Camerasv1.0.1

Decision Tree

| Task | Reference |

|------|-----------|

| Connect to security cameras (Ring, Nest, IP cams) | Check security-integration.md |

| Capture from webcams or USB cameras | Check capture.md |

| Control DSLR/mirrorless (tethering, remote shoot) | Check photography-control.md |

| Process video (detection, recognition) | Check processing.md |

| Choose or compare cameras (buying guide) | Check buying-guide.md |

---

Core Capabilities

What an agent with this skill can do:

1. Capture snapshots from any connected camera on demand
2. Record short clips (10-60s) for review or sending
3. List available cameras on the system or network
4. Receive motion alerts from security systems
5. Control photography cameras (shoot, adjust settings, download)
6. Describe what the camera sees (using vision models)

---

Protocol Quick Reference

| Protocol | Use Case | Access Method |

|----------|----------|---------------|

| RTSP | IP cameras, NVRs | rtsp://user:pass@ip:554/stream |

| ONVIF | Discovery + control | python-onvif-zeep, auto-discover |

| HTTP/MJPEG | Simple IP cams | /snapshot.jpg, /video.mjpg |

| Home Assistant | Unified access | REST API /api/camera_proxy/ |

| Frigate | Motion events + clips | MQTT + HTTP API |

| USB/V4L2 | Webcams, capture cards | ffmpeg, opencv, device index |

| gPhoto2 | DSLR/mirrorless control | USB PTP protocol |

---

Common Commands

# List cameras
ffmpeg -list_devices true -f avfoundation -i dummy  # macOS
v4l2-ctl --list-devices                              # Linux

# Snapshot from RTSP
ffmpeg -i "rtsp://user:pass@ip/stream" -frames:v 1 snapshot.jpg

# Snapshot from webcam
ffmpeg -f avfoundation -i "0" -frames:v 1 webcam.jpg  # macOS
ffmpeg -f v4l2 -i /dev/video0 -frames:v 1 webcam.jpg  # Linux

# Record 10s clip
ffmpeg -i "rtsp://ip/stream" -t 10 -c copy clip.mp4

---

Integration Patterns

With Home Assistant

If cameras are already in HA, use the REST API:

GET /api/camera_proxy/camera.front_door
→ Returns JPEG snapshot

With Frigate (recommended for security)

Frigate handles detection. Agent just listens:

• MQTT: frigate/events for motion alerts
• HTTP: /api/events/{id}/snapshot.jpg

With Vision Models

Capture snapshot → send to vision model for description:

1. ffmpeg → snapshot.jpg
2. Vision API → "A person standing at the front door"
3. Return description to user

---

Troubleshooting

| Problem | Solution |

|---------|----------|

| Camera offline | Check network, power, IP hasn't changed |

| RTSP timeout | Try adding ?tcp or use port 8554 |

| Permission denied | Run with sudo or add user to video group |

| No video, only audio | Wrong stream path, try /stream1, /ch01/main |

| gPhoto2 camera busy | Close other apps using camera, replug USB |