Skip to content

pi-subagent-jev

Pi package that gates pi-subagents dispatches with one Jev call against configurable rule sets and adds a jev_ask tool.

Pluginbeta

Overview

pi-subagent-jev gates subagent dispatches in Pi (with the pi-subagents package) through a JEV System One endpoint. When the main agent dispatches a subagent, the task text is evaluated in one typed call against configured rule sets, each rule a probability question with a below or above threshold and a message. If any rule fires, the dispatch is blocked and every violation is returned to the main agent; per the list summary, it fails open to allow on errors.

It also provides a jev_ask tool for typed noul, choice and score questions, a bundled skill that guides configuring endpoints and per-subagent compliance rules (including in Chinese), and opt-in runtime monitoring of running subagents for bash stalls, unproductive tool loops and repeated failures with advisory alerts. Since v0.7.0 there are no built-in rules, so an install without a rules file performs no checks; v0.8.0 changed the rules schema. The default model is oc/jev-1.13-free at a configurable endpoint.

What the decision model decides

TypeQuestionAnswers
Yes / noDoes this subagent task violate a configured compliance rule?yes · no

How it works

  1. Main agent dispatches a subagent.
  2. Send the task text and rule questions to Jev in one call.
  3. Block with violation messages if any rule fires.
  4. Optionally monitor running subagents and alert.

Strengths

  • Configurable per-subagent compliance rules.
  • One call per dispatch.
  • Optional monitoring for stalls and loops.

Caveats

  • No checks without a rules file.
  • Only works with pi-subagents; fails open on errors.

Getting started

pi install git:github.com/G0-0000/pi-subagent-jev, create ~/.config/jev-comp/env with JEV_AI_API_KEY and JEV_AI_BASE_URL, then /reload.

At a glance

Type
Plugin
Status
beta
Language
TypeScript
Runtime
Pi package
Has code
yes

Use cases and integrations

How curated lists describe it

  • Agent governance: when the Pi main agent dispatches a subagent, evaluates the task text against configurable rule sets in one typed Jev call (per-rule probability questions with below/above thresholds), blocks the dispatch with per-rule reasons on any hit, and fails open to allow on errors.

    yibie/awesome-jev
  • A pi package that gates subagent dispatches through a JEV System One decision model, plus general-purpose tools to query that model.

    hellogumbo/awesome-jev

Frequently asked questions

Are there default rules?

No, since v0.7.0 you must provide a rules file.

What happens on a violation?

The dispatch is blocked and violation messages go back to the main agent.

Which agents are supported?

Only Pi with the pi-subagents package.

Related projects

  • Asynchronous System-1 supervisory plugin for Hermes Agent with typed Jev decisions, ranking, verification and an opt-in tool gate.

    ProjectbetaPython
  • Kernel coordinating several Claude Code agents on one repo: Jev rates collision risk before each write to pass, wait or escalate.

    ProjectresearchPython
  • Governance proxy between Jev or local Laya and agents that turns probabilities into allow, ask, deny, keep, truncate or drop actions.

    MCP serverbeta
  • Decision governor for Pi routing plan strategy, model tier, tool permissions and outcome (finish, retry, replan) through Jev.

    ProjectbetaTypeScript
  • Agent harness that compiles an environment's finite actions into typed questions for Jev and gates each step by confidence.

    CLIbetaPython
  • Software-factory supervisor where Jev independently judges coding agents' completion, test sufficiency, progress and need for a human.

    ProjectbetaPython