Super interesting - and it looks pretty clean too. I was thinking about building something like this just for fun and out of curiosity, to see how they would perform. (I also wanted to have say 5 instances of each model to compare their decisions - See if, say Gemini would end up going more or less for the same things or not across all the 5 instances, etc) I'll keep checking it daily!