Jev Cookbook
Fifteen practical recipes show how typed Jev answers drive routing, classification and evaluation.
Added to Jevfast
How it uses Jev
Recipe 01 sends each ticket message as state and asks for its department, urgency, frustration, and refund request in one call. The department is a Choice over five named teams. The script then uses department confidence, urgency, and frustration to route the ticket and assign a priority; its example data and result labels are also used for recipe evaluation.
What Jev decides
Illustrative support-ticket department choiceExample answers · not a recorded Jev response · Source ↗
Question 1 · department
YOUR APP
INSTRUCTION
Choose which team should handle the ticket message.
STATE
For each ticket, state contains its message. The department options and descriptions are fixed in the recipe; message contents vary per ticket. This is an illustrative request shape, not an observed answer.
JEV · CHOICE
- billing
- technical
- account
- shipping
- sales
App workflow
- Load sample tickets
The script reads its ticket collection from the recipe's tickets.json file.
- Ask Jev for each ticket
It processes tickets with a concurrency limit of eight and asks the department Choice, urgency Noul, frustration Score, and refund Noul together.
- Route and set priority
Code routes using confidence thresholds of 0.8 and 0.5, then sets priority from urgency and frustration.
- Compare and save results
The script compares answers with the ticket labels, prints summary metrics, and saves the results for the recipe.
Why it is interesting
The recipe separates classification from action. Jev supplies typed answers, while ordinary code applies confidence bands, sets priority, checks the result against labels, and records metrics. This makes the routing policy visible and adjustable in code.