I'm trying out

uv

as a venv backend, and it has been great up until running the

notebook

utility. Some of our notebooks contain an initial cell to

pip install

dependencies to the venv - with

uv

, this no longer works properly because

pip

isn't bundled with the venv, so dependencies are installed to the system Python instead and then cannot be imported later on in the context of the venv Python. I'm keen to avoid

uv

as a direct dependency of our notebooks, so I've been looking at other ways to work around this issue. There is a

--seed

option for uv venv that would make

pip

available in the venv, but this does seem to slow down venv creation a bit, so I can see why this isn't the default in the Meltano implementation. Is there any way to make this configurable, or an alternative approach that might work?

Yeah, installing uv in the plugin's venv is less than ideal and goes counter to uv's philosophy. If

--seed

had a corresponding env var, it'd be really easy. It's probably worth raising an issue with the Astral folks.

installing uv in the plugin's venv is less than ideal and goes counter to uv's philosophy

To clarify, do you mean installing

pip

in the venv here?

Oh sorry, installing

pip

is alright.

I'm keen to avoid

uv

as a direct dependency of our notebooks

That would require installing

uv

in the plugin's venv, and that's was trying to refer to.

Gotcha, thanks. 👍

Thanks for creating it 👍. I also commented on the issue: another option might be to add

pip

to the

pip_url

of the notebook utility. Would that work?

Thanks, just testing that out now - I'll reply in the issue