What it is
A pi coding agent extension that shortens long tool output before it enters the context, so cached prompt prefixes are never invalidated. For pi users who want smaller contexts without losing information.
How it uses Jev
The output is cut into blocks of up to 1,500 characters. One Jev call gives each block a probability that the current request needs it, and says whether every line is needed. Only confidently useless blocks are hidden; code guarantees failure lines, request terms and top-ranked blocks survive.
Primitives:score
Technique worth stealing
Trim at write time, before caching, so the cached prompt prefix is never invalidated.
Try it
npm install && npm test; node bench/run.ts --dry-run
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.91
- Shows a System One patternyes0.75
- Handles uncertaintyunclear0.51
- Measuredyes0.76
- Runnableyes0.93
- Worth recommendingunclear0.56
- Model replicano0.06
- Problem scopescore on a 0–2 scale1.06
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.