Context compaction with decision models

Pi port of fast-jev-compaction using Pi's context hook, with a savings gate against low-value pruning and retained status notes.

Pi extension suite using Jev for selective verbatim compaction and per-turn routing between cheap and strong models.

Selective verbatim context compaction for Pi: Jev scores each tool call and result, with optional stricter pruning of assistant prose.

Automatic Pi context clearing: drops superseded tool outputs for free and asks Jev about large older ones, with originals retrievable.

Extractive compaction for Pi: Jev judges complete read-only tool pairs, code removes stale ones and keeps the rest as an extractive packet.

Opt-in, reversible Pi context filter where Jev scores older messages and low scorers are hidden from requests, not deleted.

Pi extension that judges each new message unit once with Jev, omitting unneeded units from context and logging removal reasons.

Pi extension that strips the inlined skill list and asks Jev which single skill, if any, to load per turn.

Pi observational memory where Jev answers keep/kind questions on history chunks, building a verbatim ledger and topic files.

DeepSeek Harness plugins that filter tool output, prune stale context and disclose skills progressively, with Jev or Laya judging what goes.

Windows Codex skill that safely installs and verifies save-token-jev hooks, where Jev picks which tool records to keep across compaction.

Pi coding agent extensions, including pi-jev-router, which asks Jev whether a prompt should continue, fork or start a new session.

Codex and Claude Code skills that narrow a repository locally and let Jev judge which candidate files the coding model needs.

Pairs Jev with the get-fable lifecycle harness to give coding agents fast tool and skill routing and verbatim-preserving compaction.

Codex port of fast-jev-compaction that re-injects Jev-selected verbatim history after compaction; the author advises against using it.

OpenCode V2 port of fast-jev-compaction that asks Jev or a local Jev-style model which old tool outputs still matter.

Gives an agent one tool that uses Jev to search a nested capability tree and return only the documented commands a task needs.

Coding agent where Jev handles context selection, tool routing, plan gating and verification, and the big model sees only key context.

Three Italian-language Claude Code prompts that build Jev-powered model routing, context compaction and code review into your project.

Claude Code compaction plugin where Jev decides per old tool result whether to keep it verbatim or move it to a drawer file.

Codex search plugin that runs ripgrep and keeps only passages Jev judges relevant to the question.

Enhancement pack for MiniMax Code CLI on local DGX Spark hardware, including a Jev-style local decision browser agent and context patches.

OMP extension keeping one endless chat as a summary tree, with five Jev judge gates for rules, prefetch, compression, tools and recall.

Local context-pruning proxy for Claude Code and Codex where Jev decides which history spans can be safely omitted.

Deterministic evidence selector for noisy Codex investigations; its earlier Jev subagent router is kept for the record, not recommended.

Paper code where Jev screens agent tool output for relevance and prompt injection before the agent reads it.

Archived Go harness finding Jev tool steering does not reliably cut agent tokens in Claude Code or Codex; context trimming helped.

Research code evaluating Jev as an independent per-tool relevance judge for building task-level tool spaces on TGB, BFCL and GTA.

Chinese translation of 'Jev Engineering for coding agents', design notes on a harness where Jev makes every per-turn decision.

Fork of fast-jev-compaction adding env-configurable endpoints, a CLI, an MCP server and a Devin post-compaction restore hook.