test suites are more like simulations, we will have one vapi agent call another and actually do the call, evals are text based, so we're only testing the llm prompt behavior