Skip to content

GitHub Actions

We created prefix-dev/setup-pixi to facilitate using pixi in CI.

Usage#

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    pixi-version: v0.58.0
    cache: true
    auth-host: prefix.dev
    auth-token: ${{ secrets.PREFIX_DEV_TOKEN }}
- run: pixi run test

Pin your action versions

Since pixi is not yet stable, the API of this action may change between minor versions. Please pin the versions of this action to a specific version (i.e., prefix-dev/setup-pixi@v0.9.2) to avoid breaking changes. You can automatically update the version of this action by using Dependabot.

Put the following in your .github/dependabot.yml file to enable Dependabot for your GitHub Actions:

.github/dependabot.yml
version: 2
updates:
  - package-ecosystem: github-actions
    directory: /
    schedule:
      interval: monthly # (1)!
    groups:
      dependencies:
        patterns:
          - "*"
  1. or daily, weekly

Features#

To see all available input arguments, see the action.yml file in setup-pixi. The most important features are described below.

Caching#

The action supports caching of the project and global pixi environments. By default, project environment caching is enabled if a pixi.lock file is present. It will then use the pixi.lock file to generate a hash of the environment and cache it. If the cache is hit, the action will skip the installation and use the cached environment. You can specify the behavior by setting the cache input argument.

Global environment caching is disabled by default and can be enabled by setting the global-cache input to true. As there is no lockfile for global environments, the cache will expire at the end of every month to ensure it does not go stale.

Customize your cache key

If you need to customize your cache-key, you can use the cache-key and global-cache-key input arguments. These will be the prefixes of the cache keys. The full cache keys will be <cache-key><conda-arch>-<hash> and <global-cache-key><conda-arch>-<YYYY-MM>-<hash> respectively.

Only save caches on main

In order to not exceed the 10 GB cache size limit as fast, you might want to restrict when the cache is saved. This can be done by setting the cache-write argument.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    cache: true
    cache-write: ${{ github.event_name == 'push' && github.ref_name == 'main' }}

Multiple environments#

With pixi, you can create multiple environments for different requirements. You can also specify which environment(s) you want to install by setting the environments input argument. This will install all environments that are specified and cache them.

[workspace]
name = "my-package"
channels = ["conda-forge"]
platforms = ["linux-64"]

[dependencies]
python = ">=3.11"
pip = "*"
polars = ">=0.14.24,<0.21"

[feature.py311.dependencies]
python = "3.11.*"
[feature.py312.dependencies]
python = "3.12.*"

[environments]
py311 = ["py311"]
py312 = ["py312"]

Multiple environments using a matrix#

The following example will install the py311 and py312 environments in different jobs.

test:
  runs-on: ubuntu-latest
  strategy:
    matrix:
      environment: [py311, py312]
  steps:
  - uses: actions/checkout@v4
  - uses: prefix-dev/setup-pixi@v0.9.2
    with:
      environments: ${{ matrix.environment }}

Install multiple environments in one job#

The following example will install both the py311 and the py312 environment on the runner.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    environments: >- # (1)!
      py311
      py312
- run: |
  pixi run -e py311 test
  pixi run -e py312 test

  1. separated by spaces, equivalent to

    environments: py311 py312

Caching behavior if you don't specify environments

If you don't specify any environment, the default environment will be installed and cached, even if you use other environments.

Global Environments#

You can specify pixi global install commands by setting the global-environments input argument. This will create one environment per line, and install them. This is useful in particular to install executables that are needed for pixi install to work properly. For instance, the keyring, or gcloud executables. The following example shows how to install both in separate global environments. By default, global environments are not cached. You can enable caching by setting the global-cache input to true.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    global-environments: |
      google-cloud-sdk
      keyring --with keyrings.google-artifactregistry-auth
- run: |
    gcloud --version
    keyring --list-backends

Authentication#

There are currently five ways to authenticate with pixi:

  • using a token
  • using a username and password
  • using a conda-token
  • using an S3 key pair
  • using keyring for PyPI registries

For more information, see Authentication.

Handle secrets with care

Please only store sensitive information using GitHub secrets. Do not store them in your repository. When your sensitive information is stored in a GitHub secret, you can access it using the ${{ secrets.SECRET_NAME }} syntax. These secrets will always be masked in the logs.

Token#

Specify the token using the auth-token input argument. This form of authentication (bearer token in the request headers) is mainly used at prefix.dev.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    auth-host: prefix.dev
    auth-token: ${{ secrets.PREFIX_DEV_TOKEN }}

Username and password#

Specify the username and password using the auth-username and auth-password input arguments. This form of authentication (HTTP Basic Auth) is used in some enterprise environments with artifactory for example.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    auth-host: custom-artifactory.com
    auth-username: ${{ secrets.PIXI_USERNAME }}
    auth-password: ${{ secrets.PIXI_PASSWORD }}

Conda-token#

Specify the conda-token using the conda-token input argument. This form of authentication (token is encoded in URL: https://my-quetz-instance.com/t/<token>/get/custom-channel) is used at anaconda.org or with quetz instances.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    auth-host: anaconda.org # (1)!
    auth-conda-token: ${{ secrets.CONDA_TOKEN }}
  1. or my-quetz-instance.com

S3#

Specify the S3 key pair using the auth-access-key-id and auth-secret-access-key input arguments. You can also specify the session token using the auth-session-token input argument.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    auth-host: s3://my-s3-bucket
    auth-s3-access-key-id: ${{ secrets.ACCESS_KEY_ID }}
    auth-s3-secret-access-key: ${{ secrets.SECRET_ACCESS_KEY }}
    auth-s3-session-token: ${{ secrets.SESSION_TOKEN }} # (1)!
  1. only needed if your key uses a session token

See the S3 section for more information about S3 authentication.

PyPI keyring provider#

You can specify whether to use keyring to look up credentials for PyPI.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    pypi-keyring-provider: subprocess # one of 'subprocess', 'disabled'

Custom shell wrapper#

setup-pixi allows you to run command inside of the pixi environment by specifying a custom shell wrapper with shell: pixi run bash -e {0}. This can be useful if you want to run commands inside of the pixi environment, but don't want to use the pixi run command for each command.

- run: | # (1)!
    python --version
    pip install --no-deps -e .
  shell: pixi run bash -e {0}
  1. everything here will be run inside of the pixi environment

You can even run Python scripts like this:

- run: | # (1)!
    import my_package
    print("Hello world!")
  shell: pixi run python {0}
  1. everything here will be run inside of the pixi environment

If you want to use PowerShell, you need to specify -Command as well.

- run: | # (1)!
    python --version | Select-String "3.11"
  shell: pixi run pwsh -Command {0} # pwsh works on all platforms
  1. everything here will be run inside of the pixi environment

How does it work under the hood?

Under the hood, the shell: xyz {0} option is implemented by creating a temporary script file and calling xyz with that script file as an argument. This file does not have the executable bit set, so you cannot use shell: pixi run {0} directly but instead have to use shell: pixi run bash {0}. There are some custom shells provided by GitHub that have slightly different behavior, see jobs.<job_id>.steps[*].shell in the documentation. See the official documentation and ADR 0277 for more information about how the shell: input works in GitHub Actions.

One-off shell wrapper using pixi exec#

With pixi exec, you can also run a one-off command inside a temporary pixi environment.

- run: | # (1)!
    zstd --version
  shell: pixi exec --spec zstd -- bash -e {0}
  1. everything here will be run inside of the temporary pixi environment
- run: | # (1)!
    import ruamel.yaml
    # ...
  shell: pixi exec --spec python=3.11.* --spec ruamel.yaml -- python {0}
  1. everything here will be run inside of the temporary pixi environment

See here for more information about pixi exec.

Environment activation#

Instead of using a custom shell wrapper, you can also make all pixi-installed binaries available to subsequent steps by "activating" the installed environment in the currently running job. To this end, setup-pixi adds all environment variables set when executing pixi run to $GITHUB_ENV and, similarly, adds all path modifications to $GITHUB_PATH. As a result, all installed binaries can be accessed without having to call pixi run.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    activate-environment: true

If you are installing multiple environments, you will need to specify the name of the environment that you want to be activated.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    environments: >-
      py311
      py312
    activate-environment: py311

Activating an environment may be more useful than using a custom shell wrapper as it allows non-shell based steps to access binaries on the path. However, be aware that this option augments the environment of your job.

--frozen and --locked#

You can specify whether setup-pixi should run pixi install --frozen or pixi install --locked depending on the frozen or the locked input argument. See the official documentation for more information about the --frozen and --locked flags.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    locked: true
    # or
    frozen: true

If you don't specify anything, the default behavior is to run pixi install --locked if a pixi.lock file is present and pixi install otherwise.

Debugging#

There are two types of debug logging that you can enable.

Debug logging of the action#

The first one is the debug logging of the action itself. This can be enabled by for the action by re-running the action in debug mode:

Re-run in debug mode Re-run in debug mode

Debug logging documentation

For more information about debug logging in GitHub Actions, see the official documentation.

Debug logging of pixi#

The second type is the debug logging of the pixi executable. This can be specified by setting the log-level input.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    log-level: vvv # (1)!
  1. One of q, default, v, vv, or vvv.

If nothing is specified, log-level will default to default or vv depending on if debug logging is enabled for the action.

Self-hosted runners#

On self-hosted runners, it may happen that some files are persisted between jobs. This can lead to problems or secrets getting leaked between job runs. To avoid this, you can use the post-cleanup input to specify the post cleanup behavior of the action (i.e., what happens after all your commands have been executed).

If you set post-cleanup to true, the action will delete the following files:

  • .pixi environment
  • the pixi binary
  • the rattler cache
  • other rattler files in ~/.rattler

If nothing is specified, post-cleanup will default to true.

On self-hosted runners, you also might want to alter the default pixi install location to a temporary location. You can use pixi-bin-path: ${{ runner.temp }}/bin/pixi to do this.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    post-cleanup: true
    pixi-bin-path: ${{ runner.temp }}/bin/pixi # (1)!
  1. ${{ runner.temp }}\Scripts\pixi.exe on Windows

You can also use a preinstalled local version of pixi on the runner by not setting any of the pixi-version, pixi-url or pixi-bin-path inputs. This action will then try to find a local version of pixi in the runner's PATH.

Using the pyproject.toml as a manifest file for pixi.#

setup-pixi will automatically pick up the pyproject.toml if it contains a [tool.pixi.workspace] section and no pixi.toml. This can be overwritten by setting the manifest-path input argument.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    manifest-path: pyproject.toml

Only install pixi#

If you only want to install pixi and not install the current workspace, you can use the run-install option.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    run-install: false

Download pixi from a custom URL#

You can also download pixi from a custom URL by setting the pixi-url input argument. Optionally, you can combine this with the pixi-url-headers input argument to supply additional headers for the download request, such as a bearer token.

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    pixi-url: https://pixi-mirror.example.com/releases/download/v0.48.0/pixi-x86_64-unknown-linux-musl
    pixi-url-headers: '{"Authorization": "Bearer ${{ secrets.PIXI_MIRROR_BEARER_TOKEN }}"}'

The pixi-url input argument can also be a Handlebars template string. It will be rendered with the following variables:

  • version: The version of pixi that is being installed (latest or a version like v0.48.0).
  • latest: A boolean indicating if the version is latest.
  • pixiFile: The name of the pixi binary to download, as determined by the system of the runner (e.g., pixi-x86_64-unknown-linux-musl).

By default, pixi-url is equivalent to the following template:

- uses: prefix-dev/setup-pixi@v0.9.2
  with:
    pixi-url: |
      {{#if latest~}}
      https://github.com/prefix-dev/pixi/releases/latest/download/{{pixiFile}}
      {{~else~}}
      https://github.com/prefix-dev/pixi/releases/download/{{version}}/{{pixiFile}}
      {{~/if}}

Setting inputs from environment variables#

Alternatively to setting the inputs in the with section, you can also set each of them using environment variables. The corresponding environment variable names are derived from the input names by converting them to uppercase, replacing hyphens with underscores, and prefixing them with SETUP_PIXI_.

For example, the pixi-bin-path input can be set using the SETUP_PIXI_PIXI_BIN_PATH environment variable.

This is particularly useful if executing the action on a self-hosted runner.

Inputs always take precedence over environment variables.

More examples#

If you want to see more examples, you can take a look at the GitHub Workflows of the setup-pixi repository.