Classify: customer message sentiment

The job

Jev judge eval (1 choice) over a fixed state : runnable via experimental_evaluate.

The prompt

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

// Classify: customer message sentiment — Jev judge eval (typesafe-ai/jev).
// A terse reply after a refund lands; the judge labels its sentiment.
// 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: "Customer message: Fine. Got the refund. Still took two weeks and three chats, but whatever — at least it is over.",
  questions: {
    "sentiment": {
      "type": "choice",
      "instructions": "Label the overall sentiment of the customer message.",
      "criteria": {
        "positive": "Praise, thanks, or clear satisfaction with no residual complaint.",
        "neutral": "Purely informational, no discernible emotion either way.",
        "negative": "Frustration, complaint, or dissatisfaction, even if partly resolved."
      }
    }
  },
});

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