e1171e8ff8
Motion detection now optionally runs MobileNet V2 SSD (COCO, quantized) on frames that trigger motion, identifying objects like people, cats, and cars. Events without detected objects are suppressed by default. Snapshots include bounding box annotations. New MCP tool vision_get_detections() enables label-based queries. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
vixy-vision Event Collector
Receives motion events from camera servers and stores them for Vixy to review.
Quick Start (macOS)
./setup-macos.sh
launchctl load ~/Library/LaunchAgents/com.vixy.vision-collector.plist
How It Works
Pi (camera) Mac mini (collector) Vixy (MCP)
┌──────────┐ ┌──────────────────┐ ┌──────────┐
│ motion │ POST │ collector.py │ read │ query │
│ detected ├────────►│ ├─events.db │◄────────┤ annotate │
│ │ /events │ └─snapshots/ │ │ review │
└──────────┘ └──────────────────┘ └──────────┘
Data Storage
Events are stored in ~/Documents/Vixy/events/:
events/
├── events.db # SQLite database
└── snapshots/
└── 2024-12-16/ # Date-organized images
└── basement-*.jpg
API Endpoints
| Endpoint | Method | Description |
|---|---|---|
/ |
GET | Service info |
/health |
GET | Health check |
/events |
POST | Receive event from camera |
/events |
GET | List events (debug) |
/stats |
GET | Event statistics |
Event Payload
Camera servers POST to /events:
{
"event": {
"timestamp": "2024-12-16T14:23:01Z",
"camera_id": "basement",
"event_type": "motion",
"confidence": 0.75,
"area_percent": 7.5
},
"snapshot": "<base64 JPEG>"
}
MCP Tools
Once events are collected, Vixy can:
vision_get_events()- Query eventsvision_get_event_snapshot(id)- View snapshotvision_annotate_event(id, text, tags)- Add meaningvision_event_stats()- See statistics