"""Eval-harness CLI: ``python -m src.eval`` runs the golden dataset and

prints the markdown report. The default ``answer_fn`` echoes the question

verbatim — wire your real agent loop in by importing this module and

constructing ``EvalRunner(answer_fn=your_callable)``.

"""

from __future__ import annotations

from src.eval.report import generate_report

from src.eval.runner import EvalRunner

def _identity_answer(question: str) -> str:

"""Default placeholder agent — returns the question verbatim.

Real users wire a ``Callable[[str], str]`` that hits their LLM / agent

loop. This default keeps the CLI runnable without LLM credentials.

"""

return question

def main() -> int:

runner = EvalRunner(answer_fn=_identity_answer)

results = runner.evaluate_all()

report = generate_report(results)

print(report)

return 0 if all(r.pass_result for r in results) else 1

if __name__ == "__main__":

raise SystemExit(main())