eval

type Expectation struct {
	// ExpectedTools lists tool names that must appear in order among executed tools.
	ExpectedTools []string
	// RequiredArgs maps tool name to required argument keys that must be non-empty.
	RequiredArgs map[string][]string
	// MaxSteps fails the scenario when the loop exceeds this step count (0 disables).
	MaxSteps int
	// RequireCompletion requires a non-error final assistant text response.
	RequireCompletion bool
}

type Metrics struct {
	// ToolSelectionCorrect is true when expected tools were selected in order.
	ToolSelectionCorrect bool
	// ArgsValid is true when required tool arguments were present and non-empty.
	ArgsValid bool
	// StepCount is the number of assistant turns observed.
	StepCount int
	// Completed is true when the run finished with a final assistant message and no error.
	Completed bool
	// ToolsCalled lists tool names executed during the run.
	ToolsCalled []string
}

type Scenario struct {
	// Name identifies the scenario in table tests.
	Name string
	// Prompt is the user message.
	Prompt string
	// Scripts are FakeModel responses in order.
	Scripts []agentllm.ResponseScript
	// Tools are registered for the run.
	Tools []tool.Tool
	// Expect defines scoring criteria.
	Expect Expectation
}