shipwithjev

Catalog / Tools & apps

0092GitHub

jev-skill-router

Skill routing for an agent, decided by Jev.

Jiaqi Gu’s router picks which skill an agent loads, using Jev for the decision instead of an LLM turn.

droid-Q/jev-skill-routerREADME ↗
# Jev Skill Router for Codex

**English** | [简体中文](README-cn.md)

## Features

- **Jev skill routing** — evaluate eligible skills and pass the selected names, paths, and relevance probabilities to Codex.
- **Local selection history** — record selections, no matches, empty catalogs, and routing failures, together with project, session, model, threshold, and latency.
- **Web analytics** — inspect selection rates, daily trends, skill rankings, average relevance, and expandable records. Filter by time, project, skill, or outcome.
- **English / Chinese UI** — switch language and light/dark appearance; the dashboard also works on mobile.
- **Dashboard on app launch** — on macOS, register the included startup helper once. Opening Codex starts the dashboard without opening a task or sending a message; its port comes from the Jev configuration file.

### Dashboard preview

The screenshots below use **synthetic demo data**, not measured Jev accuracy or real user history. The visual style takes inspiration from [TypeSafe AI](https://typesafe.ai/).



<details>
<summary>Inspect an individual routing decision</summary>



</details>

## Requirements

- Codex CLI with plugin hooks, `SessionStart`, `UserPromptSubmit`, and app-server `skills/list`. Developed against **0.154.0**.
- **Node.js 22+** available to the process running Codex.
- A TypeSafe API key with access to Jev.

The shipped hook is bundled. Plugin users do **not** need `npm install`, Python, or an MCP server. App-launch startup uses the included macOS user LaunchAgent; session/prompt startup remains available on other platforms.

## Install

```sh
codex plugin marketplace add https://github.com/droid-Q/jev-skill-router.git
codex plugin add jev-skill-router@jev-skill-router
```

Configure the key as described below. Open 

Also filed under Tools & apps

  1. 0573

    Support chat: agent offered a human escalation path

    A frustrated customer hits a bot loop twice; the judge checks whether a human handoff was offered.

    everyai-com · Tools & apps

  2. 0572

    Subscription cancel: effective date is explicit

    An agent cancels a subscription; the judge checks the exact date the cancellation takes effect.

    everyai-com · Tools & apps

  3. 0571

    Sales quote: numbers match the official price list

    An agent quotes $29/seat against a price list; the judge checks the quote matches exactly.

    everyai-com · Tools & apps

  4. 0570

    Refund chat: agent stated the amount before issuing it

    A support agent issues a $42.50 refund; the judge checks the exact amount was stated before the refund was confirmed.

    everyai-com · Tools & apps