Function _evaluate_legacy

python/agents/invoice-processing/eval/eval.py:742–777

Evaluate a case using hardcoded invoice-specific comparators (legacy fallback).

(
    case_id: str,
    gt: dict,
    agent: dict,
    decision: dict,
    tolerance: float,
)

def _evaluate_legacy(
    case_id: str,
    gt: dict,
    agent: dict,
    decision: dict,
    tolerance: float,
) -> dict:
    """Evaluate a case using hardcoded invoice-specific comparators (legacy fallback)."""
    financials = compare_financials(gt, agent, tolerance)
    vendor = compare_vendor(gt, agent)
    header = compare_header(gt, agent)
    line_items = compare_line_items(gt, agent, tolerance)

    all_correct = (
        decision["match"]
        and financials["all_match"]
        and vendor["all_match"]
        and header["all_match"]
        and line_items["all_match"]
    )

    mismatches = _collect_legacy_mismatches(
        decision, financials, vendor, header, line_items
    )

    return {
        "case_id": case_id,
        "status": "PASS" if all_correct else "FAIL",
        "all_correct": all_correct,
        "decision": decision,
        "financials": financials,
        "vendor": vendor,
        "header": header,
        "line_items": line_items,
        "mismatches": mismatches,
    }


def _collect_field_mismatches(section: dict, labels: list[str]) -> list[str]: