Files
agent-template/evals/assert_rules.py
T
Admin_IT-one d406da1312
Deploy / validate-image (push) Successful in 45s
Eval Gate / evaluate (push) Failing after 1m42s
Add IT-One AI agent template, eval gate, and deployment skeleton
2026-08-21 12:14:55 +02:00

9 lines
432 B
Python

def score_case(case: dict, result: dict) -> tuple[bool, str]:
answer = str(result.get("answer", ""))
expected = str(case.get("expected_answer_contains", ""))
if expected and expected.lower() not in answer.lower():
return False, "answer substring mismatch"
if len(result.get("tool_calls", [])) != int(case.get("expected_tool_calls", 0)):
return False, "tool call count mismatch"
return True, "ok"