Add detection threshold calibration comparison

backend/tests/test_sprint133_detection_threshold_calibration_ux.py

@@ -0,0 +1,37 @@
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[2]
+
+
+def test_detection_lab_exposes_persisted_threshold_calibration_comparison() -> None:
+    detection_lab = ROOT / "frontend" / "src" / "components" / "detection" / "DetectionLab.tsx"
+    app = ROOT / "frontend" / "src" / "App.tsx"
+    todo = ROOT / "docs" / "TODO.md"
+
+    source = detection_lab.read_text(encoding="utf-8")
+    app_source = app.read_text(encoding="utf-8")
+    todo_source = todo.read_text(encoding="utf-8")
+
+    assert "qualityChecks: QualityCheckRead[]" in source
+    assert "buildCalibrationRows(detectionRuns, qualityChecks)" in source
+    assert "Calibration comparison" in source
+    assert "Compare persisted detection runs by confidence threshold" in source
+    assert "Best F1 candidate" in source
+    assert "Best precision candidate" in source
+    assert "Lowest false-positive pressure" in source
+    assert "Threshold" in source
+    assert "Precision" in source
+    assert "Recall" in source
+    assert "F1" in source
+    assert "False positives" in source
+    assert "False negatives" in source
+    assert "Promote only after checking evidence across AOIs" in source
+    assert "No calibration comparison available yet" in source
+    assert "metricValue(check, 'f1')" in source
+    assert "metricValue(check, 'precision')" in source
+    assert "metricValue(check, 'recall')" in source
+    assert "metricValue(check, 'false_positives')" in source
+    assert "confidenceThresholdForRun(run)" in source
+    assert "qualityChecks={qualityChecks}" in app_source
+    assert "[x] Add full threshold calibration comparison UX" in todo_source