VibeNVR – open-source self-hosted NVR with native MQTT + Home Assistant auto-discovery (v1.28)

Hi everyone!

I wanted to share VibeNVR, an open-source self-hosted NVR I’ve been building for IP cameras. The v1.28.x release introduces native MQTT integration with Home Assistant auto-discovery, which I think a lot of people here will find useful.

What it does

VibeNVR runs 100% locally via Docker Compose — no cloud, no subscription, no vendor lock-in. It connects to IP cameras via RTSP and ONVIF, records continuously or on motion/AI events, and now publishes everything to HA automatically.

Home Assistant integration (v1.28)

  • Native MQTT with auto-discovery: cameras, motion events and AI detections are automatically exposed as Home Assistant entities — no manual YAML configuration needed. Just point VibeNVR at your MQTT broker and entities appear.
  • Works alongside your existing HA setup, cameras remain independent

AI detection

  • Switch between YOLOv8 (high accuracy) or MobileNet SSD (faster, lighter) at runtime
  • Google Coral TPU supported as singleton inference engine
  • Lean Mode: no AI accelerator needed, runs on standard hardware
  • Universal AI Master Switch: disable TFLite entirely to free memory, cameras auto-fallback to OpenCV motion detection

Other highlights

  • ONVIF auto-discovery for compatible cameras (Tapo, Reolink, Hikvision, Dahua tested)
  • Dual-stream support: sub-stream for live view (~80% less CPU), main stream for recording
  • Timeline with motion/AI event filtering
  • RBAC with viewer/admin isolation
  • Built with Python + FastAPI + React

Links

Happy to answer any questions or hear feedback from the HA community!