Deploying & Using Clusters

Deploying an AI model in a Cluster

Navigator's Canvas is ideal for designing and testing AI Flows, but these flows only run when Navigator is open on your machine. For production-ready, persistent deployments that continue running regardless of Navigator's status, you'll need to use Clusters.

Clusters enable you to:

  • Run AI Flows continuously without keeping Navigator open
  • Distribute processing across multiple computers for better performance
  • Connect to and manage remote devices from a central interface
  • Scale your AI solutions from development to production

Prerequisites
Before deploying with Clusters, ensure you have:
• Navigator installed
• A completed Flow created in Canvas
• One or more computers running the webAI Runtime Agent
• Input devices (e.g., cameras), if needed by your Flow

Understanding the Architecture

Key Components

  • Navigator: The user interface where you create, manage, and deploy Flows
  • webAI Runtime Agent: The background service that executes your AI Flows
  • Cluster: A coordinated group of computers and devices that run your AI Flows

Key Concepts

  • Cluster Controller: Manages the distributed processing environment and allows for persistent deployments even when Navigator is closed
  • Nodes: Computers running the webAI Runtime Agent
  • Input Devices: Hardware components (like cameras) that provide data to your Flow
  • Deployment: The unit of execution in a Cluster - similar to "running" a Flow in Canvas

Deployment Process

  1. Design Your Flow

    • Create your AI Flow in the Canvas environment
    • Test thoroughly to ensure it works as expected
    • Navigator will automatically assign local devices during testing

  2. Save a Version for Deployment

    1. Click the three-dot context menu on your project
    2. Select "Save for Deployment"
    3. Enter a meaningful name and optional description

    To manage saved versions:

    1. Click the context menu → "View Canvas Versions"
    2. From here, you can rename or delete versions as needed

  3. Create or Connect to a Cluster

    To Create a New Cluster:

    1. Click the Network button in the bottom left corner
    2. Navigate to the Clusters tab
    3. Click Add Cluster → Host a new Cluster
    4. Follow the prompts to set up your local Runtime Agent as a Cluster controller

    To Connect to an Existing Cluster:

    1. Navigate to the Clusters tab
    2. Navigator automatically detects Clusters on the same network
    3. Select the desired Cluster from the list
    4. Alternatively, connect directly via IP address if the Cluster is on a different network

  4. Set Up Your Devices

    Adding Compute Nodes:

    1. Go to your Cluster and select the Devices tab
    2. Click Add Node
    3. Follow the prompts to connect a computer running the webAI Runtime Agent

    Adding Input Devices:

    1. From the Devices tab, click Add Input Device
    2. Select the type of device (e.g., camera, microphone)
    3. Configure the connection settings

    All provisioned devices become available to any user connected to the Cluster.

  5. Create a Deployment

    1. Go to your desired Cluster and click Create Deployment
    2. Enter a deployment name and optional description
    3. Select the Canvas Version you want to deploy

    4. Assign resources to each Element in your Flow:

      • Map cameras and other input devices to provisioned Input Devices
      • Assign AI Elements (e.g., Object Detector) to provisioned Compute Nodes
    5. Review your configuration
    6. Click Deploy

Your Flow is now running as a persistent deployment in the Cluster and will continue operating even when Navigator is closed.

Managing Deployments

Monitoring Deployments

  • Navigate to your Cluster
  • Select the Deployments tab
  • View the status and performance metrics of all active deployments

Stopping or Restarting Deployments

  • From the Deployments tab, locate your deployment
  • Use the control buttons to stop, restart, or delete the deployment
  • You can also edit deployment settings by selecting the deployment and clicking Edit

Best Practices

  • Resource Allocation: Distribute compute-intensive elements across multiple nodes for better performance
  • Redundancy: For critical applications, consider deploying to multiple nodes for failover capability
  • Monitoring: Regularly check deployment status and performance metrics
  • Updates: When updating a Flow, save a new version and create a new deployment rather than modifying existing ones

Troubleshooting

If your deployment isn't working as expected:

  • Check that all nodes are online and connected to the Cluster
  • Verify that input devices are properly configured and accessible
  • Review the logs for any error messages
  • Ensure the Flow runs correctly in Canvas before deployment
  • Confirm that nodes have sufficient resources (RAM, CPU) for their assigned elements

By following these steps, you can successfully deploy your AI Flows to production using Clusters, enabling persistent operation without requiring Navigator to remain open.