Uv

uv Skill

Extremely fast Python package and project manager by Astral (Ruff creators).

Overview

uv is a single tool that replaces:

• pip/pip-tools - Package installation and dependency resolution
• virtualenv/venv - Virtual environment creation
• pyenv - Python version management
• pipx - Tool installation and execution
• poetry/pdm - Project and dependency management
• twine - Package publishing

Key Features:

• 10-100x faster than pip
• Universal lockfile (uv.lock) for reproducible builds
• Automatic Python version management
• Built-in tool execution (uvx)
• PEP 723 inline script dependencies
• Drop-in pip compatibility

Quick Reference

Quick Start

Installation

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

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

# Via pip/pipx
pipx install uv
pip install uv

# Homebrew
brew install uv

Shell Completion

# Bash
echo 'eval "$(uv generate-shell-completion bash)"' >> ~/.bashrc

# Zsh
echo 'eval "$(uv generate-shell-completion zsh)"' >> ~/.zshrc

# Fish
echo 'uv generate-shell-completion fish | source' > ~/.config/fish/completions/uv.fish

Essential Commands

1. Starting a Project

# Create new project
uv init my-project          # Application (default)
uv init --lib my-library    # Library (src layout, build backend)
uv init --app my-app        # Explicit application

# Set Python version
uv python pin 3.12          # Creates .python-version

# Install Python if needed
uv python install 3.12

2. Managing Dependencies

# Add packages
uv add requests flask       # Production dependencies
uv add --dev pytest ruff    # Development dependencies
uv add --group test pytest  # Specific dependency group
uv add --optional api flask # Optional extra
uv add "httpx>=0.20"        # With version constraint

# Remove packages
uv remove requests
uv remove --dev pytest

# Update packages
uv lock --upgrade                    # All packages
uv lock --upgrade-package requests   # Single package

# View dependencies
uv tree                     # Full tree
uv tree --depth 2           # Limited depth

3. Syncing Environment

# Install all dependencies
uv sync                     # Default (includes dev)
uv sync --locked            # CI/production (strict)
uv sync --frozen            # Don't update lockfile
uv sync --no-dev            # Exclude dev dependencies
uv sync --all-extras        # Include optional extras

4. Running Code

# Run in project environment
uv run python script.py
uv run pytest
uv run flask run

# Run with temporary dependency
uv run --with pandas script.py

# Run tools without installing (uvx)
uvx ruff check .
uvx black --check .
uvx --from httpie http https://example.com

5. Python Version Management

# Install Python versions
uv python install           # Latest version
uv python install 3.12      # Specific version
uv python install 3.11 3.12 3.13  # Multiple

# List versions
uv python list
uv python list --only-installed

# Pin version
uv python pin 3.12          # Project (.python-version)
uv python pin --global 3.12 # User default

# Find Python
uv python find
uv python find ">=3.11"

6. Virtual Environments

# Create (usually automatic)
uv venv                     # Creates .venv
uv venv my-env              # Custom name
uv venv --python 3.12       # Specific Python

# Activate (optional - uv run auto-detects)
source .venv/bin/activate   # macOS/Linux
.venv\Scripts\activate      # Windows

7. Global Tools

# Install tools globally
uv tool install ruff
uv tool install "ruff==0.5.0"
uv tool install --python 3.12 mypy

# Manage tools
uv tool list
uv tool upgrade ruff
uv tool upgrade --all
uv tool uninstall ruff

# Setup PATH
uv tool update-shell

Scripts with Inline Dependencies (PEP 723)

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

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

# Run script (dependencies auto-installed)
uv run example.py

Script format:

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

import requests
from rich import print
print(requests.get("https://api.example.com").json())

pip-Compatible Interface

# Install packages (requires virtual environment)
uv pip install flask
uv pip install -r requirements.txt
uv pip install -e .         # Editable install

# Uninstall
uv pip uninstall flask

# Compile requirements
uv pip compile requirements.in -o requirements.txt

# Sync from requirements
uv pip sync requirements.txt

# Freeze environment
uv pip freeze

Building and Publishing

# Build distributions
uv build                    # Creates dist/ with wheel and sdist

# Publish to PyPI
uv publish                  # Requires authentication setup

# Authenticate
uv auth login pypi          # Interactive login
uv auth login pypi --token  # API token

Project Structure

my-project/
├── pyproject.toml          # Project definition (required)
├── uv.lock                 # Lock file (auto-generated)
├── .venv/                  # Virtual environment (gitignored)
├── .python-version         # Python version pin (optional)
└── src/
    └── my_project/
        └── __init__.py

pyproject.toml Example

[project]
name = "my-project"
version = "0.1.0"
description = "My awesome project"
readme = "README.md"
requires-python = ">=3.11"
dependencies = [
    "requests>=2.28",
    "click>=8.0",
]

[project.optional-dependencies]
api = ["fastapi", "uvicorn"]
dev = ["pytest", "ruff"]

[project.scripts]
my-cli = "my_project.cli:main"

[dependency-groups]
dev = ["pytest>=8", "ruff", "mypy"]
test = ["pytest-cov"]
docs = ["sphinx", "myst-parser"]

[tool.uv]
dev-dependencies = ["pytest", "ruff"]  # Alternative to dependency-groups
default-groups = ["dev"]

[tool.uv.sources]
# Git dependency
my-lib = { git = "https://github.com/user/my-lib" }
# Local path
local-pkg = { path = "./packages/local-pkg" }
# Specific index
torch = { index = "pytorch" }

[[tool.uv.index]]
name = "pytorch"
url = "https://download.pytorch.org/whl/cpu"
explicit = true

Configuration

Configuration Files

Project-level (highest priority):

• uv.toml - Standalone config (preferred)
• pyproject.toml - Under [tool.uv] section

User-level:

• ~/.config/uv/uv.toml (macOS/Linux)
• %APPDATA%\uv\uv.toml (Windows)

Key Environment Variables

Common Workflows

New Project Setup

# Create and enter project
uv init my-project
cd my-project

# Add dependencies
uv add flask sqlalchemy
uv add --dev pytest ruff mypy

# Run application
uv run flask run

# Run tests
uv run pytest

Existing Project (from requirements.txt)

# Initialize uv project
uv init

# Import dependencies from requirements.txt
uv add $(cat requirements.txt | grep -v "^#" | tr '\n' ' ')

# Or use pip interface
uv venv
uv pip install -r requirements.txt

CI/CD Pipeline

# GitHub Actions
- uses: astral-sh/setup-uv@v7
  with:
    enable-cache: true
- run: uv sync --locked
- run: uv run pytest

# GitLab CI
variables:
  UV_CACHE_DIR: .uv-cache
  UV_LINK_MODE: copy
image: ghcr.io/astral-sh/uv:latest
script:
  - uv sync --locked
  - uv run pytest

Docker

FROM python:3.12-slim
COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/

WORKDIR /app
COPY pyproject.toml uv.lock ./

# Install dependencies only (for caching)
RUN --mount=type=cache,target=/root/.cache/uv \
    uv sync --locked --no-install-project

# Copy source and install project
COPY . .
RUN --mount=type=cache,target=/root/.cache/uv \
    uv sync --locked

ENV PATH="/app/.venv/bin:$PATH"
CMD ["python", "-m", "my_app"]

direnv Integration

# .envrc
if has uv; then
  VIRTUAL_ENV="$(pwd)/.venv"
  if [[ ! -d "$VIRTUAL_ENV" ]]; then
    uv venv
  fi
  PATH_add "$VIRTUAL_ENV/bin"
  export VIRTUAL_ENV
fi

Migration Guide

From pip + requirements.txt

# Option 1: Initialize new uv project and import dependencies
uv init
# Extract package names from requirements.txt
cat requirements.txt | grep -v "^#" | grep -v "^-e" | \
  cut -d'=' -f1 | cut -d'>' -f1 | xargs uv add

# Option 2: Keep using requirements.txt with uv's pip interface
uv venv
uv pip install -r requirements.txt

# Option 3: Generate lockfile from requirements.txt
uv pip compile requirements.in -o requirements.txt

Key differences:

• uv.lock replaces requirements.txt for locking
• pyproject.toml replaces requirements.in for declaring dependencies
• Use uv run instead of activating virtualenv

From Poetry

# uv can read Poetry's pyproject.toml format
cd existing-poetry-project

# Initialize uv (keeps existing pyproject.toml)
uv sync

# Poetry sections are automatically recognized:
# [tool.poetry.dependencies] -> project dependencies
# [tool.poetry.group.dev.dependencies] -> dev dependencies

Migrate pyproject.toml (optional but recommended):

# Poetry format
[tool.poetry.dependencies]
python = "^3.11"
requests = "^2.28"

[tool.poetry.group.dev.dependencies]
pytest = "^8.0"

# Standard format (uv native)
[project]
requires-python = ">=3.11"
dependencies = ["requests>=2.28"]

[dependency-groups]
dev = ["pytest>=8.0"]

Command equivalents:

From Pipenv

# Convert Pipfile to requirements.txt first
pipenv requirements > requirements.txt
pipenv requirements --dev > requirements-dev.txt

# Initialize uv project and import
uv init
cat requirements.txt | grep -v "^#" | xargs uv add
cat requirements-dev.txt | grep -v "^#" | xargs uv add --dev

# Remove old Pipenv files after verification
rm Pipfile Pipfile.lock

Command equivalents:

From pyenv (Python version management only)

# Install Python versions with uv instead
uv python install 3.11 3.12 3.13

# Pin version for project (creates .python-version)
uv python pin 3.12

# List installed versions
uv python list --only-installed

# uv reads existing .python-version files from pyenv

Note: You can use pyenv and uv together - uv will detect pyenv-installed Pythons.

From conda

uv does not replace conda for:

• Non-Python dependencies (C libraries, CUDA, etc.)
• Conda-specific packages not on PyPI

For pure Python projects:

# Export conda environment
conda list --export > conda-packages.txt

# Extract Python packages (filter conda-specific ones)
grep -v "^#" conda-packages.txt | grep -v "conda" | cut -d'=' -f1 > packages.txt

# Initialize uv and add packages
uv init
uv add $(cat packages.txt | tr '\n' ' ')

Hybrid approach: Use conda for system dependencies, uv for Python packages:

# Conda for non-Python deps
conda create -n myenv python=3.12 cudatoolkit

# Activate conda env, then use uv
conda activate myenv
export UV_SYSTEM_PYTHON=1
uv pip install -r requirements.txt

Common Pitfalls

1. Forgetting --locked in CI/Production

# Wrong - may update lockfile unexpectedly
uv sync

# Correct - fails if lockfile is outdated (reproducible builds)
uv sync --locked

2. Mixing uv and pip Commands

# Don't do this - breaks uv's dependency tracking
pip install some-package
source .venv/bin/activate && pip install another-package

# Do this instead - uv tracks all dependencies
uv add some-package
# Or use uv's pip interface if needed
uv pip install some-package

3. Not Committing uv.lock

The uv.lock file must be committed to version control for reproducible builds. Without it, environments may resolve differently.

# .gitignore - DON'T ignore uv.lock
.venv/
__pycache__/
# uv.lock  <-- DO NOT ADD THIS LINE

4. Running Commands Outside Project Environment

# Wrong - uses system Python, not project environment
python script.py
pytest

# Correct - runs within project's virtual environment
uv run python script.py
uv run pytest

5. Editing uv.lock Manually

Never edit uv.lock by hand. It's auto-generated and managed by uv.

# To update a specific package
uv lock --upgrade-package requests

# To upgrade all packages
uv lock --upgrade

# To regenerate from scratch
rm uv.lock && uv lock

6. Using --frozen When You Mean --locked

# --frozen: Don't update lockfile, but don't verify it either
uv sync --frozen

# --locked: Verify lockfile matches pyproject.toml (use this in CI)
uv sync --locked

Troubleshooting

Common Issues

Debug Commands

# Check current environment
uv python find
uv pip list

# Verbose output
uv sync -v
uv sync -vv  # More verbose

# Check lockfile status
uv lock --check

# Clear cache
uv cache clean
uv cache prune --ci  # For CI environments

References

• references/cli-commands.md - Complete CLI reference
• references/project-management.md - Project and dependency management
• references/python-versions.md - Python version management
• references/integrations.md - Docker, CI/CD, and tool integrations

External Links

• Official docs: https://docs.astral.sh/uv/
• GitHub: https://github.com/astral-sh/uv
• Docker images: https://github.com/astral-sh/uv/pkgs/container/uv
• Changelog: https://github.com/astral-sh/uv/blob/main/CHANGELOG.md

Absorbed sub-skill (post-consolidation)

This skill now subsumes the former python-project skill (project scaffolding). Original content preserved under:

Gotchas

• uv pip install -e . works but doesn't write the editable install to uv.lock — the lockfile is wrong on next uv sync. Use uv add --editable instead.
• uv sync --frozen fails if dependencies upgraded outside the lock — won't auto-update; emit a clear error message but blocks CI until you uv lock --upgrade-package <name>.
• uv venv creates .venv in CWD — missing --directory after cd puts the venv in the wrong place. Pin with UV_PROJECT_ENVIRONMENT env var or absolute paths in scripts.
• uv tool install and uv pip install go to DIFFERENT environments — uv tools are isolated per-tool; pip-installed packages aren't visible to uv-installed tools.
• Pin via ==X.Y.Z in pyproject.toml doesn't lock; only uv.lock does — dependencies = ["foo==1.2.3"] allows uv to pick any compatible 1.2.3, including yanked versions.