uv-dependency-management（uv 依赖管理）

uv Dependency Management

Purpose

Master dependency management with uv, including adding/removing packages, version constraints, dependency groups for development and production, lock files, and resolving version conflicts.

Quick Start

Add a dependency and see it in your lock file:

# Add a package with version constraint
uv add "requests>=2.31.0"

# View your dependency tree
uv tree

# Update all dependencies to latest compatible versions
uv lock --upgrade

Your pyproject.toml is automatically updated, and uv.lock ensures reproducible installs.

Instructions

Step 1: Understand Dependency Scopes

uv manages three types of dependencies:

Main Dependencies (project requirements)

[project]
dependencies = [
    "fastapi>=0.104.0",
    "sqlalchemy>=2.0.0",
]

Optional Dependencies (extras for features)

[project.optional-dependencies]
database = ["psycopg2-binary>=2.9.0"]
aws = ["boto3>=1.34.0"]

Development Groups (dev-only tools)

[dependency-groups]
dev = ["pytest>=8.0.0", "black>=23.0.0"]
test = ["pytest-cov>=4.1.0"]
lint = ["ruff>=0.1.0", "mypy>=1.7.0"]

Step 2: Add Dependencies

Add to main project:

uv add requests httpx
uv add "fastapi>=0.104.0"
uv add "django>=4.2,<5.0"

Add to specific group:

uv add --dev pytest                    # Add to 'dev' group
uv add --group test pytest-cov        # Add to 'test' group
uv add --group lint ruff mypy         # Add to 'lint' group

Add with extras:

uv add "pandas[excel,plot]"
uv add --dev "pytest[cov]"

Step 3: Use Version Constraints

Common patterns (semantic versioning):

uv add "requests"                   # Latest
uv add "requests>=2.31.0"          # Minimum version
uv add "requests>=2.31.0,<3.0.0"   # Range (major bump)
uv add "requests~=2.31.0"          # Compatible release (~= 2.31.x)
uv add "requests==2.31.0"          # Exact version (production)

Recommended approach:

• Development: >= (flexible, want latest)
• Production: >=X,<Y (range, tested with that version)
• Stable packages: ~= (compatible releases)

Step 4: Remove and Update Dependencies

Remove a package:

uv remove requests
uv remove pytest black              # Remove multiple

Update to specific version:

uv add "requests@2.32.0"

Update all dependencies:

uv sync --upgrade                   # Update and sync
uv lock --upgrade                   # Just update lock file

Step 5: Organize with Dependency Groups

Define logical groupings in pyproject.toml:

[dependency-groups]
dev = [
    "pytest>=8.0.0",
    "pytest-cov>=4.1.0",
    "black>=23.0.0",
]
lint = [
    "ruff>=0.1.0",
    "mypy>=1.7.0",
    "pylint>=3.0.0",
]
docs = [
    "mkdocs>=1.5.0",
    "mkdocs-material>=9.0.0",
]

Install specific groups:

uv sync --group dev --group lint    # Install groups
uv sync --all-groups                # Install everything
uv sync --no-dev                    # Only main deps (production)

Step 6: Understand Lock Files

Lock file (uv.lock):

• Records exact versions of all dependencies
• Ensures reproducible installs across machines
• Should be committed to version control
• Automatically updated when you change pyproject.toml

Workflow:

# After changing pyproject.toml
uv sync              # Installs from lock or creates new lock

# To explicitly update lock file
uv lock --upgrade    # Update all to latest compatible

# In CI/CD (use frozen to prevent surprises)
uv sync --frozen     # Fails if lock is out of sync

Step 7: Handle Dependency Conflicts

Problem: Two packages need incompatible versions of same library

Solution 1: Check if newer versions are compatible

uv add "package-a"
uv add "package-b"   # Might fail with version conflict

# Try adding with different constraints
uv add "package-a>=1.0,<2.0"
uv add "package-b>=2.0,<3.0"

Solution 2: Use separate dependency groups

[dependency-groups]
ml-cpu = ["torch-cpu>=2.0"]
ml-gpu = ["torch-gpu>=2.0"]

# Install one group at a time
uv sync --group ml-cpu

Solution 3: Investigate and pick winning version

# See what's needed
uv add --dry-run "package-a" "package-b"

# One might need updating
uv add "package-a>=2.0"
uv add "package-b>=3.0"

Examples

Example 1: Web API Project Setup

# Initialize and add web dependencies
uv init my-api
cd my-api
uv add fastapi uvicorn sqlalchemy pydantic

# Add development tools
uv add --group dev pytest pytest-asyncio
uv add --group lint ruff mypy black

# Verify setup
uv tree

Resulting pyproject.toml:

[project]
name = "my-api"
version = "0.1.0"
dependencies = [
    "fastapi>=0.104.0",
    "uvicorn>=0.24.0",
    "sqlalchemy>=2.0.0",
    "pydantic>=2.5.0",
]

[dependency-groups]
dev = ["pytest>=8.0.0", "pytest-asyncio>=0.21.0"]
lint = ["ruff>=0.1.0", "mypy>=1.7.0", "black>=23.0.0"]

Example 2: Library with Optional Features

# Create library with core deps
uv init my-library
uv add "click>=8.1.0"

# Add optional features as extras
uv add --group excel "openpyxl>=3.0.0"
uv add --group database "sqlalchemy>=2.0.0"
uv add --group async "aiohttp>=3.8.0"

# Setup development tools
uv add --dev pytest

Example 3: Resolving Version Conflicts

# Trying to add packages with conflicting requirements
uv add package-a                    # Works fine
uv add package-b                    # Fails - needs different version of shared lib

# View what's going on
uv add --dry-run package-b

# Try different constraints
uv add "package-a>=1.0,<1.5"
uv add "package-b>=2.0,<2.5"

# Check if it works
uv tree

Example 4: Data Science Project

# Create project
uv init data-pipeline
cd data-pipeline

# Data science core
uv add pandas numpy scikit-learn

# Analysis tools
uv add jupyter matplotlib seaborn plotly

# Development
uv add --group dev pytest pytest-cov

# Final tree
uv tree

Example 5: Update Strategy for Production

# Initial setup with compatible ranges
uv add "requests>=2.31.0,<3.0.0"
uv add "fastapi>=0.104.0,<1.0.0"

# After testing on latest patch
uv add "requests@2.31.3"
uv add "fastapi@0.104.1"

# Lock file is frozen for production
git add uv.lock
git commit -m "Pin exact versions for production"

# CI/CD uses frozen lock
uv sync --frozen

Example 6: Monorepo with Path Dependencies

# Main project
uv init main-app
cd main-app

# Add local packages (editable)
uv add --editable ../shared-lib
uv add --editable ../utils-lib

# Now imports work from both packages

Requirements

• uv installed (install: curl -LsSf https://astral.sh/uv/install.sh | sh)
• Understanding of PEP 508 version specifiers (recommended)
• Project with pyproject.toml (created by uv init)
• Python 3.8+ available

See Also

• uv-project-setup - Initial project creation
• uv-python-version-management - Managing Python versions
• uv-troubleshooting - Resolving dependency issues
• uv Documentation - Official guide