| import pytest |
|
|
| from aegis.agents import EvaluatorAgent |
| from aegis.core import AgentContext, AgentResult, Task |
|
|
|
|
| def make_context(result: AgentResult | dict[str, object] | None) -> AgentContext: |
| inputs: dict[str, object] = {} |
|
|
| if result is not None: |
| inputs["result"] = result.model_dump() if isinstance(result, AgentResult) else result |
|
|
| return AgentContext( |
| task=Task( |
| name="Evaluate result", |
| description="Evaluate an agent execution result.", |
| ), |
| inputs=inputs, |
| ) |
|
|
|
|
| @pytest.mark.asyncio |
| async def test_evaluator_passes_valid_result() -> None: |
| evaluator = EvaluatorAgent() |
| context = make_context( |
| AgentResult( |
| agent_name="planner", |
| success=True, |
| output={"steps": ["analyze", "execute"]}, |
| ) |
| ) |
|
|
| result = await evaluator.run(context) |
|
|
| assert result.success is True |
| assert result.output["passed"] is True |
| assert result.output["score"] == 100 |
| assert result.output["issues"] == () |
| assert result.output["evaluated_agent"] == "planner" |
|
|
|
|
| @pytest.mark.asyncio |
| async def test_evaluator_fails_unsuccessful_result() -> None: |
| evaluator = EvaluatorAgent() |
| context = make_context( |
| AgentResult( |
| agent_name="tool", |
| success=False, |
| errors=["Execution failed."], |
| ) |
| ) |
|
|
| result = await evaluator.run(context) |
|
|
| assert result.success is True |
| assert result.output["passed"] is False |
| assert result.output["score"] == 25 |
| assert len(result.output["issues"]) == 3 |
|
|
|
|
| @pytest.mark.asyncio |
| async def test_evaluator_rejects_missing_result() -> None: |
| evaluator = EvaluatorAgent() |
| context = make_context(None) |
|
|
| result = await evaluator.run(context) |
|
|
| assert result.success is False |
| assert result.errors == ["Missing required input: result"] |
|
|
|
|
| @pytest.mark.asyncio |
| async def test_evaluator_rejects_invalid_result() -> None: |
| evaluator = EvaluatorAgent() |
| context = make_context( |
| { |
| "agent_name": "", |
| "success": True, |
| } |
| ) |
|
|
| result = await evaluator.run(context) |
|
|
| assert result.success is False |
| assert result.errors |
| assert result.errors[0].startswith("Invalid agent result:") |
|
|
