Does every clearing decision cost a model call?

Superseded reads, repeated test runs, repeated skill reads, and similar restatements clear with no network call. Only remaining large outputs go to Jev, and only when their still-needed probability falls below threshold.

Can I get the original output back without rerunning the command?

Yes. The original body is replaced by a stable marker while call, order, and metadata stay intact. A retrieval tool returns the stored output in pages without rerunning commands or rereading changed files.

Why does it sometimes keep an old output instead of clearing it?

Clearing rewrites cached prefixes, so a batch commits only when payback lands within the configured request window. Overlap with provider-native Codex compaction pauses this extension on that branch.

Will this work in plain Claude Code or Codex CLI sessions?

No. It needs a Pi session, a recent Node runtime, and a TypeSafe key where Pi launches. Without the key it stays dormant, and without the retrieval tool enabled it pauses clearing.

What data does pi-jev-compaction send to TypeSafe?

Up to eight recent user/assistant excerpts of about 600 characters, candidate tool names and truncated arguments of about 400 characters, and result head/tail excerpts of about 800 characters plus output length, all capped at 24,000 UTF-8 bytes per request. Thinking, signatures, images and system instructions are excluded.

When does pi-jev-compaction start clearing context?

Once the next request reaches 120,000 tokens or 45% of the model's context window, whichever is smaller, while protecting the latest 12,000 estimated tokens. On a 1M-token window it starts at 120k rather than 450k.

Which environment variables tune pi-jev-compaction?

PI_JEV_MODEL (default jev-1.13.0), PI_JEV_THRESHOLD (0.45), PI_JEV_TRIGGER_TOKENS (120000), PI_JEV_MAX_PAYBACK_TURNS (20), PI_JEV_KEEP_THRESHOLD (0.25), PI_JEV_KEEP_RECENT_TOKENS (12000) and PI_JEV_TIMEOUT_MS (5000). Invalid numeric settings fall back to defaults.

CommunityJev Agents & MCP

Pi Jev Compaction

Automatic Pi extension that clears safely superseded reads and test runs outright and asks Jev about other large older tool outputs, replacing them with retrievable markers.

See highlights

Open github.com

Stars: 4
Language: TypeScript
License: MIT
Published by: Community
Author: nourhelmi
Last verified: Oct 2026

nourhelmi/pi-jev-compaction

User repository on GitHub

View on GitHub

Automatic Jev context clearing for Pi. Keep the conversation, prune stale tool output, retrieve originals without rerunning commands.

Forks: 2
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Keep the conversation. Clear stale tool output. Get the original back without running the command again. An automatic Pi extension powered by TypeSafe Jev. It clears safely superseded reads and test runs directly; Jev scores other older tool outputs. Original session messages stay intact. No extra agent. No summary model. No runtime dependencies beyond Pi.

Quoted from the project's README. Read the full README.

Tagscommunity awesomejev catalog-import

Works forJev for Classification Jev for Tool calling

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Clears superseded reads and identical test runs locally with no network call.
Scores up to sixteen large outputs per evaluation through one Jev request.
Markers preserve order and metadata while originals stay retrievable in-session.
Payback gate avoids cache rewrites that would cost more than they save.

Watch out

Evaluations use bounded excerpts and can misjudge relevance; retrieval exists for that reason and secret handling is heuristic. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

JevLoop

Runnable agent harness that moves pick, score and yes-no loop choices to Jev or Laya in milliseconds while reserving the language model for writing, compiling DECISION.md into executable questions.

Jev Agents & MCP#community#jevbest#catalog-import

Community

zjunlp/JevLoopTypeScript

patdown

Node CLI that lints a file tree against fuzzy natural-language rules in one Markdown file, judging each matched file with a swappable TypeSafe-backed judge.

Jev Agents & MCP#community#jevbest#catalog-import

Community

tyler-dot-earth/patdownTypeScript

jev-cli

Command-line toolkit that wraps Jev judgments into scriptable verify, screen, classify, extract, find, rerank, match, route, ask, compact, and batch commands.

Jev Agents & MCP#community#jevbest#catalog-import

Community

Nasrallah-AL/jev-cliTypeScript

jevwire

Wires Jev into agents as seven MCP tools, an embeddable JevDecisionModel library with run functions, and Claude Code hooks at tool and turn boundaries.

Jev Agents & MCP#community#jevbest#catalog-import

Community

Brainwires/jevwireTypeScript

jev-recipes

TypeScript recipe library with 248 focused functions that call Jev for routing, reranking, gating and labelling, returning structured ready-or-review decisions for Node backends.

Jev Agents & MCP#community#awesomejev#catalog-import

Community