Screen: product review authenticity + helpfulness

The job

Jev judge eval (1 choice + 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.

// Screen: product review authenticity + helpfulness — Jev judge eval (typesafe-ai/jev).
// A five-star review with oddly generic praise; the judge screens it.
// 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: "Review: ★★★★★ Best product ever!!! I love it so much, everyone should buy this amazing product. Fast shipping too. — Verified purchase, account created yesterday, 14 identical reviews posted today.",
  questions: {
    "authenticity": {
      "type": "choice",
      "instructions": "Judge the authenticity of this review.",
      "criteria": {
        "genuine": "Reads like a real customer with specific experience.",
        "suspicious": "Generic or incentivized-looking, but not clearly fake.",
        "fake": "Clear abuse signals (bulk posting, new account, template text)."
      }
    },
    "helpfulness": {
      "type": "score",
      "instructions": "Score how helpful this review is to a shopper. Higher is better.",
      "criteria": [
        "No usable information at all.",
        "Vague sentiment with one generic detail.",
        "Some specifics but missing key purchase factors.",
        "Specific, balanced, and decision-useful."
      ]
    }
  },
});

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