Also check out DeepEval... our team has been using it for a while, and it's been working well for us because we can evaluate any LLMs, something this library doesn't seem to support (https://github.com/confident-ai/deepeval).