pixi（pixi 搜索）

Pixi Package Manager

Overview

Pixi is a fast, modern package manager built on Rattler (Rust-based conda implementation) that provides reproducible, cross-platform environments. It combines conda and PyPI ecosystems, supports multiple languages, and includes built-in task management.

Quick Start

Check Installation

Before working with pixi, verify it's installed:

bash scripts/check_pixi.sh

If not installed, follow the installation instructions provided by the script.

Common Workflows

Initialize new project:

pixi init my-project
cd my-project
pixi add python numpy pandas

Add packages:

pixi add <package>              # Conda package
pixi add --pypi <package>        # PyPI package
pixi add --feature dev pytest    # Add to specific feature

Run commands:

pixi run <task>                  # Run defined task
pixi exec <command>              # Execute command in environment
pixi shell                       # Activate environment shell

Manage environments:

pixi shell --environment <name>  # Activate specific environment
pixi run -e <name> <task>        # Run task in environment

Core Capabilities

1. Project Initialization

Initialize new pixi projects with automatic manifest creation:

# Standard pixi.toml format
pixi init my-project

# Use pyproject.toml format
pixi init --format pyproject my-project

The manifest file (pixi.toml or pyproject.toml) defines:

• Project metadata (name, version, description)
• Dependencies (conda and PyPI packages)
• Tasks (commands to run)
• Features (optional dependency sets)
• Environments (combinations of features)
• Channels (package sources)
• Platforms (target operating systems)

2. Package Management

Add, remove, and manage packages from conda-forge and PyPI:

# Add packages
pixi add numpy pandas matplotlib
pixi add "python>=3.11,<3.12"
pixi add --pypi requests flask

# Add to specific feature
pixi add --feature dev pytest black ruff

# Remove packages
pixi remove numpy
pixi remove --feature dev pytest

# Update packages
pixi update                      # Update all
pixi update numpy                # Update specific package
pixi upgrade                     # Upgrade in manifest

Package types:

• Regular dependencies: Runtime requirements
• --pypi: PyPI packages (via uv integration)
• --host: Host dependencies (available at runtime)
• --build: Build dependencies (only during build)

3. Environment Management

Pixi supports multiple isolated environments within a single project. Each environment is a combination of features.

Key concepts:

• Default feature: Always included automatically in every environment
• Custom features: Named sets of dependencies, tasks, and configurations
• Environments: Combinations of features for different purposes

Example configuration:

[dependencies]
python = "3.11.*"
numpy = "*"

[feature.test.dependencies]
pytest = "*"
pytest-cov = "*"

[feature.dev.dependencies]
black = "*"
ruff = "*"

[environments]
test = ["test"]                  # Includes: default + test
dev = ["dev", "test"]            # Includes: default + dev + test

Working with environments:

# Activate environment
pixi shell --environment test

# Run task in environment
pixi run --environment test pytest

# Install specific environment
pixi install --environment dev

# List packages in environment
pixi list --environment test

For detailed environment and feature patterns, see references/features-guide.md.

4. Feature Management

Features are building blocks that define sets of packages and configurations. They enable:

• Separating development from production dependencies
• Supporting multiple Python versions
• Creating platform-specific configurations
• Managing hardware-specific dependencies (CUDA vs CPU)

Common patterns:

Development features:

[feature.dev.dependencies]
pytest = "*"
black = "*"
ruff = "*"
mypy = "*"

[feature.dev.tasks]
test = "pytest tests/"
format = "black ."
lint = "ruff check ."

Python version features:

[feature.py310.dependencies]
python = "3.10.*"

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

[environments]
py310 = ["py310"]
py311 = ["py311"]

Hardware features:

[feature.cuda.dependencies]
pytorch-cuda = { version = "*", channel = "pytorch" }

[feature.cpu.dependencies]
pytorch-cpu = { version = "*", channel = "pytorch" }

[environments]
cuda = ["cuda"]
cpu = ["cpu"]

For comprehensive feature patterns and best practices, see references/features-guide.md.

5. Task Management

Define and run tasks within the pixi environment:

# Add tasks
pixi task add test "pytest tests/"
pixi task add format "black ."
pixi task add lint "ruff check ."

# Run tasks
pixi run test
pixi run format

# List tasks
pixi task list

Task configuration in manifest:

[tasks]
start = "python app.py"
test = "pytest tests/"
format = "black ."

# Task with dependencies
build = { cmd = "python setup.py build", depends-on = ["install"] }

# Feature-specific tasks
[feature.dev.tasks]
dev = "python app.py --reload"

6. Command Execution

Execute commands within the pixi environment:

# Run command in activated environment
pixi exec python script.py
pixi exec pytest tests/

# Temporary package installation and execution
pixi exec --spec ruff ruff check .
pixi exec --spec black black .

# Run in specific environment
pixi exec --environment test pytest

7. Global Tool Management

Install CLI tools system-wide, safely isolated:

# Install global tools
pixi global install gh ripgrep fd-find bat
pixi global install ruff black mypy

# List global tools
pixi global list

# Upgrade tools
pixi global upgrade ruff
pixi global upgrade-all

# Remove tools
pixi global remove bat

8. Project Information

Get information about the current project:

# Show project info
pixi info

# Detailed information
pixi info --extended

# JSON output
pixi info --json

# Use helper script
python scripts/pixi_info.py

9. Lock File Management

Pixi maintains a pixi.lock file for reproducible environments:

# Update lock file without installing
pixi lock

# Update lock for specific environment
pixi lock --environment test

# Install from lock file
pixi install --frozen

10. Maintenance

Clean up and maintain pixi projects:

# Remove environment directory
pixi clean

# Clean package cache
pixi clean cache

# Reinstall environment from scratch
pixi reinstall

Reference Documentation

For detailed information, consult these references:

• cli-reference.md - Complete CLI command reference with all options and flags
• features-guide.md - Comprehensive guide to features and multi-environment setup
• examples.md - Real-world usage examples and common scenarios

Workflow Decision Guide

Starting a new project? → Use pixi init and add dependencies with pixi add

Adding packages? → Use pixi add for conda packages, pixi add --pypi for PyPI packages → Use --feature flag to add to specific features

Need multiple environments? → Define features in manifest, compose them into environments → See references/features-guide.md for patterns

Running commands? → Use pixi run for defined tasks → Use pixi exec for ad-hoc commands → Use pixi shell to activate environment interactively

Managing dependencies? → Use pixi update to update to newer compatible versions → Use pixi upgrade to upgrade and modify manifest → Use pixi lock to update lock file without installing

Need global tools? → Use pixi global install for system-wide CLI tools

Troubleshooting? → Run pixi info to check project status → Run python scripts/pixi_info.py for detailed information → Check references/examples.md for common solutions

Best Practices

1. Use features for optional dependencies - Keep default feature minimal with only production dependencies
2. Leverage solve groups - Ensure consistent versions across related environments
3. Define tasks in manifest - Make common operations easily repeatable
4. Use semantic versioning - Specify version constraints appropriately
5. Commit lock file - Ensure reproducible environments across team
6. Use global install for tools - Keep project dependencies clean
7. Document environment purposes - Add comments explaining each environment's use case

Common Scenarios

Data Science Project

pixi init data-project
pixi add python numpy pandas matplotlib jupyter scikit-learn
pixi add --feature dev pytest black
pixi task add notebook "jupyter lab"
pixi run notebook

Web Application

pixi init web-app
pixi add python
pixi add --pypi fastapi uvicorn sqlalchemy
pixi add --feature dev --pypi pytest httpx
pixi task add dev "uvicorn app:app --reload"
pixi run dev

Multi-Python Testing

[feature.py310.dependencies]
python = "3.10.*"

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

[environments]
py310 = ["py310"]
py311 = ["py311"]

pixi run -e py310 pytest
pixi run -e py311 pytest

For more scenarios, see references/examples.md.

Resources

Scripts

• scripts/check_pixi.sh - Verify pixi installation
• scripts/pixi_info.py - Get detailed project information

References

• references/cli-reference.md - Complete CLI command reference
• references/features-guide.md - Features and multi-environment guide
• references/examples.md - Usage examples and common scenarios

Assets

• assets/pixi.toml.template - Basic project template

External Resources

• Official documentation: https://pixi.prefix.dev/latest/
• GitHub repository: https://github.com/prefix-dev/pixi
• Conda-forge packages: https://conda-forge.org/