That's really cool you guys made your repo public! The personality dials are pretty neat - it looks like you're operating off percentages which should make the AI a bit less predictable. I think the really strong part is in your PokerManager too, which seems to allow decisions to be made of percentage? I'm impressed by that logic - the one in my game was much simpler since the hand sizes were fixed at 5 dice total with the 3 common ones being static and the 2 private ones having to reroll together. I precomputed the possible hands given the three common dice, but fought with not having a central way to evaluate hand rank (oof). The possible hands were put into probability buckets and tied to the amount of rerolls that the player and the AI had done to measure "believability" of a given bluff.
So there are a bunch of things in your decision functions and poker functions that I will try to learn from. Thanks for sharing!