Uv

uv - Fast Python Package and Project Manager

Overview

This skill provides guidance for using uv, an extremely fast Python package and project manager written in Rust. uv replaces pip, pip-tools, pipx, poetry, pyenv, virtualenv, and more while being 10-100x faster. The skill focuses particularly on uv run for executing code within isolated project environments.

When to Use This Skill

Use this skill when:

• Working in a Python project with a pyproject.toml file
• Running Python scripts or CLI tools that require dependencies
• Managing project dependencies and virtual environments
• Executing commands within an isolated project environment
• Creating new Python projects or initializing scripts
• Converting from pip, poetry, or other Python tooling to uv

Core Concept: uv run

The most important command in uv is uv run, which executes commands within an isolated project environment.

Why Use uv run

ALWAYS use uv run instead of bare python or direct script execution when:

1. Running Python scripts that depend on project dependencies
2. Executing Python CLI tools installed in the project
3. Working within a uv-managed project (has pyproject.toml)
4. Running commands that need access to the project's virtual environment

Key behavior: uv run automatically ensures the project environment is current before executing the command. It installs the project into .venv and keeps it isolated from the system Python.

Basic Usage

# Run a Python script
uv run example.py

# Run a Python module
uv run python -m mymodule

# Run a CLI tool from project dependencies
uv run pytest
uv run ruff check
uv run black .

# Run external commands that need project access
uv run bash scripts/deploy.sh

Adding Dependencies Per-Invocation

Use --with to include or override dependencies for a single execution without modifying pyproject.toml:

# Add a dependency temporarily
uv run --with httpx python -c "import httpx; print(httpx.__version__)"

# Specify version constraints
uv run --with 'httpx==0.26.0' python script.py

# Multiple dependencies
uv run --with httpx --with rich python script.py

This is useful for:

• Testing code with different dependency versions
• One-off operations requiring special packages
• Running scripts without permanently adding dependencies

Running Scripts with Inline Dependencies

Python scripts can declare their own dependencies using inline metadata. This creates isolated environments separate from the project:

# /// script
# dependencies = [
#   "requests<3",
#   "rich",
# ]
# requires-python = ">=3.12"
# ///

import requests
import rich

# Script code here...

Important: When a script has inline metadata, uv run uses ONLY those dependencies and ignores the project's dependencies, even when run inside a project directory.

Initialize a script with metadata:

# Create script with inline metadata
uv init --script example.py --python 3.12

# Add dependencies to script
uv add --script example.py 'requests<3' 'rich'

# Run the script (uses its own dependencies)
uv run example.py

Common Tasks and Workflows

Creating a New Project

# Create a new application project
uv init my-project
cd my-project

# Create a library project
uv init --lib my-library

# Create project in current directory
uv init

# Specify Python version
uv init --python 3.12

This creates:

• pyproject.toml - Project configuration and dependencies
• .python-version - Python version specification
• README.md - Basic documentation
• src/my_project/ - Source code directory (for apps)

Managing Dependencies

# Add a dependency
uv add requests

# Add with version constraint
uv add 'httpx>=0.25,<0.27'

# Add development dependency
uv add --dev pytest

# Add optional dependency group
uv add --optional docs sphinx

# Remove a dependency
uv remove requests

# Update all dependencies
uv lock --upgrade

# Sync environment to match lockfile
uv sync

Virtual Environment Management

# uv automatically creates and manages .venv/
# Most operations don't require manual venv activation

# To manually create a venv (rarely needed)
uv venv

# To activate manually (usually unnecessary)
source .venv/bin/activate  # Linux/Mac
.venv\Scripts\activate     # Windows

Best practice: Use uv run instead of manually activating virtual environments. This ensures environment consistency and automatic updates.

Python Version Management

# Install a specific Python version
uv python install 3.12

# List available Python versions
uv python list

# Find Python installations
uv python find

# Pin project to specific Python version
uv python pin 3.12

Running Tools Without Installation

# Run a tool ephemerally (like pipx)
uvx ruff check .
uvx black --check .

# Equivalent to:
uv tool run ruff check .

# Install tool globally
uv tool install ruff

pip-Compatible Commands

uv provides drop-in pip replacements:

# Install packages (like pip install)
uv pip install requests

# Generate requirements.txt (like pip freeze)
uv pip freeze > requirements.txt

# Install from requirements.txt
uv pip install -r requirements.txt

# Compile dependencies (like pip-compile)
uv pip compile pyproject.toml -o requirements.txt

# Show dependency tree
uv pip tree

Decision Tree: Which Command to Use

Need to run code in this project? → Use uv run <command>

Need to add a package to this project? → Use uv add <package>

Need to run a tool once without installing? → Use uvx <tool> or uv tool run <tool>

Need to create a new project? → Use uv init [project-name]

Need to update dependencies to latest versions? → Use uv lock --upgrade

Need to sync environment after pulling changes? → Use uv sync

Need to install a specific Python version? → Use uv python install <version>

Working with legacy pip workflows? → Use uv pip commands as drop-in replacements

Key Differences from Other Tools

vs pip

• 10-100x faster due to Rust implementation
• Automatic virtual environment management
• Built-in lockfile support
• No need for pip-tools or virtualenv

vs poetry

• Faster dependency resolution
• Compatible with standard pyproject.toml
• Simpler command structure
• No separate poetry.lock format (uses uv.lock)

vs pipx

• uvx provides same functionality
• Faster execution
• Better caching and deduplication

vs pyenv

• uv python manages Python versions
• Integrated with project management
• Faster installation and switching

Best Practices

1. Always use uv run in projects - Don't manually activate virtual environments; let uv handle environment management automatically.

2. Commit lockfiles - Include uv.lock in version control for reproducible builds.

3. Pin Python versions - Use uv python pin to specify project Python requirements.

4. Use --with for experiments - Test dependencies temporarily without modifying pyproject.toml.

5. Leverage inline metadata for scripts - Single-file scripts should declare dependencies inline for portability.

6. Use uvx for one-off tools - Don't pollute project dependencies with tools you run occasionally.

7. Let uv manage environments - Trust uv sync to keep environments in sync; avoid manual pip installs in uv projects.

8. Check .python-version files - Respect project Python version specifications.

Common Patterns

Running Tests

# Add pytest as dev dependency
uv add --dev pytest pytest-cov

# Run tests
uv run pytest

# Run with coverage
uv run pytest --cov

Linting and Formatting

# Add dev tools
uv add --dev ruff black mypy

# Run tools
uv run ruff check .
uv run black .
uv run mypy src/

Building and Publishing

# Build distributions
uv build

# Publish to PyPI
uv publish

# Publish to test PyPI
uv publish --index-url https://test.pypi.org/legacy/

Working with Scripts

# Create standalone script with dependencies
uv init --script analyze.py
uv add --script analyze.py pandas numpy

# Run script (uses its inline dependencies)
uv run analyze.py

# Lock script dependencies for reproducibility
uv lock --script analyze.py

Troubleshooting

Command not found after uv add

Problem: Added a package with a CLI tool, but command doesn't work.

Solution: Use uv run <command> to execute within project environment:

uv add ruff
uv run ruff check .  # Not just "ruff check ."

Environment out of sync

Problem: Dependencies changed but environment hasn't updated.

Solution: Run uv sync to synchronize:

uv sync

Wrong Python version

Problem: Project requires different Python version.

Solution: Install and pin correct version:

uv python install 3.12
uv python pin 3.12

Script ignores project dependencies

Problem: Script with inline metadata doesn't use project packages.

Solution: This is intentional. Scripts with inline metadata are isolated. Either:

• Remove inline metadata to use project dependencies
• Add needed dependencies to script's inline metadata

References

For detailed command options and advanced usage, see references/cli_reference.md.