ToolCallJudge

ToolCallJudge(config?): Judge<ToolCallJudgeOptions>

Creates a deterministic judge that checks expected tool calls.

Parameters

config?

ToolCallJudgeConfig = {}

Matching behavior shared by every assessment from this judge.

Returns

Judge<ToolCallJudgeOptions>

Example

describeEval("refund agent", {
  harness: refundHarness,
  judges: [ToolCallJudge({ ordered: true })],
}, (it) => {
  it("creates a refund after lookup", async ({ run }) => {
    await run("Refund invoice inv_123", {
      metadata: {
        expectedTools: ["lookupInvoice", "createRefund"],
      },
    });
  });
});