Package Management in Python

pip is the installer that ships with Python. uv offers a pip-compatible interface that is much faster, and also manages whole projects with a uv.lock lockfile, Python versions and tools. Poetry was the first widely used tool to manage dependencies, lockfiles and publishing from pyproject.toml, and remains common. Hatch manages environments, scripts and builds, and pipx installs command-line applications into isolated environments. conda installs binary packages beyond Python, such as C libraries and CUDA toolkits, which makes it popular for scientific and GPU stacks.

Build backends turn your project into wheels and source distributions, and pyproject.toml declares which one to use. setuptools is the historical default and still the most widely used. hatchling is Hatch's backend, poetry-core is Poetry's backend and works without Poetry installed, and flit-core is a minimal backend for pure-Python packages. uv also provides its own backend. For a pure-Python library, any of them works; extensions in C or Rust need a backend that supports compiled code.