Smart NVR - Getting Started#

Overview#

Smart NVR is a GenAI-powered video analytics application that transforms traditional network video recorders with intelligent event detection and real-time insights at the edge. This guide will walk you through deploying and configuring the application to extract valuable insights from your video data.

Prerequisites#

System Requirements#

System must meet minimum requirements
3-4 devices for distributed deployment

Smart NVR operates in a distributed architecture requiring multiple services across 3-4 devices for optimal performance:

Device	Service	Purpose
Device 1	VSS Search	Video search functionality
Device 2	VSS Summary	Video summarization
Device 3	VLM Microservice	AI-powered event descriptions (optional)
Device 3/4	Smart NVR App	Main application interface

Software Dependencies#

Docker: Installation Guide
- Must be configured to run without sudo (Post-install guide)
Git: Installation Guide

Required Services#

Before setting up Smart NVR, ensure these services are running on their respective devices:

1. VSS (Video Search and Summarization) Services#

Deploy these on separate devices:

VSS Search: Handles video search functionality
VSS Summary: Provides video summarization capabilities

📖 VSS Documentation

2. VLM Microservice (Optional)#

Required only when enabling AI-powered event descriptions (NVR_GENAI=true):

Runs the VLM model defined in the frigate config file
Use VLM_MAX_COMPLETION_TOKENS to limit response length during deployment

📖 VLM Serving Documentation

Quick Start#

Step 1: Clone the repo#

# Clone the repository
git clone https://github.com/open-edge-platform/edge-ai-suites.git
cd edge-ai-suites/metro-ai-suite/smart-nvr

Step 2: Configure Environment#

Set up the required environment variables:

# Docker Registry Details
export REGISTRY_URL="intel"
export TAG="1.2.3"

# VSS Service Endpoints
export http_proxy=<http-proxy>
export https_proxy=<https-proxy>
export no_proxy=<no_proxy>

# VSS Service Endpoints
export VSS_SUMMARY_IP=<vss-summary-device-ip>
export VSS_SUMMARY_PORT=<vss-summary-port>        # Default: 12345
export VSS_SEARCH_IP=<vss-search-device-ip>
export VSS_SEARCH_PORT=<vss-search-port>          # Default: 12345

# MQTT Configuration
export MQTT_USER=<mqtt-username>
export MQTT_PASSWORD=<mqtt-password>

# Feature Toggles
export NVR_GENAI=false                  # Set to 'true' to enable AI-powered event descriptions
export NVR_SCENESCAPE=false             # Set to 'true' to enable Scenescape integration

Step 3: Launch Application#

# Start all services
source setup.sh start

This launches all required containers:

Services overview

Step 4: Access the Interface#

Open your browser and navigate to:

http://<host-ip>:7860

Step 5: Stop Services#

# Stop all services when done
source setup.sh stop

Advanced Configuration#

Enabling AI-Powered Event Descriptions#

To enable Smart NVR’s GenAI capabilities for intelligent event descriptions:

1. Update Frigate Configuration#

Modify resources/frigate-config/config.yml:

genai:
  enabled: true

2. Ensure VLM Service Availability#

Verify the VLM microservice is running and accessible at the configured endpoint.

3. Set Environment Variable#

export NVR_GENAI=true
export VLM_SERVING_IP=<vlm-serving-device-ip>
export VLM_SERVING_PORT=<vlm-serving-port>

4. Run the application#

Re-run the application after configuring the rest of environment variables. Ensure that the environment value export NVR_GENAI=true is set.

⚠️ Important Notes:

This feature is experimental and may be unstable due to underlying Frigate GenAI implementation

Requires VLM microservice to be running

Disabled by default for system stability

Running Tests and Generating Coverage Report#

To ensure the functionality of the microservice and measure test coverage, follow these steps:

Install Dependencies Install the required dependencies, including development dependencies, using:
```
poetry install --with test
```
Run Tests with Poetry Use the following command to run all tests:
```
poetry run pytest
```

Run Tests with Coverage To collect coverage data while running tests, use:

poetry run pytest --cov=src --cov=ui --cov-report=term-missing:skip-covered

Generate Coverage Report After running the tests, generate a coverage report:
```
poetry run coverage report -m
```
Generate HTML Coverage Report (Optional) For a detailed view, generate an HTML report:
```
poetry run coverage html
```
Open the htmlcov/index.html file in your browser to view the report.

These steps will help you verify the functionality of the microservice and ensure adequate test coverage.

Scenescape Integration#

For traffic analytics capabilities with Intel Scenescape (vehicle counting, traffic flow analysis), see the Scenescape Integration Guide.

Custom Build Configuration#

If using custom build flags, ensure the same environment variables are set before running the setup script.

Next Steps#

Explore Features: Learn about application capabilities in the How to Use Guide
Troubleshooting: If you encounter issues, check the Troubleshooting Guide