1Show HN: Keep large tool output out of LLM context: 3x accuracy 95% fewer tokens (opens in new tab)(github.com)10loumaciel2mo ago1