GitHub - chopratejas/headroom: The Context Optimization Layer for LLM Applications
kept by eddie
Headroom compresses tool outputs, logs, and RAG chunks before they reach an LLM, reducing token count by 60–95% without changing answers.