Triage: crash report routing + reproducibility

The job

Jev judge eval (1 boolean + 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.

// Triage: crash report routing + reproducibility — Jev judge eval (typesafe-ai/jev).
// A crash report with steps and logs; the judge routes it and checks reproducibility.
// 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: "Report: app crashes on export when the board exceeds 500 cards. Steps: 1) open Big Board (612 cards), 2) Export → CSV, 3) crash at ~80%. Logs attached, crash ID 7f2a. Repro: 3/3 attempts on iOS 17.2 and 17.3.",
  questions: {
    "owner": {
      "type": "choice",
      "instructions": "Which team should own this crash report?",
      "criteria": {
        "mobile": "Client-side crash in the app itself.",
        "backend": "Server or API failure behind the symptom.",
        "data": "Data corruption or migration issue.",
        "docs": "Not a bug — documentation or usage question."
      }
    },
    "reproducible": {
      "type": "boolean",
      "instructions": "True only if the report includes concrete steps plus evidence of repeated reproduction (counts, versions, or logs)."
    }
  },
});

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