Skip to main content

support-ticket

id: support-ticket sidebar_position: 1 title: Support ticket triage​

This example mirrors examples/support-ticket.ts. It classifies inbound emails and assigns a priority.

examples/support-ticket.ts
import {definePrompt} from 'prompt-fn';
import {z} from 'zod';
import {google} from '@ai-sdk/google';

const classifyTicket = definePrompt({
  name: 'Support Ticket',
  description: 'Classify support tickets by intent and urgency',
  inputSchema: z.object({
    subject: z.string(),
    body: z.string(),
  }),
  outputSchema: z.object({
    intent: z.enum(['bug', 'billing', 'question']),
    priority: z.enum(['low', 'normal', 'urgent']),
    summary: z.string(),
  }),
  template: ({subject, body}) => `Summarize and classify\nSubject: ${subject}\nBody: ${body}`,
  model: google('models/gemini-1.5-flash'),
});

const response = await classifyTicket({
  subject: 'Refund not received',
  body: 'I cancelled last week and still have not seen the refund.',
});

Why it is interesting​

  • Demonstrates string templates with interpolated variables.
  • Shows how logger + recovery keep the response usable even when Gemini or Ollama return quoted JSON.
  • Input/output types are inferred, so response.priority narrows to 'low' | 'normal' | 'urgent'.

Try it locally​

bun test test/integration/integration.test.ts --filter "Using Google Gemini"

Set GOOGLE_GENERATIVE_AI_API_KEY before running. See test/integration/integration.test.ts for the env variable guards.