Dear Ravithree,
Thank you for the question - it is something I planned to announce in the near future anyway. My answer would be - definitely use Libra. We have recently implemented an nbra workflow module in it (e.g. from libra_py.workflows.nbra import *), which essentially implements the Pyxaid-on-steroids scheme, if you will. Basically, it can do the same things Pyxaid can + bunch of other things, such as wavefunction phase consistency correction, state tracking, new schemes for decoherence and accepted hops. The nbra module apparently relies on all the features Libra provides and we consistently improve it. This is where we implement new things and aim to further develop. We don't do the developments of the original Pyxaid much. Having said this, Pyxaid is still a viable option - it does work, but some effect in it are not accounted for (e.g. phase corrections).
If you decide to go with Libra, please take a look at our collection of Jupyter tutorials which explain how to run the Pyxaid-like calculations with Libra, e.g.
The Libra itself should be relatively simple to install using Conda.
On the user interface and conventions side, some things are changed in comparison to those of the original Pyxaid. For instance, there are 4 steps now, not 3. Excitations are defined in a slightly different way, etc.
We are still working on manuscripts that would describe the new features available in Libra and I hope you'll be able to see them before too long.
Hope this helps to solve some of your questions,
Alexey