jevbooks

← All projects

jev-demo

SrPio/jev-demo

Jev picks a Claude model by task complexity.

Classification & Routing100%Runner-up: Verification & GuardrailsQuestionnaire

What it is

A small TypeScript web app where you type a task or requirement and Jev, TypeSafe's "System One" model, decides which Claude model fits best based on complexity. It only queries Jev through Vercel AI Gateway; it never calls Claude or Anthropic.

How it uses Jev

The API route calls experimental_evaluate from the ai package with the typesafe-ai/jev model, passing the user's text as state and three typed questions: a choice for the recommended Claude model, a score for complexity, and a boolean for whether deep reasoning is needed. The result is shown in a result card.

Primitives:choicescorenoul

Technique worth stealing

Use Jev as a router: pass a catalog of models as choice criteria and let typed questions pick one.

Try it

Copy .env.example to .env.local with AI_GATEWAY_API_KEY, then pnpm install and pnpm dev; open http://localhost:3000.

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricyes0.86
  • Shows a System One patternno0.10
  • Handles uncertaintyno0.04
  • Measuredno0.04
  • Runnableyes0.97
  • Worth recommendingno0.32
  • Model replicano0.02
  • Problem scopescore on a 0–2 scale0.61
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 23 Sept 2026.