When setting up Python projects, the traditional workflow of using pip, virtualenv, and pip-tools can be slow.

uv, a modern Rust-based package manager, is a lot faster.

Why UV?

• Up to 8-10x faster than pip for package installation
• Built in Rust for maximum performance
• Compatible with existing Python tooling
• Automatic dependency resolution

Installation

For macOS and Linux:

curl -LsSf https://astral.sh/uv/install.sh | sh

For Windows:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Quick Project Templates

Here are two templates you can use directly in your terminal for quick project setup:

mkdir -p src/$(basename $(pwd)) && \
touch src/$(basename $(pwd))/__init__.py && \
touch src/$(basename $(pwd))/main.py && \
echo "[project]
name = \"$(basename $(pwd))\"
version = \"0.1.0\"
requires-python = \">=3.12\"
dependencies = []

[tool.hatch.build.targets.wheel]
packages = [\"src/$(basename $(pwd))\"]" > pyproject.toml && \
uv venv && \
source .venv/bin/activate

This basic template sets up:

1. A src/ directory with your project name (based on current directory)
2. Initializes a Python package with __init__.py and main.py
3. Creates a pyproject.toml with basic configuration
4. Sets up a UV virtual environment and activates it

mkdir -p src/$(basename $(pwd))/models src/$(basename $(pwd))/data src/$(basename $(pwd))/notebooks && \
touch src/$(basename $(pwd))/__init__.py && \
touch src/$(basename $(pwd))/main.py && \
touch src/$(basename $(pwd))/app.py && \
touch src/$(basename $(pwd))/notebooks/exploration.ipynb && \
echo "[project]
name = \"$(basename $(pwd))\"
version = \"0.1.0\"
requires-python = \">=3.12\"
dependencies = []

[tool.hatch.build.targets.wheel]
packages = [\"src/$(basename $(pwd))\"]" > pyproject.toml && \
uv venv && \
source .venv/bin/activate

This ML research template creates:

1. A full ML project structure with dedicated directories for:
   • models/: Trained models and model definitions
   • data/: Dataset storage and processing
   • notebooks/: Jupyter notebooks for experimentation
2. Adds app.py for Streamlit dashboard development
3. Creates an empty Jupyter notebook for exploration
4. Sets up the same basic Python package structure with UV

You can save these templates in two ways:

1. Traditional Shell Aliases: Add them to your shell configuration file (.zshrc or .bashrc):

# Add to your .zshrc or .bashrc
alias pyquick='[first template above]'
alias pyml='[second template above]'

# Then reload your configuration
source ~/.zshrc  # or source ~/.bashrc

2. Raycast Snippets (Recommended): I prefer using Raycast to manage these snippets:

• Create two snippets:
  • Name: "Python Quick Project", Keyword: pyquick
  • Name: "Python ML Project", Keyword: pyml
• Paste the respective template code into each snippet
• Trigger anywhere with pyquick or pyml keyword

Installing Packages with UV

After setting up your project with either snippet, you can quickly add dependencies using UV's faster package installation:

Instead of using pip install pandas numpy, use:

uv add pandas numpy torch

For dev dependencies like pytest:

uv add --dev pytest black

Installing from requirements.txt:

uv pip install -r requirements.txt

UV's package installation is significantly faster than pip, especially when installing multiple packages simultaneously. The uv add command will automatically:

• Update your virtual environment
• Add the package to your dependencies
• Handle dependency resolution more efficiently than pip

I recommend trying uv out!