Quick Start

Follow these steps to build and run the DSX Blueprint for AI Factories locally. You need two terminal windows. Use one for the Kit application (streaming server). Use the other for the web frontend.

Time expectations:

• Initial setup and build: 15-20 minutes

• Subsequent launches: 2-3 minutes

• Shader compilation on first launch: 5-8 minutes

Step 1: Clone the Repository

1. Open a terminal window.

2. Clone the repository:

git clone https://github.com/NVIDIA-Omniverse-blueprints/omniverse-dsx-blueprint-for-ai-factories.git
cd omniverse-dsx-blueprint-for-ai-factories

Verify the repository cloned successfully:

ls -la

# Compare this with the contents of the directory you cloned from. If they match, the repository cloned successfully.

Step 2: Download and Configure USD Scene Data

The blueprint requires a USD scene dataset that is hosted separately from the repository.

1. Download the DSX Content Pack from NGC.

2. Extract the archive to a location on your disk, for example:

Linux

/data/dsx/

Windows

C:\data\dsx\

3. Open source/apps/dsx.kit and set the auto_load_usd path to point to your extracted USD scene file. For example, if you extracted the content pack to /data/dsx/ (Linux) or C:\data\dsx\ (Windows):

[settings.app]
auto_load_usd = "<your_extract_path>/DSX_BP/Assembly/DSX_Main_BP.usda"

Replace <your_extract_path> with the location you chose in step 2.

Note

The auto_load_usd value is empty by default. If you skip this step, the application will start but no scene will load.

Step 3: Verify Environment Variables

Before starting the application, verify that your API key is set correctly:

Linux

echo $NVIDIA_API_KEY

# Should display your API key (starting with "nvapi-")
# If it shows empty, return to :ref:`Prerequisites <prerequisites>` to set it up

Windows (PowerShell)

$env:NVIDIA_API_KEY

# Should display your API key (starting with "nvapi-")
# If it shows empty, return to :ref:`Prerequisites <prerequisites>` to set it up

Step 4: Start the Kit Application

1. In the first terminal, run the streaming script in the dsx directory.

Linux

./run_streaming.sh

Windows

.\run_streaming.bat

The script builds the application (if needed) and starts the streaming server.

Note

Initial startup takes 5-8 minutes for shader compilation. Subsequent launches are faster.

2. Wait for the message indicating the streaming server is ready.

Verify the streaming server started successfully:

Look for these messages in the terminal output:

• RTX Ready

• No error messages about GPU or display issues

If you see errors:

• GPU errors: Verify GPU driver version with nvidia-smi (must be 537.58+)

• Build failures: See Build and Compilation Issues in Troubleshooting

Step 5: Start the Web Frontend

1. Open a second terminal window.

2. Ensure you are in the dsx directory, and run the web development script.

Linux

./run_web.sh

Windows

.\run_web.bat

The script installs dependencies (if needed) and starts the development server.

3. Open the streaming URL in your browser. The URL format is http://your-ip-address:8081/streaming.html, where your-ip-address is your local machine’s IP address (shown in the terminal output).

Terminal output displaying available URLs for local and network access

The web interface connects to the Kit application and displays the 3D scene.

Expected browser behavior:

• Page loads without errors

• Interactive globe visible with site markers (Sweden, Virginia, New Mexico)

• Left toolbar visible with navigation icons

• Configurator panel appears on the right side

• No console errors related to WebRTC or connection failures

If the browser shows connection issues:

• Verify the streaming server is still running (check first terminal)

• Check that you’re using the correct IP address from terminal output

Optional: Configuration Options

You can configure the streaming connection using URL query parameters or environment variables:

Parameter	Environment Variable	Default	Description
server	VITE_OMNIVERSE_SERVER	localhost	Kit server address
signalingPort	VITE_SIGNALING_PORT	49100	WebRTC signaling port
width	—	1920	Stream width in pixels
height	—	1080	Stream height in pixels
fps	—	60	Target frame rate

Example with custom settings:

http://localhost:8080?server=192.168.1.100&width=1280&height=720

For build issues, see Troubleshooting.