Audit: weekend itinerary feasibility + quality

The job

Jev judge eval (1 boolean + 1 score) over a fixed state : runnable via experimental_evaluate.

The prompt

Copy this into Muse and replace anything in [square brackets] with your own details.

// Audit: weekend itinerary feasibility + quality — Jev judge eval (typesafe-ai/jev).
// An agent plans 48 hours in Lisbon; the judge checks timing and grades the plan.
// Model: typesafe-ai/jev via the Vercel AI Gateway ($0.042/1M input tokens).
// Auth: set AI_GATEWAY_API_KEY (vck_…) in the environment — never paste the key here.
// For zero-data-retention / no-training workloads, add
// providerOptions: { gateway: { … } } to the evaluate call below.

import { experimental_evaluate } from 'ai';

const result = await experimental_evaluate({
  model: 'typesafe-ai/jev',
  state: "Itinerary: Fri 18:00 land LIS, hotel Alfama. Sat: Belém (09:00), lunch Time Out Market, Alfama + castle (15:00), fado show 21:00 (booked). Sun: day trip Sintra 08:30–14:00, fly 19:20. All legs under 40 min by metro/taxi; buffers included.",
  questions: {
    "timing_feasible": {
      "type": "boolean",
      "instructions": "True only if every leg leaves realistic travel time and nothing overlaps the 19:20 Sunday flight."
    },
    "itinerary_quality": {
      "type": "score",
      "instructions": "Score this itinerary as a weekend plan. Higher is better.",
      "criteria": [
        "Infeasible or missing logistics entirely.",
        "Feasible but a rushed death-march with no buffers.",
        "Feasible and paced, but generic or missing bookings.",
        "Feasible, paced, booked where it matters, with buffers."
      ]
    }
  },
});

console.log(JSON.stringify(result, null, 2));