2 comments

[ 0.22 ms ] story [ 3.8 ms ] thread
Hey there, nice one. Care to share a bit what is under the hood? What are the challenges of local AI? It is RAG-ging? What small models do you use?
To avoid misquoting scripture it’s using a couple techniques including tool calls for grounding and a client-side parsing layer to find and replace citations.

It currently uses deepseek flash. We’ll evaluate local models in the future.