Ask HN: Open-Source AI chat/Q&A over your own dataset?

I'm seeing a lot of open-source projects for training an LLM on your own dataset/documents, but they always seem to rely on an OpenAI key.

Apologies if this is a question with an obvious answer, but are there any good open-source projects for doing this without leveraging a 3rd party service? Something that would be 100% self-hosted?