Two basic questions regarding CLI use and global package installation

[fyalcin]

Hey all,

I've mainly been using conda to handle my python environments but rye seemed interesting so I started playing around with it. There are a few things that I do with python that I can't seem to manage with rye though, as much as I tried to troubleshoot it I'm not getting anywhere. So I decided to turn to the community for help. Here I describe the two situations where I am having trouble with rye;

1- I have a project in which I use streamlit for a simple GUI. Usually, I just would have streamlit installed in the conda environment that is active and I would run streamlit run app.py and be done with it. With rye, even when streamlit is installed and can be imported in a python console, I don't have access to the command line tool that comes with the installation. How would one solve this? If instead I install streamlit as a tool with rye install streamlit and have access to the command line tool, since this tool runs separately from the project it does not have access to the other libraries installed in the project.

2- I work with multiple different projects but there are certain libraries that I install in every python environment I use (for instance, pandas). When I installed rye, it asked if I wanted to use a python managed by rye as the system python and I chose yes. Is there a way to install certain packages into this "base" python environment so I have access to them all the time, even outside rye-managed python projects?

I hope my questions are clear and I appreciate anyone who can help me with these. Thanks!

[geofft]

Hi! I think at this point if you're starting new you should check out uv instead of Rye, which is much more actively maintained and is explicitly the successor to Rye. Rye is mostly still around for people who are using specific features that aren't yet in uv, a list that has gotten pretty small.

Here's how I'd answer these questions for uv:

1- If you're inside a project directory, uv run streamlit will run the streamlit command from the current environment. Or you can activate the environment with souce .venv/bin/activate and then streamlit will be there.

Also, a fun trick is that if your Streamlit app happens to be one file, you can do uvx streamlit run yourapp.py, or uvx --with numpy streamlit run yourapp.py if you need dependencies, without having to create a project at all. This command will create a temporary environment when you run it.

2- No, largely because, what should it do if you have a project that uses a different version of pandas, or has a conflict with pandas? How do you make sure you don't accidentally put an import pandas into your code without declaring the dependency, and it works for you but not someone else working on the same project? (For what it's worth the conda folks also really wish they could disable the ability to install into the base environment, but too many people have workflows where they do, see e.g. conda/conda#12245 which calls it "malpractice" :) )

However, with uv, you can do uvx --with pandas python to quickly get a shell, and inside a project you can do uv run --with pandas python to get your project dependencies and also pandas. This will also do the same thing of creating a temporary environment (if needed) with your project dependencies as well as the additional dependencies you specify on the command line. (And if there is a dependency conflict, it will tell you.)