Skip to content

EvalRunOptions

Per-run metadata forwarded to the harness alongside the test input.

await run("Refund invoice inv_123", {
metadata: {
expected: { status: "approved" },
expectedTools: ["lookupInvoice", "createRefund"],
},
});

TMetadata extends HarnessMetadata = HarnessMetadata

optional metadata?: TMetadata

Per-run expectations or configuration forwarded to harnesses and judges.