Meta recently released SAM-Audio, an exciting new audio LLM capable of separating sounds from complex mixtures using text or visual prompts. Like many research models, it isn’t installable out of the box for macOS users. To get it running on my M1 Mac was a bit of an adventure, but I managed in the end.

Here’s a copier project template that sets up everything I need for a modern Python CLI project in about 30 seconds.

If you have the Python extension in VSCode, you can open and run Jupyter notebook directly in the IDE. But what if you want to work on the actual JSON source?

While looking at ways to run MLflow with uv on Github, I came across this gem of a comment which is the perfect, most compact introduction to uv

I finally had the chance to test Astral’s uv tool. It’s very good, from what I’ve seen so far.

pip freeze > requirements.txt is the canonical way to capture all the packages you have installed in a Python project. I am not sure why; it doesn’t do what you expect.