I'm working on a RAG app using fastapi and llama-index and I'd like to know if any of you knows a way to handle to concurrent requests given that the chat engine is a shared global variable. Right now I'm using python Lock class but it seem to slow down the process.
