And as I know how to run local LLMs on the same Mac where DT runs, there is no fear to get into privacy issues. Here it is OpenAI gpt-oss 120B a reasonig model about as performat as OpenAI GPT o4-mini, which was SOTA a few month ago. Pretty strong!
Local RAG is not likely to be very performant. Elephas is a good example.
Cloud-based RAG is not consistent with DT’s principles of data security.