I was playing around with OpenAI's API, and found I could get the same output from the same input if I turned the temperature down to 0. Does anyone know if this is a feasible way of getting deterministic output from an LLM?