Damus
franzap · 5w
gpt-oss:20b if you can run it, it's pretty good You would use a RAG for augmenting the local model, I'll experiment with that soon