Show HN: KV‑psi Uses Linux PSI to Trim Large Language Model KV Cache

KV‑psi is an open‑source project announced on Hacker News. It applies Linux Pressure Stall Information (PSI) to monitor memory pressure. The tool dynamically trims the key‑value cache of large language models. By reducing cache size, it aims to improve overall memory efficiency. The codebase is hosted on GitHub for community access. Users can integrate KV‑psi with existing LLM inference pipelines. The project includes documentation on configuration and usage. Early results suggest lower memory consumption without major performance loss.