thanks for checking it out. i imagine 2 things, 1 - any edit actions require a "human in the loop" approval. 2 - for sensitive info, internal instances of llms or only local models accessing info