The README proposes using `pip install futa`, but there is no actual project published on PyPI for this to work - nor any releases in the repository (you'd have to build the wheel yourself). Although given that they once famously took down a "slut" package I have to wonder if they'd actually accept "futa". Although they did accept "fuckit", so.

I did notice that you have a Python interface to ollama listed in requirements.txt (which doesn't really help with packaging anyway) but aren't using it in the code. (In case it's what you were looking for: there aren't currently any ways to describe dependencies that are installed for the system and aren't available as Python packages, but PEP 725 could mean they happen eventually, maybe.) Also, you can use `text=True` (or specify an `encoding` or `errors`) in the arguments for the `subprocess.run` call, to get textual stdout/stderr directly rather than decoding them afterward; and `capture_output=True` is an easier way to get both streams than setting them both to `PIPE`. And you can iterate over a file in text mode directly (which gives a line at a time), rather than reading the whole thing and calling `.splitlines`, simplifying further:

  denials = [line.strip() for line in f if line.strip()]

Anyway, I appreciate the commitment to the bit.

Unfortunate choice of project name should you google this with safe search disabled