Normally, you have to create a separate env for each project to avoid issues with incompatible versions of packages/libraries/etc. Sometimes, it's even impossible to install desired version of Python itself, thus one of the best tool here - pyenv (manages different python versions). Next step - create virtualenv (venv, conda, poetry - doesn't matter too much at the start, but poetry is quite popular on real projects) for any project where you would install dependencies. There is no sense to use virtual environment for a single script with builtins.