.com/tidyverse/data-dict data-dict-cli --features lspUninstall
Delete the binary:
rm ~/.cargo/bin/data-dictThe install script also leaves a receipt in ~/.config/data-dict-cli/data-dict-cli-receipt.json recording what it installed and where; you can delete that too.