Jev Support Triage

Why Jev

Jev returns typed choices and scores your code can branch on — route to billing, set urgency, and escalate — in one low-latency call without parsing prose. Example input: “I’ve been charged twice for my order.” → category, urgency, escalate recommendation.

How to define the decision

Pass the raw ticket subject + body as state. Define a Choice for team/topic (billing, bug, account…), a Score for urgency or severity, and a Noul for escalate_now. Keep criteria labels stable so your router code stays schema-safe across releases.

Sample input and output

From the live tool template. Run it in the playground for a real API response — sample labels below are illustrative until you click Run.

subject: App is down

body: Production dashboard returns 500 for all users since 10am.

Treat high-confidence Choice labels as safe auto-routes. If urgency Score or escalate Noul sits near your threshold, queue a human instead of auto-refunding or auto-closing. Log ticket id + decisions for audit — never invent a label when the API errors.

Failure modes and human fallback

• Upstream or network errors: retry with backoff; do not invent a decision.
• Low confidence: route to a human queue or safer default — never auto-act on money/destructive tools.
• Schema drift: keep a golden fixture set in CI so label changes fail the build before production.
• Do not auto-issue refunds or account deletes from Jev alone.
• Fail closed to a human queue on low confidence or upstream errors.
• Keep reply drafting in templates or a generative LLM — Jev owns the decision layer only.

API example

Same prepaid jv_live_ balance as tools. Output tokens are free on this host's metering model.

curl https://jevtypesafe.org/api/v1/decide \
  -H "Authorization: Bearer jv_live_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -d '{"state":"Support ticket\nSubject: App is down\nBody: Production dashboard returns 500 for all users since 10am.","questions":{"team":{"type":"choice","instructions":"Which team should own this ticket?","criteria":{"technical":"bugs, outages, engineering","billing":"payments, invoices, refunds","account":"login, access, permissions","general":"questions that do not fit elsewhere"}},"issue_type":{"type":"choice","instructions":"What kind of issue is this?","criteria":{"outage":"service is down or broken for many users","bug":"product defect for one or few users","question":"how-to or information request","billing":"money, charges, subscriptions","abuse":"security, spam, or policy violation"}},"severity":{"type":"score","instructions":"How severe is the impact?","criteria":["cosmetic or minor","limited impact","major impact","critical / production down"]},"urgency":{"type":"choice","instructions":"How urgently should this be handled?","criteria":{"routine":"can wait days","today":"should be handled today","urgent":"needs attention within hours","now":"immediate response required"}},"escalate":{"type":"noul","instructions":"Should this be escalated to a human manager or on-call immediately?"}}}'

FAQ

Can Jev replace my whole support bot?

No. Use Jev for the decision layer (route, urgency, escalate). Keep an LLM or templates for drafting replies.

How do I try support triage here?

Open the Support Triage tool, run the sample ticket free, then copy the same questions into /api/v1/decide with a jv_live_ key.

What if confidence is low?

Fail closed to a human queue when confidence is below your threshold. Log the ticket id and decision — do not auto-refund or auto-close on low-confidence paths.

Related use cases

• Jev Email Triage
• Jev Agent Safety
• Jev Model Routing