i think the question is really: can I turn my search problem into a in-process vector search problem where I can scale with the number of processes.