Thanks for this.

Built this myself using Gemma4, but the branching is neat. Will have to try it.

If you're packing the questions into the branches, is there a chance that the ordering of the questions can change the probability of each output?

I see you have packs, but as close to deterministic as possible is what I'm after.