Pixi: exercises - Reproducible Machine Learning Workflows for Scientists with Pixi

This page contains exercises to help you practice the concepts learned in the Pixi introduction. Each exercise is designed to reinforce your understanding of Pixi’s features and capabilities.

Exercise 1: Creating a project

Create a new Pixi project named my-project using the command line.
Add python as a dependency to your project.
Use the Pixi installed Python to print "Hello, Pixi!" to the console.
Run a command to print the version of Python used in your project. You can see the location of the interpreter by running: import sys;print(sys.executable)

Solution

# 1
pixi init ~/reproducible-ml-scipy-2025/my-project
cd ~/reproducible-ml-scipy-2025/my-project
# 2
pixi add python
# 3
pixi run python -c 'print("Hello, Pixi!")'
# 4
pixi run python -c 'import sys; print(sys.version); print(sys.executable)'

Resulting pixi.toml file:

pixi.toml

1
2
3
4
5
6
7
8
9
10
11
[workspace]
authors = ["Jane Doe <jane.doe@example.com>"]
channels = ["conda-forge"]
name = "my-project"
platforms = ["osx-arm64"]
version = "0.1.0"

[tasks]

[dependencies]
python = ">=3.13.5,<3.14"

Exercise 2: Dependencies

Make your project work for Windows, macOS, and Linux.
Add scipy as a dependency through the command line.
Add numpy as a dependency through the command line.
Add fastqc as a dependency from bioconda. Hint: the channel has to be added first.
Add pandas as a pypi dependency through the command line.
Add a pypi + git dependency on pytest to your project.
Visualize the locked dependencies in the command line.

Solution

# 1
pixi workspace platform add win-64 osx-arm64 osx-64 linux-64
# 2
pixi add scipy
# 3
pixi add numpy
# 4
pixi workspace channel add bioconda
pixi add bioconda::fastqc
# 5
pixi add pandas --pypi
# 6
pixi add pytest --pypi --git https://github.com/pytest-dev/pytest
# 7: Any of:
pixi list
pixi list -x
pixi tree

Resulting pixi.toml file:

pixi.toml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
[workspace]
authors = ["Jane Doe <jane.doe@example.com>"]
channels = ["conda-forge", "bioconda"]
name = "my-project"
platforms = ["osx-arm64", "win-64", "osx-64", "linux-64"]
version = "0.1.0"

[tasks]

[dependencies]
python = ">=3.13.5,<3.14"
scipy = ">=1.16.0,<2"
numpy = ">=2.3.0,<3"
fastqc = { version = ">=0.12.1,<0.13", channel = "bioconda" }

[pypi-dependencies]
pandas = ">=2.3.0, <3"
pytest = { git = "https://github.com/pytest-dev/pytest" }

Exercise 3: Modifying dependencies

Replace pandas with a conda dependency on pandas from the conda-forge channel. Hint: make sure it worked by checking the environment with pixi list pandas
Require the scipy dependency to be version 1.15.1.
Make the pytest use version 8.3.1 from the git repository.

Solution

# 1
pixi add pandas
pixi list pandas
# Optionally, you can remove the old pandas dependency:
pixi remove pandas --pypi
# 2
pixi add scipy==1.15.1
# 3
pixi add pytest --pypi --git https://github.com/pytest-dev/pytest.git --tag 8.3.1

Resulting pixi.toml file:

pixi.toml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
[workspace]
authors = ["Jane Doe <jane.doe@example.com>"]
channels = ["conda-forge", "bioconda"]
name = "my-project"
platforms = ["osx-arm64", "win-64", "osx-64", "linux-64"]
version = "0.1.0"

[tasks]

[dependencies]
python = ">=3.13.5,<3.14"
scipy = "==1.15.1"
numpy = ">=2.3.0,<3"
fastqc = { version = ">=0.12.1,<0.13", channel = "bioconda" }
pandas = ">=2.3.0,<3"

[pypi-dependencies]
pytest = { git = "https://github.com/pytest-dev/pytest.git", tag = "8.3.1" }

Exercise 4: Multiple environments

Make pytest part of the test feature.
Create a new environment named test that includes the test feature.
Install the test environment.
Run the pytest command in the test environment to ensure it works correctly.
Create a format feature that includes ruff.
Create a new environment named format that includes the format feature but not the default and test features.
Run ruff --version.
Now make sure the test and default environments are using the same solve group.
Compare the dependencies with pixi list --explicit --environment ENV_NAME. Validate the versions of the dependencies in both environments are the same.

Solution

1: move pytest from [pypi-dependencies] to [feature.test.pypi-dependencies] in pixi.toml:

pixi.toml

-[pypi-dependencies]
+[feature.test.pypi-dependencies]
pytest = { git = "https://github.com/pytest-dev/pytest.git", tag = "8.3.1" }

# 2
pixi workspace environment add test --feature test
# 3
pixi install --environment test
# Or short:
pixi i -e test
# 4
pixi run -e test pytest # Expect the test to fail as there are no tests ;).
# 5
pixi add ruff --feature format
# 6
pixi workspace environment add format --feature format --no-default-feature
# 7
pixi run -e format ruff --version
# 8
pixi workspace environment add test --feature test --solve-group group1 --force
pixi workspace environment add default --solve-group group1 --force
# 9
pixi list -x -e test
pixi list -x

Resulting pixi.toml file:

pixi.toml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
[workspace]
authors = ["Jane Doe <jane.doe@example.com>"]
channels = ["conda-forge", "bioconda"]
name = "my-project"
platforms = ["osx-arm64", "win-64", "osx-64", "linux-64"]
version = "0.1.0"

[tasks]

[dependencies]
python = ">=3.13.5,<3.14"
scipy = "==1.15.1"
numpy = ">=2.3.0,<3"
fastqc = { version = ">=0.12.1,<0.13", channel = "bioconda" }
pandas = ">=2.3.0,<3"

[feature.test.pypi-dependencies]
pytest = { git = "https://github.com/pytest-dev/pytest.git", tag = "8.3.1" }

[feature.format.dependencies]
ruff = "*"

[environments]
test = { features = ["test"], solve-group = "group1" }
format = { features = ["format"], no-default-feature = true }
default = { solve-group = "group1" }

Exercise 5: Task running

Create a task named hello that prints "Hello, Pixi!" to the console.
Create a task named greet that prints "Hello, <name>!" to the console, where <name> is a argument passed to the task.
Create a task named greet-YOUR_NAME that prints "Hello, YOUR_NAME!" to the console, where it depends on the greet task and gives it an argument.
Create a start task that runs the hello task first and then the greet task, but with the arguments passed to the start task, passed to the greet task.
Run all these tasks.
Add the task fmt to the format feature that runs ruff --version.
Run the fmt task.

Solution

# 1
pixi task add hello 'echo "Hello, Pixi!"'
# 2
pixi task add greet 'echo "Hello, {{ name }}!"' --arg name
# 3
pixi task add greet-YOUR_NAME --depends-on greet::YOUR_NAME
# 4
pixi task add start --depends-on hello --depends-on "greet::{{ name }}" --arg name
# 5
pixi run hello
pixi run greet your_name
pixi run greet-YOUR_NAME
pixi run start your_name
# 6
pixi task add fmt 'ruff --version' --feature format
# 7 Notice that it automatically runs in the `format` environment.
pixi run fmt

Resulting pixi.toml file:

pixi.toml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
[workspace]
authors = ["Jane Doe <jane.doe@example.com>"]
channels = ["conda-forge", "bioconda"]
name = "my-project"
platforms = ["osx-arm64", "win-64", "osx-64", "linux-64"]
version = "0.1.0"

[tasks]
hello = 'echo "Hello, Pixi!"'
greet = { cmd = 'echo "Hello, {{ name }}!"', args = ["name"] }
greet-YOUR_NAME = [{ task = "greet", args = ["YOUR_NAME"] }]
start = { args = ["name"], depends-on = [{ task = "hello" }, { task = "greet", args = ["{{ name }}"] }] }

[dependencies]
python = ">=3.13.5,<3.14"
scipy = "==1.15.1"
numpy = ">=2.3.0,<3"
fastqc = { version = ">=0.12.1,<0.13", channel = "bioconda" }
pandas = ">=2.3.0,<3"

[feature.test.pypi-dependencies]
pytest = { git = "https://github.com/pytest-dev/pytest.git", tag = "8.3.1" }

[feature.format.dependencies]
ruff = "*"

[feature.format.tasks]
fmt = "ruff --version"

[environments]
test = { features = ["test"], solve-group = "group1" }
format = { features = ["format"], no-default-feature = true }
default = { solve-group = "group1" }

Exercise 6: Activation

Set an environment variable named MY_ENV_VAR to the value Hello, Pixi! in the default environment.
Create a task named print-env that prints the value of MY_ENV_VAR to the console.
Run the print-env task in the default environment to ensure it prints the correct value.
Use pixi shell to activate the environment in the current shell.
Verify that the MY_ENV_VAR is set in the current shell by running echo $MY_ENV_VAR (or echo %MY_ENV_VAR% on Windows).

Solution

Set the environment variable in the pixi.toml file:

pixi.toml

[activation.env]
MY_ENV_VAR = "Hello, Pixi!"

# 2
pixi task add print-env 'echo $MY_ENV_VAR'
# 3
pixi run print-env
# 4
pixi shell
# 5
echo $MY_ENV_VAR
# Or on Windows:
echo %MY_ENV_VAR%

Resulting pixi.toml file:

pixi.toml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
[workspace]
authors = ["Jane Doe <jane.doe@example.com>"]
channels = ["conda-forge", "bioconda"]
name = "my-project"
platforms = ["osx-arm64", "win-64", "osx-64", "linux-64"]
version = "0.1.0"

[activation.env]
MY_ENV_VAR = "Hello, Pixi!"

[tasks]
hello = 'echo "Hello, Pixi!"'
greet = { cmd = 'echo "Hello, {{ name }}!"', args = ["name"] }
greet-YOUR_NAME = [{ task = "greet", args = ["YOUR_NAME"] }]
start = { args = ["name"], depends-on = [{ task = "hello" }, { task = "greet", args = ["{{ name }}"] }] }
print-env = "echo $MY_ENV_VAR"

[dependencies]
python = ">=3.13.5,<3.14"
scipy = "==1.15.1"
numpy = ">=2.3.0,<3"
fastqc = { version = ">=0.12.1,<0.13", channel = "bioconda" }
pandas = ">=2.3.0,<3"

[feature.test.pypi-dependencies]
pytest = { git = "https://github.com/pytest-dev/pytest.git", tag = "8.3.1" }

[feature.format.dependencies]
ruff = "*"

[feature.format.tasks]
fmt = "ruff --version"

[environments]
test = { features = ["test"], solve-group = "group1" }
format = { features = ["format"], no-default-feature = true }
default = { solve-group = "group1" }

Exercise 7: Push the project to GitHub

Make sure you have a GitHub repository created for this workshop, like described in the setup section.
Commit your changes to the local repository.
Push your changes to the remote GitHub repository.

Solution

# 1:
git commit -am "Completed Pixi exercises"
# 2:
git push origin main

Reproducible Machine Learning Workflows for Scientists with Pixi

Pixi: Introduction

Reproducible Machine Learning Workflows for Scientists with Pixi

Introduction to CUDA conda packages