Using UV to manage the system installation

[guillemc23]

I don't think that this is an issue, it's most likely a question.

In my case, I am using the GDAL official docker image that already comes with its bindings and Python packages properly installed. I would like to use UV to manage this system's Python installation and to add other packages on top of it. I have seen the flag --system but I'm not certain what it does, it definitely doesn't let me to do this.

For example, the system's installation has the gdal package properly installed but not pandas, so I would like to use UV to install pandas on the system's installation and run my scripts without the need for creating a virtual environment.

I have tested as well setting the following environment variables:

ENV UV_SYSTEM_PYTHON "true"
ENV UV_PYTHON_PREFERENCE "only-system"

But still, when I run uv run --frozen python main.py, it creates a virtual environment instead of running over the system instalation. This also happens when running uv sync 🤔

FYI: GDAL is a geospatial set of tools that are usually quite tedious to install, so using the official docker image is a practical way to do so, better than just starting from the Python image and installing GDAL to it

[Super1Windcloud]

I think that you can uv pip install --system

[guillemc23]

What about running something using the system python?

[charliermarsh]

You can set UV_PROJECT_ENVIRONMENT to a system interpreter -- see here: #6834 (comment)