Evaluation patterns

import { Agent } from "@agentium/core";
import { custom, EvalSuite } from "@agentium/eval";

export async function evaluateRouting(agent: Agent) {
  const result = await new EvalSuite({
    name: "ticket-routing",
    agent,
    concurrency: 1,
    timeoutMs: 10_000,
    cases: [
      { name: "invoice", input: "Reply with only billing for an invoice question.", expected: "billing" },
      { name: "delivery", input: "Reply with only shipping for a delivery question.", expected: "shipping" },
    ],
    scorers: [custom("correct-route", (_input, output, expected) => {
      const actual = output.text.trim().toLowerCase();
      const pass = typeof expected === "string" && actual === expected;
      return { score: pass ? 1 : 0, pass, reason: `Expected ${expected}; received ${actual}` };
    })],
  }).run();
  return { passed: result.passed, failed: result.failed, cases: result.results };
}