I’m at the point where I’m migrating all my projects to uv, and new Python projects don’t use any other package manager.

I finally got around to migrating this site to use it, using the very handy migrate-to-uv tool. So it’s time to update my recommended Python setup.

My old Python setup was very much built around the complications of managing environments and dependencies, and the conflicting set of tools to deal with those two problems. There are still a few places where I’ll use pipx, but otherwise everything is on uv.

This guide is still aimed at a recent Apple or Linux computer, or WSL if you’re on Windows. I’m writing this on a MacBook Pro with an M2 chip, if that matters to you.

Tools and helpful links:

• Python docs: https://docs.python.org/3/
• Python Standard Library: https://docs.python.org/3/library/index.html - Start here when you’re trying to solve a specific problem
• uv: https://docs.astral.sh/uv/ - For packaging, dependency and environment management, plus standalone scripts
• pipx: https://pipxproject.github.io/pipx/ - Global script installer, built on top of pip
• homebrew: https://brew.sh/ - Package manager for macOs, use for specific needs
• Postgres.app: https://postgresapp.com/ - Preferred database server for macOs

You probably don’t need both uv and pipx. I have a bunch of existing tools I installed with pipx, and those work fine, so I haven’t migrated them to uvx.

There is one set of tools that stays on pipx, though: Datasette and its SQLite toolchain. Simon Willison built those to install their own plugins, using datasette install <plugin> or llm install <plugin>. Those use pip internally and sometimes uv can cause problems upgrading, so I’ve kept them on pipx.

Installing the right Python

Use uv and nothing else for this. Run uv python list to see what’s already installed or otherwise available. If you’re not using pipx, it’s fine to just let uv install the right version of Python for each project.

If you want a specific version of Python installed globally, use uv python install <version>. The docs are good.

For pipx, stick to my instructions from a couple years ago:

pip install --user pipx # install it
pipx ensurepath # make sure your system can find it

That’s assuming your system already comes with a vesion of Python and pip installed. If not, try Homebrew. Maybe it’s better now, especially with uv managing everything else.

Virtual environments and local dependencies

Everything is now part of uv. Run uv init to create a project, uv add for each dependency and uv sync to install everything from an existing project.

Use uv run to run scripts inside the virtual environment that uv creates.

This is easier now

I never managed to write the post about why Python’s setup is so hard. It ultimately comes down to dependencies, both libraries and Python itself. For the most part, uv has made this a non-issue. It’s also significantly faster than the tools it replaced, which means I can iterate faster and don’t lose focus waiting for dependencies to download and install.

Now, to migrate more projects …