2026-11-09 · America/Los_Angeles · 社区动态 · #8

I built Prefixwise (MIT): tells you why your Claude Code, Codex, or agent calls miss the prompt cache, fixes common misses in flight, and blocks cache regressions in CI

Prefixwise tells you why your LLM calls miss the prompt cache, and what each miss costs you. Put it between your agent and the model, or inside vLLM itself, and it identifies the cause of every miss in plain English, fixes the common ones for you, and prevents new ones from reaching production. Your prompts never leave your machine: it stores hashes and counts, nothing else. Repo: https://github.com/tejachavali/prefixwise Install: pip install prefixwise (0.10.0 is out today) Identify: 19 causes, each with its evidence and fix. Tools sent in a changed order, a timestamp in the system prompt,…

热度 40.5 / 100；排名与评分保留该期记录。