Jev examples
Jev examples: decision patterns that work
The quality of a Jev decision depends mostly on how you write the question. These examples show patterns that hold up well in production.
Get Early Access to ClassifierHub: 2× credits in your first paid month.
Ticket routing (choice)
Describe each option with concrete vocabulary that appears in real tickets, and always include an explicit catch-all option so unusual inputs don't get forced into the wrong bucket.
"team": {
"type": "choice",
"instructions": "Which team should handle this ticket?",
"criteria": {
"billing": "Invoices, charges, refunds, plan changes",
"technical": "Errors, bugs, outages, API problems",
"account": "Login, password, permissions",
"other": "Anything not covered above"
}
}Refund requests (noul)
Describe both sides of a yes/no question. Ask about the fact, not the wording: "asks for money back" is more robust than "mentions refunds", which also matches "I don't want a refund".
Lead fit (score)
Describe every level as a concrete situation, from lowest to highest. Use the per-level probabilities to find borderline leads that deserve a human look.
"fit": {
"type": "score",
"instructions": "How well does this lead fit a B2B analytics product?",
"criteria": [
"Personal use or unrelated need",
"Small need, unclear budget",
"Real need, small team",
"Clear need, growing team",
"Urgent need, established company"
]
}Agent guardrails (choice + score)
Before an agent runs a tool call, ask whether the action is allowed for the task, and how risky it is. Auto-approve only high-confidence allows with low risk, block clear rejections, and send the rest to review.
Frequently asked questions
Related guides
Last updated 2026-09-25. ClassifierHub is an independent product built on top of the Jev decision model, accessed through OpenRouter. It is not affiliated with or endorsed by TypeSafe or OpenRouter.