Python Package Template

Do you like building python packages? Do you hate configuring all of the things? This is for you.

Create a new folder with your package name (this makes it easier to infer the package name from the folder name via copier scripts):

mkdir your-project
cd your-project

Then run the copier script to set up the package:

uv tool run --with jinja2_shell_extension copier@latest copy --trust --vcs-ref=HEAD https://github.com/iloveitaly/python-package-template .

What is Provided?

Here are the key tools and features provided:

1. uv. Configured in the .envrc file with layout uv.
2. Justfile. Provides commands for setting up (just setup) and cleaning (just clean) the project environment.
3. GitHub Actions. Repo metadata sync, package build and publish.
4. Dependabot.:
5. VS Code Settings. Some important py config that is hard to figure out sometimes.
6. CLI Script Hook. Remove if you aren't publishing a package with a CLI.
7. LLM IDE Rules. Optional setup for better AI-assisted development with GitHub Copilot, Cursor, etc.
8. GitHub Configuration. just recipes to configure your GitHub repo.

Prompts

Prompt to take this template, combined with code extracted from a project, and create an open source project.

This repo:

1. Has a generic/empty template for a python package
2. Has a file extracted from a project @extract.py

Can you:

1. $INTEGRATE_CODE_INSTRUCTIONS
2. Add any `uv add` packages required for this to run
3. Make sure all tests pass. Add some basic tests.
4. Update the @README.md
5. Update the `pyproject.toml` keywords (maximum of 4) and description based on the contents of this repo. Do not use `.` in a keyword.
6. Make sure all linting issues are corrected by running `just lint`

Trusted Publishers

This is setup to be used with PyPi trusted publishers. You can easily change this configuration by adding --token ${{ secrets.PYPI_API_TOKEN }} to the build workflows.

Related

• https://github.com/superlinear-ai/substrate