Expand golden QA benchmark scenarios

backend/tests/test_sprint12_golden_qa_benchmark.py

@@ -27,7 +27,21 @@ def test_golden_qa_expected_metrics_are_documented() -> None:
     assert expected["tolerance"] <= 1e-9
 
 
-def test_golden_qa_benchmark_command_passes_and_reports_persistence() -> None:
+def test_golden_qa_benchmark_manifest_covers_multiple_regression_scenarios() -> None:
+    manifest_path = ROOT / "fixtures" / "golden" / "golden_qa_benchmarks.json"
+    manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
+    scenario_ids = {scenario["benchmark_id"] for scenario in manifest["scenarios"]}
+
+    assert manifest["version"] == 1
+    assert {
+        "golden-buildings-partial-match-v1",
+        "golden-buildings-perfect-match-v1",
+        "golden-buildings-no-overlap-v1",
+        "golden-buildings-multipolygon-match-v1",
+    }.issubset(scenario_ids)
+
+
+def test_golden_qa_benchmark_command_passes_all_scenarios_and_reports_persistence() -> None:
     script = ROOT / "scripts" / "run_golden_qa_benchmark.py"
     result = subprocess.run(
         [sys.executable, str(script), "--json"],
@@ -39,14 +53,33 @@ def test_golden_qa_benchmark_command_passes_and_reports_persistence() -> None:
     payload = json.loads(result.stdout)
 
     assert payload["status"] == "passed"
-    assert payload["benchmark_id"] == "golden-buildings-partial-match-v1"
-    assert payload["metrics"]["precision"] == 0.5
-    assert payload["metrics"]["recall"] == 0.5
-    assert payload["metrics"]["f1"] == 0.5
-    assert payload["metrics"]["false_positive_count"] == 1
-    assert payload["metrics"]["false_negative_count"] == 1
-    assert payload["persistence"]["quality_check_count"] == 1
-    assert payload["persistence"]["metric_count"] == 6
+    assert payload["scenario_count"] >= 4
+    scenarios = {scenario["benchmark_id"]: scenario for scenario in payload["scenarios"]}
+    partial = scenarios["golden-buildings-partial-match-v1"]
+    perfect = scenarios["golden-buildings-perfect-match-v1"]
+    no_overlap = scenarios["golden-buildings-no-overlap-v1"]
+    multipolygon = scenarios["golden-buildings-multipolygon-match-v1"]
+
+    assert partial["metrics"]["precision"] == 0.5
+    assert partial["metrics"]["recall"] == 0.5
+    assert partial["metrics"]["f1"] == 0.5
+    assert partial["metrics"]["false_positive_count"] == 1
+    assert partial["metrics"]["false_negative_count"] == 1
+    assert perfect["metrics"]["precision"] == 1.0
+    assert perfect["metrics"]["recall"] == 1.0
+    assert perfect["metrics"]["f1"] == 1.0
+    assert perfect["metrics"]["mean_iou"] == 1.0
+    assert no_overlap["metrics"]["precision"] == 0.0
+    assert no_overlap["metrics"]["recall"] == 0.0
+    assert no_overlap["metrics"]["f1"] is None
+    assert no_overlap["metrics"]["mean_iou"] is None
+    assert multipolygon["metrics"]["precision"] == 1.0
+    assert multipolygon["metrics"]["recall"] == 1.0
+    assert multipolygon["metrics"]["f1"] == 1.0
+    assert multipolygon["metrics"]["mean_iou"] == 1.0
+
+    assert payload["persistence"]["quality_check_count"] == payload["scenario_count"]
+    assert payload["persistence"]["metric_count"] == payload["scenario_count"] * 6
     assert sorted(payload["persistence"]["metric_keys"]) == [
         "f1",
         "false_negative_count",
@@ -63,6 +96,7 @@ def test_golden_qa_shell_wrapper_is_safe_and_documented() -> None:
 
     assert "set -euo pipefail" in content
     assert "run_golden_qa_benchmark.py --json" in content
+    assert "import sys, geoalchemy2" in content
 
     result = subprocess.run(
         ["bash", "-n", "scripts/verify_golden_qa_benchmark.sh"],