Most RAG (Retrieval-Augmented Generation) systems call out to hosted embedding APIs and vector databases at query time. This project flips that model: embeddings and search happen locally in the ...