vibrantlabsai/ragas ? reverse-engineered prompt

Build me a Python library for evaluating LLM apps, especially RAG systems.

I want it to help me score outputs with objective metrics, generate test data when I do not already have a good dataset, and give me clear feedback I can use to improve prompts and retrieval quality. It should be easy to install and use from code, with a simple quickstart flow that can create a ready to run evaluation project for me. Support common LLM workflows, and make it work smoothly with popular tools like LangChain and other observability setups if possible.

I also want a way to define custom checks for things like summary quality or answer correctness, using an LLM to judge the result when needed. If you need to, look up the current docs online and make sure the examples and defaults feel modern and practical. Keep the project polished, documented, and ready for people to use right away.