jevbooks

← All projects

pi-jev-context

kevinpita/pi-jev-context

Reversible context pruning for Pi, powered by TypeSafe Jev.

Scoring & Ranking89%Runner-up: Data Labeling & Curation

What it is

An opt-in Pi extension that filters older messages from model requests while preserving the original session history. It is for Pi users running long sessions who want to keep useful context without deleting history.

How it uses Jev

Jev judges which older eligible messages are still useful. Candidates include complete text-only tool call/result pairs, assistant text, readable thinking, and user shell history. A candidate stays when its keep probability is greater than the threshold (default 0.8). Low-scoring content is hidden from future model requests, not deleted.

Primitives:score

Technique worth stealing

Branch-aware caching of Jev relevance judgments so only new or changed candidates are rejudged.

Try it

pi install npm:pi-jev-context, restart Pi or run /reload, then /login typesafe and /jev on.

View on GitHub

judged by Jevjev-1.13.0

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-centricyes0.83
  • Shows a System One patternno0.22
  • Handles uncertaintyno0.06
  • Measuredno0.05
  • Runnableyes0.99
  • Worth recommendingno0.31
  • Model replicano0.03
  • Problem scopescore on a 0–2 scale1.05
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.