You can use models to collect details of your app that you'd like to evaluate, like the prompts or the LLM settings that you're using.