User repository on GitHub
Verbatim Jev-scored context reduction for omp, over TypeSafe or OpenRouter
- Forks
- 2
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
OMP Jev Compaction
An omp plugin that does verbatim Jev-scored context reduction: stale tool calls and results are truncated to a short head plus a recoverable note, while sticky rewrites keep the provider's prompt cache intact.
- Stars
- 12
- Language
- TypeScript
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- jerryfane
- Last verified
- Sep 2026
About this project
Verbatim context reduction for omp, scored by TypeSafe's Jev decision model, over either the TypeSafe API or OpenRouter. Nothing is rewritten or summarized. Every tool call and tool result is scored; the ones Jev says are no longer needed are truncated to a short head plus a recoverable note, and everything else, including all user and assistant text, is passed through exactly as omp built it.
Quoted from the project's README. Read the full README.
Works forJev for Tool calling
Used inJev for Software & DevTools
Highlights
- Nothing is rewritten or summarized: stale tool calls are truncated to a short head plus a recoverable note.
- A 16,198-message replay produced 3 rewrites and 37 reuses over 40 requests, with 0 prefix breaks across 39 checks.
- A cache guard skips sessions at 80% or more cache reads; on 14 live sessions it protected $1.81 per request round while capturing $2.61.
- One client covers the TypeSafe API and OpenRouter; OpenRouter measured about 430 ms and $0.0000198 per decision request.
- Dropped payloads are written to ~/.omp/jev-spill/<hash>.txt and recovered with one read instead of re-running the tool.
Quickstart
bash
omp plugin install jerryfane/omp-jev-compaction
ompWatch out
MIT-licensed, as is the vendored fast-jev-compaction core; requires omp plus a TypeSafe or OpenRouter key, and every reduction sends conversation state, including tool names, inputs, and message text, to the decision endpoint.
Questions about this entry
Short answers drawn from the linked project or page.
Nothing is rewritten or summarized. Every tool call and tool result is scored, and the ones Jev says are no longer needed are truncated to a short head plus a recoverable note. Everything else, including all user and assistant text, is passed through exactly as omp built it.
Decisions are remembered rather than recomputed. A rewrite needs context growth past 40% and at least 15 requests since the last one, or it fires unconditionally after 40. Replaying a 16,198-message session produced 3 rewrites and 37 reuses with 0 prefix breaks across 39 checks. On the non-sticky path a cache guard skips any session already at 80% or more cache reads.
It does lose facts: full context answered 100% of planted questions in a recall script, reduced context answered 63-75%. With payload parking on, the misses become recoverable, with 0 permanent losses across both thresholds tested. Dropped payloads are written to ~/.omp/jev-spill/<hash>.txt and fetched back with one read instead of re-running the tool. Setting OMP_JEV_SPILL=0 makes the misses permanent.
Often not, and the README says so before you rely on it. omp prunes tool output before the session_before_compact hook runs, so in live sessions the region handed over was usually already reduced by omp's own shake to markers like [shaken ~516 tokens - recover: artifact://6], or held no tool calls at all. The hook then declines and omp compacts normally, so the continuous context path is the one that pays.
Install it as an omp plugin with omp plugin install jerryfane/omp-jev-compaction, then just run omp. Installing is the opt-in, so continuous reduction is on by default with threshold 0.2 and a 150,000-character floor; turn it off with OMP_JEV_CONTEXT=0 or omp plugin disable omp-jev-compaction.
A dropped payload is written to ~/.omp/jev-spill/<hash>.txt and the replacement text names that file, so the agent recovers it with one read instead of re-running the tool; identical payloads share one file. With OMP_JEV_SPILL=0 the misses become permanent.
Rewrites are rare and remembered: one happens only when context grows past +40% (rewriteGrowth) and at least 15 requests have passed, or unconditionally after 40. Replaying a real 16,198-message session, 40 requests produced 3 rewrites and 37 reuses with 0 prefix breaks across 39 checks.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: The compaction plugin, in Chinese
X: Decision outputs in Grep
Hacker News: jevmem keeps Claude Code project memory by asking Jev which turns are worth saving.
avinashjetwani61 pointsJevmem – automatic project memory for Claude Code, built on JevRead the thread on Hacker NewsX: A usage router for Grok Bot
More Jev resources like this
A Claude Code plugin and npm library that replaces transcript summarization with Jev keep-or-drop judgments, deleting stale tool calls and results while kept text stays verbatim.
Jev Tools & CLIs#community#typescript#claude-code
A Jev-guided context compaction library for coding agents: instead of summarizing old context, Jev decides which tool calls and results still matter while user and assistant text is kept verbatim.
Jev Tools & CLIs#community#typescript#compaction
A Pi extension that keeps conversation text verbatim while pruning stale tool history with Jev keep-or-drop decisions, falling back to Pi's built-in summary compaction when pruning cannot free enough context.
Jev Tools & CLIs#community#typescript#compaction
A Codex plugin and npm library that wraps native session compaction: Jev scores every tool call and result, and the verbatim history it keeps is re-injected after Codex summarizes. The README does not recommend using it.
Jev Tools & CLIs#community#typescript#codex
An MCP server and agent skill that verifies coding-agent work by driving the real running app and returning pass, fail, or couldn't-tell verdicts with the file:line to fix; its listing frames the approach as Jev-style machine-native runtime perception.
Jev Tools & CLIs#community#typescript#mcp
Community
reticlehq/reticleTypeScript
reticlehq/reticleTypeScriptA router that turns models, subagents, skills, MCP tools, CLIs, and plugins into one candidate set Jev answers with a single typed Choice, while the router code owns availability, permissions, and confirmation.
Jev Tools & CLIs#community#typescript#routing
Community
BillionsBobby/JevRouterTypeScript
BillionsBobby/JevRouterTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.