Hacker News new | ask | show | jobs
by mattfrommars 304 days ago
Hi,yes. It makes sense now. To cache the steps and reply. Very efficient strategy than to run the step each time using LLM