what is everyone using for voice observability and evals atm? there doesn't seem to be a lot of options out there