Show HN: Memory for LLM apps that cuts input tokens up to 80% (avg 68%)github.com/Tem-Degu3 pointsdegutemesgena month ago