A trial-and-error approach to temperature opimization for LLMs. Runs the same prompt at many temperatures and selects the best output automatically. - View it on GitHub