Interactive notebooks teaching how to build AI applications with Python and modern models.
This is teaching material, not a product. The bar is that a learner can clone it and run it. Optimise for a working first command and a clear error when something is missing, not for architecture.
make help
make setup
make install
make dev
make runAll targets: help, setup, install, dev, run, build, up, down, logs, shell, restart, validate, test, clean ....
Python project (pyproject.toml). Use the repo's own virtualenv; never install into the system
interpreter.
- Never break the first-run path. A learner hitting an error in step one abandons the workshop. If you change setup, run it from a clean clone before calling it done.
- Explanations belong in the README and in notebook prose, not in long code comments. The code is the lesson; it should read as the clearest version of the idea.
- No em dashes in prose, per the house rules in
/Users/param/.claude/skills/lwp-shared/house-rules.md. - Pin what you can. A workshop that worked last term and breaks this term because an upstream package moved is the most common failure here.