Support chat: agent offered a human escalation path

The job

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

The prompt

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

// Support chat: agent offered a human escalation path — Jev judge eval (typesafe-ai/jev).
// A frustrated customer hits a bot loop twice; the judge checks whether a human handoff was offered.
// 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: This is the third time I am explaining this. Your bot keeps closing my ticket.\nAgent: I understand this is frustrating. I have read the full history of ticket T-551. Here is what I will do: re-open the ticket, apply the missed credit, and stay on this chat until you confirm it looks right. If you would rather talk to a person, I can transfer you to a human specialist right now — just say the word.",
  questions: {
    "handoff_offered": {
      "type": "boolean",
      "instructions": "True only if the agent explicitly offered to transfer or escalate the customer to a human."
    }
  },
});

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