Hi,
without a full specification it's hard to implement it. One could try to reverse-engineer the protocol by checking the code (it's public afaik), but of course it might not be as simple as it sounds.
Of course the performance of the protocol should be evaluated in the target scenario, i.e., with the intended MAC/Phy protocols, or you might have completely different results.
If the protocol (like it seems to be) is based on LoRA, this means to use the LoRA model from the AppStore (caveat: I have no idea of its maintenance status).
As for the protocol implementation (see above, either you find the exact specification, or it's reverse-engineering time) you could use the same approach of 6LoWPAN, which defines a shim layer. In your case the layer can do the ruting.