fedrq has the following which uses the original version of this that @j-mracek added as a "raw prototype" after discussion in #281.

def get_releasever() -> str:
    """
    Return the system releasever
    """
    # libdnf5 >= 5.0.10
    # https://github.com/rpm-software-management/dnf5/pull/448
    base = libdnf5.base.Base()
    return libdnf5.conf.Vars.detect_release(base.get_weak_ptr(), "/").get()

Returning a proper string instead of a wrapper type is definitely nicer but does represent a breaking change.

Although, with python3-libdnf5-5.1.17-2.fc40.x86_64, it seems that is now broken and returns an empty string when it used to return the proper releasever.

Returning a proper string instead of a wrapper type is definitely nicer

Returning a proper string is nicer.
And, I tested the wrong version. It really returns a wrapped std::unique_ptr<std::string>. And get() can be used to get the string. The problem is that it doesn't distinguish between None and an empty string, which was probably the original intent of the unpleasant std::unique_ptr<string> construct.

but does represent a breaking change.

But as you write. It's breaking change.
I'm considering whether to close this PR... Anyway, it complicates my further work.