How do you manage dependencies and virtual environments in Python?

Proper dependency management prevents conflicts and ensures reproducible environments.

Virtual environments isolate each project's packages. python -m venv .venv creates one, source .venv/bin/activate enables it. Each project has its own package versions without global conflicts. Always use virtual environments—never install packages globally.

pip is the basic package installer. pip install package adds packages, pip freeze > requirements.txt captures versions. requirements.txt lists dependencies but doesn't distinguish direct from transitive dependencies.

Poetry is a modern alternative handling dependencies, virtual environments, and packaging. pyproject.toml defines your project and dependencies. poetry.lock pins exact versions for reproducibility. poetry add package adds dependencies while managing the lock file.

pip-tools bridges pip and lock files: requirements.in lists your direct dependencies, pip-compile generates requirements.txt with pinned transitive dependencies.

For production, pin versions precisely. requirements.txt should include exact versions (package==1.2.3) or use lock files. This ensures deployments match your tested environment.

Development dependencies (testing, linting) stay separate from production. Poetry supports dependency groups; with pip, use requirements-dev.txt.

Docker environments should install dependencies before copying code—this caches the dependency layer, speeding up rebuilds when only code changes.