RAG Assessment

I’m developing a RAG system for schools trying to use only open source elements (llms, frameworks etc.) in Python. By processing documents I can obtain a dataset with the following features: question, answer, context(s), ground truth. I tried to feed Ragas (sticking to their tutorials) with this dataset but an Openai key seems necessary. Is it true? If yes, can you suggest me alternatives to Ragas? Thank you.