주기 실회기 검증과 G7 종료계약 보강

scripts/check-g7-human-voice-gain.py

@@ -0,0 +1,129 @@
+#!/usr/bin/env python3
+"""Validate and evaluate one independent human-held-out G7 voice-gain pack.
+
+The command emits only aggregate metrics and PII-safe JSON pointers. It never
+echoes input paths, participant/labeler keys, labels, or raw validation values.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from pathlib import Path
+from typing import Any, Iterable
+
+from pydantic import ValidationError
+
+
+REPO_ROOT = Path(__file__).resolve().parents[1]
+API_ROOT = REPO_ROOT / "apps/api"
+if str(API_ROOT) not in sys.path:
+    sys.path.insert(0, str(API_ROOT))
+
+from app.contracts.g7_external_evidence import (  # noqa: E402
+    G7HumanVoiceGainEvidencePack,
+)
+from app.services.g7_voice_gain_evidence import (  # noqa: E402
+    evaluate_human_voice_gain,
+)
+
+
+def _json_pointer(location: tuple[int | str, ...]) -> str:
+    if not location:
+        return "/"
+    parts = []
+    for item in location:
+        value = str(item).replace("~", "~0").replace("/", "~1")
+        parts.append(value)
+    return "/" + "/".join(parts)
+
+
+def _base_report() -> dict[str, Any]:
+    return {
+        "schema_version": "vignette.g7-human-voice-gain-check.v1",
+        "passed": False,
+        "clinical_claim_allowed": False,
+        "privacy_boundary": {
+            "input_path_logged": False,
+            "participant_keys_logged": False,
+            "labeler_keys_logged": False,
+            "labels_logged": False,
+            "raw_validation_values_logged": False,
+        },
+        "validation_errors": [],
+        "result": {},
+    }
+
+
+def validate_payload(payload: object) -> dict[str, Any]:
+    report = _base_report()
+    try:
+        pack = G7HumanVoiceGainEvidencePack.model_validate(payload)
+    except ValidationError as exc:
+        report["validation_errors"] = [
+            {
+                "pointer": _json_pointer(tuple(item["loc"])),
+                "type": item["type"],
+            }
+            for item in exc.errors(
+                include_url=False,
+                include_context=False,
+                include_input=False,
+            )
+        ]
+        return report
+
+    try:
+        result = evaluate_human_voice_gain(pack)
+    except Exception as exc:
+        report["validation_errors"] = [
+            {"pointer": "/", "type": f"evaluation:{type(exc).__name__}"}
+        ]
+        return report
+
+    report["passed"] = result.passed
+    report["result"] = result.model_dump(mode="json")
+    return report
+
+
+def parser() -> argparse.ArgumentParser:
+    result = argparse.ArgumentParser(description=__doc__)
+    mode = result.add_mutually_exclusive_group(required=True)
+    mode.add_argument("--input", type=Path, help="deidentified human pack JSON")
+    mode.add_argument(
+        "--print-schema",
+        action="store_true",
+        help="print the authoritative JSON Schema and exit",
+    )
+    return result
+
+
+def main(argv: Iterable[str] | None = None) -> int:
+    args = parser().parse_args(list(argv) if argv is not None else None)
+    if args.print_schema:
+        print(
+            json.dumps(
+                G7HumanVoiceGainEvidencePack.model_json_schema(),
+                ensure_ascii=False,
+                indent=2,
+                sort_keys=True,
+            )
+        )
+        return 0
+
+    try:
+        payload = json.loads(args.input.read_text(encoding="utf-8"))
+    except (OSError, UnicodeError, json.JSONDecodeError) as exc:
+        report = _base_report()
+        report["validation_errors"] = [
+            {"pointer": "/", "type": f"input:{type(exc).__name__}"}
+        ]
+    else:
+        report = validate_payload(payload)
+    print(json.dumps(report, ensure_ascii=False, indent=2, sort_keys=True))
+    return 0 if report["passed"] is True else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())