Add presence detection, person/vehicle AI detection, health monitoring

Task 1 — Presence: ping family phones, derive household state
(EMPTY/KIDS_HOME/ADULTS_HOME/ALL_HOME), configurable departure delay,
per-member roles, auto-arm actions via MQTT.

Task 2 — Detection: MobileNet-SSD v2 via OpenCV DNN for person/vehicle
classification. Vehicle color/size fingerprinting for known car matching.
Zone-based filtering per camera. Model download script.

Task 3 — Health: periodic disk/MQTT/subsystem checks, auto-prune oldest
non-starred recordings on disk pressure, daily digest builder.

126 tests passing.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tests/unit/test_detection.py

@@ -0,0 +1,74 @@
+"""Tests for person and vehicle detection."""
+
+import numpy as np
+
+from vigilar.detection.person import Detection, PersonDetector
+from vigilar.detection.vehicle import classify_dominant_color, classify_size
+from vigilar.detection.zones import filter_detections_by_zone
+
+
+class TestPersonDetector:
+    def test_detector_initializes_without_model(self):
+        detector = PersonDetector(model_path="nonexistent.pb", config_path="nonexistent.pbtxt")
+        assert not detector.is_loaded
+
+    def test_detect_returns_empty_when_not_loaded(self):
+        detector = PersonDetector(model_path="nonexistent.pb", config_path="nonexistent.pbtxt")
+        frame = np.zeros((480, 640, 3), dtype=np.uint8)
+        detections = detector.detect(frame)
+        assert detections == []
+
+    def test_detection_dataclass(self):
+        d = Detection(class_name="person", class_id=1, confidence=0.85, bbox=(10, 20, 100, 200))
+        assert d.class_name == "person"
+        assert d.confidence == 0.85
+
+
+class TestVehicleColor:
+    def test_white_detection(self):
+        white_region = np.full((100, 100, 3), 240, dtype=np.uint8)
+        color = classify_dominant_color(white_region)
+        assert color == "white"
+
+    def test_black_detection(self):
+        black_region = np.full((100, 100, 3), 15, dtype=np.uint8)
+        color = classify_dominant_color(black_region)
+        assert color == "black"
+
+    def test_size_compact(self):
+        assert classify_size(bbox_area=5000, zone_area=100000) == "compact"
+
+    def test_size_midsize(self):
+        assert classify_size(bbox_area=15000, zone_area=100000) == "midsize"
+
+    def test_size_large(self):
+        assert classify_size(bbox_area=30000, zone_area=100000) == "large"
+
+
+class TestZoneFiltering:
+    def test_detection_inside_zone(self):
+        detections = [Detection("person", 1, 0.9, (50, 50, 80, 80))]
+        zone_region = (0, 0, 200, 200)
+        filtered = filter_detections_by_zone(detections, zone_region, ["person"])
+        assert len(filtered) == 1
+
+    def test_detection_outside_zone(self):
+        detections = [Detection("person", 1, 0.9, (300, 300, 50, 50))]
+        zone_region = (0, 0, 200, 200)
+        filtered = filter_detections_by_zone(detections, zone_region, ["person"])
+        assert len(filtered) == 0
+
+    def test_filter_by_class(self):
+        detections = [
+            Detection("person", 1, 0.9, (50, 50, 80, 80)),
+            Detection("car", 3, 0.8, (50, 50, 80, 80)),
+        ]
+        zone_region = (0, 0, 200, 200)
+        filtered = filter_detections_by_zone(detections, zone_region, ["person"])
+        assert len(filtered) == 1
+        assert filtered[0].class_name == "person"
+
+    def test_no_zone_returns_all(self):
+        detections = [Detection("person", 1, 0.9, (50, 50, 80, 80))]
+        filtered = filter_detections_by_zone(detections, None, ["person", "car"])
+        assert len(filtered) == 1