Show HN: Reduce ChatGPT costs 10x with distributed cache for LLMs