Skip to content

Latest commit

 

History

History
37 lines (28 loc) · 1.74 KB

File metadata and controls

37 lines (28 loc) · 1.74 KB

How It Works

This document describes the end‑to‑end architecture of Live Video Search and how NVR Event Router and VSS Search integrate.

High‑Level Architecture

graph TD
  A[Camera Streams] -->|RTSP/Video Feeds| B[Frigate NVR]
  B -->|Event Clips + Metadata| C[NVR Event Router]
  C -->|Watched camera ingestion| H[Pipeline Manager]
  H --> E[VSS Search‑MS]
  E --> F[VDMS DataPrep]
  F --> G[VDMS VectorDB]
  H --> I[VSS UI Configure Cameras and Search]
  K[Telemetry Collector] --> H
Loading

Data Flow

  1. Ingestion: Cameras stream into Frigate, which records clips and publishes events via MQTT.
  2. Event Routing: NVR Event Router receives events and associates clips with camera metadata.
  3. Indexing: VSS camera configuration enables watcher-based clip ingestion to Pipeline Manager, which forwards clips to DataPrep. Embeddings are stored in VDMS.
  4. Querying: Users query VSS UI with optional time‑range filters. Search‑MS retrieves and ranks relevant clips.
  5. Visualization: Results are shown directly in VSS UI.
  6. Telemetry: Collector streams system metrics to Pipeline Manager and the UI.

Integration Points

  • Watcher-based ingestion path ties enabled camera clips directly to VSS Search input.
  • Pipeline Manager endpoints unify search configuration and retrieval.
  • Telemetry WS provides live metrics for operational visibility.

Related Architecture References