Interesting how long it takes to get this “assistant” functionality on a local running models with llama for e.g.? It’s seems like many use cases require access to user data.
But building a simple RAG/Vector DB on top of llama3 is not very complicated.