VibeNVR v1.28.x: Universal AI Switch, MQTT, and Multi-Model TFLite Support

#opensource #cctv #ai #privacy

VibeNVR v1.28.0 and v1.28.1 introduce a major architectural shift in how the AI engine is managed.

v1.28.0 - The AI & Connectivity Update

This release lays the foundation for a proper multi-model AI pipeline:

Multi-model TFLite support: choose between YOLOv8 and MobileNet SSD v2, optimizing for precision or speed
Global AI Engine: model and hardware selection moved to System Settings for centralized management
NMS (Non-Maximum Suppression): eliminates redundant bounding boxes, with configurable IoU Threshold for YOLOv8
Hysteresis (1.0s persistence): smoother object tracking, no more flickering bounding boxes
Native MQTT Service: real-time event reporting with Home Assistant Auto-Discovery - binary sensors and status entities registered automatically, no YAML config needed
Global Settings Sync: MQTT, AI, and Throttling settings propagated in real-time to all camera threads
Dashboard Health Widgets: live MQTT Connection and AI Processor status indicators

v1.28.1 - Universal AI Activation Framework

The AI engine is now a managed singleton with a global master switch:

Universal AI Master Switch: a single toggle in System Settings. Turn it OFF and TFLite models are fully released from memory - significant CPU and RAM savings on constrained hardware
Intelligent OpenCV Fallback: cameras configured for AI detection automatically fall back to standard OpenCV motion detection when the engine is off
Singleton Engine Architecture: prevents redundant model loads across camera threads, with real-time config synchronization
Visual State Indicators: amber warnings and status banners in the Camera Edit modal when running in fallback mode
RBAC Hardening: global AI configuration endpoints restricted to Admin role only

Why it matters

For homelab and resource-constrained deployments, the ability to fully unload AI models from memory is significant. You can run the full AI pipeline when needed and drop back to lightweight motion detection otherwise - without reconfiguring individual cameras.

The MQTT Auto-Discovery integration means VibeNVR events now show up natively in Home Assistant as binary sensors, making automation rules straightforward to build.