Multiple Environments, Varying Libraries, Same Python Version?

[rontarrant]

My situation:
I build GUI utilities and I use either Kivy, PySide6, or Tkinter, depending on the project. I have three directories, each containing sub-directories for the projects that use a single GUI toolkit. These are the major environments.

Within each major environment, I have sub-directories, each containing a project/utility. Each of these projects needs its own libraries as well as a GUI toolkit. The GUI toolkit is the same for each, but the other libraries are all over the map.

What I know:
I know how to set things up so that when I cd into each parent directory, pyenv automatically activates a different version of Python along with the proper GUI toolkit library for the project.

My question:
Is it possible to use the same version of Python for all environments, but still have only one GUI Toolkit available within each environment?

Another question:
And once that's set up, is it possible for each project within a given environment to make use of a sub-environment with its own set of additional libraries (numpy, moviepy, etc.) available alongside the GUI toolkit used for all projects within that environment?

In other words, I'm looking to set up a three-tier complex of environments:
Level 1: a single parent environment with a single version of Python,
Level 2: Three sub-environments, each with a different GUI Toolkit, and
Level 3: A third level of sub-sub-environments with additional libraries needed on a project-by-project basis.

If any of this needs clarification, please let me know.

[native-api]

Virtualenv doesn't seem to be able to inherit packages from other virtualenvs.
Neither does Venv.

• You can install some packages into the base installation then create virtualenvs with --system-site-packages (pyenv virtualenv can pass arbitrary switches to the underlying command) thus create a 2-level hierarchy
• People have reported success using .pth files to use packages installed into another environment
• Finally, you can use a project-centric approach using Pipenv. It can create environments on demand from spec files. There'll be duplication -- but this mechanic eliminates human error as long as the spec files are correct.

[rontarrant]

Excellent. Thanks for all the info!

[rontarrant]

With some help from ChatGPT (and a lot of editing and experimentation afterward) I got it working. I have four umbrella environments, one each for Kivy, Toga, PySide6, and Tkinter. Each project under those umbrella environments has its own sub-environment. Attached is a quick doc explaining how it was done.
Virtual Environments with pyenv - ChatGPT.odt