273 lines
11 KiB
Python
273 lines
11 KiB
Python
from __future__ import annotations
|
|
|
|
import argparse
|
|
import json
|
|
import sys
|
|
from datetime import UTC, datetime
|
|
from hashlib import sha256
|
|
from pathlib import Path
|
|
from uuid import uuid4
|
|
|
|
|
|
ROOT = Path(__file__).resolve().parents[1]
|
|
BACKEND_ROOT = ROOT / "backend"
|
|
if str(BACKEND_ROOT) not in sys.path:
|
|
sys.path.insert(0, str(BACKEND_ROOT))
|
|
|
|
from app.models import Dataset, Metric, QualityCheck, SourceRegistry, SourceSnapshot # noqa: E402
|
|
from app.services.qa_service import QaService # noqa: E402
|
|
from app.services.quality_service import QualityService # noqa: E402
|
|
from app.services.source_registry_service import SourceRegistryService # noqa: E402
|
|
|
|
|
|
class BenchmarkSession:
|
|
def __init__(self, datasets: list[Dataset]) -> None:
|
|
self.datasets = {dataset.id: dataset for dataset in datasets}
|
|
self.added: list[object] = []
|
|
self.commits = 0
|
|
self.refreshes: list[object] = []
|
|
|
|
def get(self, model, item_id):
|
|
if model.__name__ == "Dataset":
|
|
return self.datasets.get(item_id)
|
|
if model.__name__ == "Area":
|
|
return None
|
|
return None
|
|
|
|
def add(self, item) -> None:
|
|
self.added.append(item)
|
|
|
|
def commit(self) -> None:
|
|
self.commits += 1
|
|
|
|
def refresh(self, item) -> None:
|
|
self.refreshes.append(item)
|
|
|
|
|
|
def _load_manifest() -> dict:
|
|
manifest_path = ROOT / "fixtures" / "golden" / "golden_qa_benchmarks.json"
|
|
if manifest_path.exists():
|
|
manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
|
|
if not isinstance(manifest.get("scenarios"), list) or not manifest["scenarios"]:
|
|
raise AssertionError("golden_qa_benchmarks.json must define at least one scenario")
|
|
return manifest
|
|
|
|
# Backwards-compatible fallback for older checkouts and local smoke scripts.
|
|
expected = json.loads((ROOT / "fixtures" / "golden" / "expected_qa_metrics.json").read_text(encoding="utf-8"))
|
|
return {"version": 0, "scenarios": [expected]}
|
|
|
|
|
|
def _dataset(dataset_id, project_id, name: str, path: Path, *, role: str) -> Dataset:
|
|
# The benchmark is still synthetic evidence, but it intentionally models
|
|
# the same complete Phase 2 provenance binding required at the QA service
|
|
# boundary. The candidate is a derived fixture; the reference simulates
|
|
# a governed GRB building snapshot. No legacy fixture exception is used.
|
|
source_key = "grb" if role == "reference" else "derived"
|
|
source = SourceRegistry(
|
|
id=uuid4(),
|
|
**SourceRegistryService.definition_for(source_key).as_model_values(),
|
|
)
|
|
checksum = sha256(path.read_bytes()).hexdigest()
|
|
snapshot = SourceSnapshot(
|
|
id=uuid4(),
|
|
source_registry_id=source.id,
|
|
snapshot_key=f"golden-qa:{source_key}:{checksum}",
|
|
checksum_sha256=checksum,
|
|
fetched_at=datetime.now(UTC),
|
|
crs="EPSG:4326",
|
|
units=source.default_units,
|
|
spatial_resolution_json={"status": "fixture"},
|
|
temporal_coverage_json={"status": "fixture"},
|
|
geographic_coverage_json={"scope": "golden-qa-fixture"},
|
|
observed_schema_json={"dataset_type": "vector", "fixture_mode": True},
|
|
freshness_status="current",
|
|
ingest_status="ingested",
|
|
known_limitations_json=["Synthetic golden benchmark fixture; not production evidence."],
|
|
snapshot_metadata_json={"fixture_mode": True, "benchmark": "golden-qa"},
|
|
)
|
|
return Dataset(
|
|
id=dataset_id,
|
|
project_id=project_id,
|
|
name=name,
|
|
dataset_type="vector",
|
|
source="governed golden benchmark fixture",
|
|
dataset_role="reference" if role == "reference" else "derived",
|
|
source_name=source.source_key,
|
|
reference_layer_name="buildings" if role == "reference" else None,
|
|
storage_path=str(path),
|
|
crs="EPSG:4326",
|
|
checksum_sha256=checksum,
|
|
source_registry_id=source.id,
|
|
source_snapshot_id=snapshot.id,
|
|
source_registry=source,
|
|
source_snapshot=snapshot,
|
|
data_contract_key="geointel.vector.geojson",
|
|
data_contract_version="1.0.0",
|
|
validation_status="passed",
|
|
provenance_status="complete",
|
|
lineage_status="not_applicable",
|
|
quarantine_status="not_quarantined",
|
|
metadata_json={"crs_assumed": False, "fixture_mode": True},
|
|
source_metadata={"fixture_mode": True, "source_registry_key": source.source_key},
|
|
provenance_metadata={"fixture_mode": True, "source_snapshot_id": str(snapshot.id)},
|
|
status="ready",
|
|
)
|
|
|
|
|
|
def _assert_close(label: str, actual: float | int | None, expected: float | int | None, tolerance: float) -> None:
|
|
if expected is None:
|
|
if actual is not None:
|
|
raise AssertionError(f"{label} drifted: actual={actual}, expected=None")
|
|
return
|
|
if actual is None:
|
|
raise AssertionError(f"{label} is None, expected {expected}")
|
|
if abs(float(actual) - float(expected)) > tolerance:
|
|
raise AssertionError(f"{label} drifted: actual={actual}, expected={expected}, tolerance={tolerance}")
|
|
|
|
|
|
def _run_scenario(expected: dict, session: BenchmarkSession, project_id) -> dict:
|
|
added_before = len(session.added)
|
|
commits_before = session.commits
|
|
candidate_dataset_id = uuid4()
|
|
reference_dataset_id = uuid4()
|
|
candidate_path = ROOT / expected["candidate_fixture"]
|
|
reference_path = ROOT / expected["reference_fixture"]
|
|
tolerance = float(expected["tolerance"])
|
|
|
|
session.datasets[candidate_dataset_id] = _dataset(
|
|
candidate_dataset_id,
|
|
project_id,
|
|
Path(expected["candidate_fixture"]).name,
|
|
candidate_path,
|
|
role="source",
|
|
)
|
|
session.datasets[reference_dataset_id] = _dataset(
|
|
reference_dataset_id,
|
|
project_id,
|
|
Path(expected["reference_fixture"]).name,
|
|
reference_path,
|
|
role="reference",
|
|
)
|
|
|
|
result = QaService.compare_candidate_with_reference(
|
|
db=session,
|
|
project_id=project_id,
|
|
candidate_dataset_id=candidate_dataset_id,
|
|
reference_dataset_id=reference_dataset_id,
|
|
iou_threshold=float(expected["iou_threshold"]),
|
|
)
|
|
|
|
metrics = {
|
|
"precision": result.precision,
|
|
"recall": result.recall,
|
|
"f1": result.f1_score,
|
|
"mean_iou": result.mean_iou,
|
|
"false_positive_count": result.false_positives,
|
|
"false_negative_count": result.false_negatives,
|
|
}
|
|
|
|
_assert_close("candidate_feature_count", result.candidate_feature_count, expected["candidate_feature_count"], 0)
|
|
_assert_close("reference_feature_count", result.reference_feature_count, expected["reference_feature_count"], 0)
|
|
_assert_close("matches", result.matches, expected["matches"], 0)
|
|
_assert_close("false_positive_count", result.false_positives, expected["false_positive_count"], 0)
|
|
_assert_close("false_negative_count", result.false_negatives, expected["false_negative_count"], 0)
|
|
_assert_close("precision", result.precision, expected["precision"], tolerance)
|
|
_assert_close("recall", result.recall, expected["recall"], tolerance)
|
|
_assert_close("f1", result.f1_score, expected["f1"], tolerance)
|
|
_assert_close("mean_iou", result.mean_iou, expected["mean_iou"], tolerance)
|
|
|
|
quality_check = QualityService.persist_quality_check(
|
|
db=session,
|
|
project_id=project_id,
|
|
reference_dataset_id=reference_dataset_id,
|
|
check_type="golden_candidate_vs_reference",
|
|
status=result.status,
|
|
score=result.f1_score,
|
|
parameters={"iou_threshold": result.iou_threshold, "benchmark_id": expected["benchmark_id"]},
|
|
findings={
|
|
"matches": result.matches,
|
|
"false_positives": result.false_positives,
|
|
"false_negatives": result.false_negatives,
|
|
},
|
|
candidate_dataset_id=candidate_dataset_id,
|
|
metrics=metrics,
|
|
)
|
|
scenario_added = session.added[added_before:]
|
|
persisted_metrics = [item for item in scenario_added if isinstance(item, Metric)]
|
|
|
|
return {
|
|
"status": "passed",
|
|
"benchmark_id": expected["benchmark_id"],
|
|
"description": expected.get("description"),
|
|
"metrics": metrics,
|
|
"result_counts": {
|
|
"candidate_feature_count": result.candidate_feature_count,
|
|
"reference_feature_count": result.reference_feature_count,
|
|
"matches": result.matches,
|
|
},
|
|
"quality_check_id": str(quality_check.id),
|
|
"persistence": {
|
|
"quality_check_count": len([item for item in scenario_added if isinstance(item, QualityCheck)]),
|
|
"metric_count": len(persisted_metrics),
|
|
"metric_keys": [metric.metric_key for metric in persisted_metrics],
|
|
"commit_count": session.commits - commits_before,
|
|
},
|
|
"fixtures": {
|
|
"candidate": expected["candidate_fixture"],
|
|
"reference": expected["reference_fixture"],
|
|
},
|
|
}
|
|
|
|
|
|
def run_benchmark() -> dict:
|
|
manifest = _load_manifest()
|
|
project_id = uuid4()
|
|
session = BenchmarkSession([])
|
|
scenarios = [_run_scenario(expected, session, project_id) for expected in manifest["scenarios"]]
|
|
persisted_metrics = [item for item in session.added if isinstance(item, Metric)]
|
|
aggregate_metric_keys = sorted({metric.metric_key for metric in persisted_metrics})
|
|
first = scenarios[0]
|
|
|
|
return {
|
|
"status": "passed",
|
|
"version": manifest.get("version"),
|
|
"scenario_count": len(scenarios),
|
|
"scenarios": scenarios,
|
|
"persistence": {
|
|
"quality_check_count": len([item for item in session.added if isinstance(item, QualityCheck)]),
|
|
"metric_count": len(persisted_metrics),
|
|
"metric_keys": aggregate_metric_keys,
|
|
"commit_count": session.commits,
|
|
},
|
|
# Compatibility fields for scripts that still read the original single-scenario shape.
|
|
"benchmark_id": first["benchmark_id"],
|
|
"metrics": first["metrics"],
|
|
"fixtures": first["fixtures"],
|
|
}
|
|
|
|
|
|
def main() -> int:
|
|
parser = argparse.ArgumentParser(description="Run GeoIntel golden QA/QC benchmark.")
|
|
parser.add_argument("--json", action="store_true", help="Print machine-readable JSON only.")
|
|
args = parser.parse_args()
|
|
|
|
try:
|
|
payload = run_benchmark()
|
|
except Exception as exc:
|
|
if args.json:
|
|
print(json.dumps({"status": "failed", "error": str(exc)}, indent=2))
|
|
else:
|
|
print(f"Golden QA/QC benchmark failed: {exc}", file=sys.stderr)
|
|
return 1
|
|
|
|
if args.json:
|
|
print(json.dumps(payload, indent=2, sort_keys=True))
|
|
else:
|
|
print("GeoIntel golden QA/QC benchmark passed")
|
|
print(json.dumps(payload, indent=2, sort_keys=True))
|
|
return 0
|
|
|
|
|
|
if __name__ == "__main__":
|
|
raise SystemExit(main())
|