Skip to content

Python

Python development setup and package management using modern tools.

Overview

This setup uses two modern package managers: - UV: Fast, reliable Python package manager and project tool - Pixi: Cross-platform package manager compatible with conda

Installing Python Build Standalone (Indigo)

For a clean, optimized Python environment, use Python Build Standalone - a high-quality standalone Python distribution that works on Linux (x86_64 and ARM64) and macOS (Intel and Apple Silicon).

Use the automated installer script that automatically detects your OS and architecture:

1
curl -fsSL "https://raw.githubusercontent.com/dirkpetersen/dok/main/scripts/python-build-standalone-install.sh?$(date +%s)" | bash

This script will: - Automatically detect your OS and architecture (x86_64 Linux, ARM64 Linux, macOS Intel, macOS Apple Silicon) - Fetch the latest release from GitHub - Download and extract Python to ~/.python - Verify the installation - Display next steps to add Python to your PATH

Custom Installation Directory

To install to a different directory, pass it as an argument: 

1
curl -fsSL "https://raw.githubusercontent.com/dirkpetersen/dok/main/scripts/python-build-standalone-install.sh?$(date +%s)" | bash -s /opt/python

Manual Installation (All Platforms)

If you prefer to install manually or the script doesn't work, follow these steps:

1. Detect your system's target identifier:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
# Linux x86_64
echo "x86_64-unknown-linux-gnu"

# Linux ARM64
echo "aarch64-unknown-linux-gnu"

# macOS Intel (x86_64)
echo "x86_64-apple-darwin"

# macOS Apple Silicon (ARM64)
echo "aarch64-apple-darwin"

2. Visit the releases page to find the latest version: github.com/indygreg/python-build-standalone/releases

3. Download the appropriate release (replace YOUR_TARGET with your system's identifier):

1
2
3
4
RELEASE_TAG="20241217"  # Replace with latest tag from releases page
TARGET="x86_64-unknown-linux-gnu"  # Replace with your target

curl -fsSL "https://github.com/indygreg/python-build-standalone/releases/download/$RELEASE_TAG/cpython-3.13.$RELEASE_TAG-$TARGET.tar.zst" -o python.tar.zst

4. Extract to your home directory:

1
2
mkdir -p ~/.python
tar -I zstd -xf python.tar.zst -C ~/.python

5. Add to PATH

Add this to your shell configuration (.bashrc, .zshrc):

1
export PATH="$HOME/.python/python/install/bin:$PATH"

Then reload your shell:

1
source ~/.bashrc  # or source ~/.zshrc

6. Verify Installation

1
python --version

UV Package Manager

UV is a fast, reliable package manager for Python projects.

Installation

1
curl -fsSL https://astral.sh/uv/install.sh | sh

Creating a New Project

1
2
uv init my-project
cd my-project

Adding Dependencies

1
2
uv add requests
uv add --dev pytest

Running Python Scripts

1
uv run script.py

Installing from pyproject.toml

1
uv sync

Creating a Virtual Environment

1
2
uv venv
. .venv/bin/activate  # On Windows: .venv\Scripts\activate

Pixi Package Manager

Pixi is a cross-platform package manager compatible with conda environments, great for data science and scientific computing.

Installation

1
curl -fsSL https://pixi.sh/install.sh | bash

Creating a New Project

1
2
pixi init my-data-project
cd my-data-project

Adding Dependencies

1
2
pixi add numpy pandas
pixi add --pypi requests

Running Commands

1
pixi run python script.py

Creating Environments

1
2
pixi create -e analysis
pixi run -e analysis python analysis.py

Conda Compatibility

Pixi automatically handles conda channels and environments:

1
pixi add -c conda-forge numpy

Choosing Between UV and Pixi

Use Case Tool
Standard Python projects UV
Data science / Jupyter Pixi
Machine learning Pixi
Web development UV
CLI tools UV
Scientific computing Pixi
Mixed dependencies (PyPI + conda) Pixi

Virtual Environments

With UV

1
2
3
4
5
6
7
8
# Create virtual environment
uv venv

# Activate
. .venv/bin/activate

# Deactivate
deactivate

With Pixi

1
2
3
4
5
# Create environment
pixi create -e myenv

# Use environment
pixi run -e myenv python script.py

Best Practices

  • Always use virtual environments to isolate project dependencies
  • Use UV for most Python projects - it's faster and simpler
  • Use Pixi for data science - better conda ecosystem support
  • Commit lock files (uv.lock or pixi.lock) to version control
  • Document dependencies in pyproject.toml or pixi.toml