How Voicegain Remotely Manages the Edge Deployments

Voicegain uses standard Kubernetes API to control the configuration of the Edge Deployments. This way all the Voicegain microservices that altogether comprise the speech recognition application can be remotely deployed into your Kubernetes Cluster.

Management is done from the Voicegain Web Console.

Step 1 is creating a new Edge cluster configuration in the Web Console - you have first to switch to the Edge Mode (if you do not see the Edge mode please contact Support.)

There are two types of configurations:

• EZ-Init which is largely automated. It assumes a vanilla Ubuntu 20.04 setup on which the EZ-Init script will install Kubernetes and will then register it with Voicegain Platform.
• User-configured Kubernetes Cluster. A typical case would be a cluster configured in the Cloud, e.g. GCP, using the Kubernetes management tools of the given Cloud platform, e.g., GKE.

To interact with a given Edge deployment you first need to load its configuration using the load button. It will open it in a new browser tab.

First part of the Edge configuration is the Kubernetis admin connection.

If you used EZ-Init script this section will already have been populated with correct data. If you configured the Kubernetes cluster yourself you will need to paste here the Kubernetes config that was created for a service account that has Admin access to the cluster.

Instructions for Google GKE on how to create a suitable kubeconfig.yaml are available here.  Note that you would likely use a service account name different from ci-cd-pipeline used in the example. We suggest using "voicegain-manage".

If the admin connection to the Kubernetes cluster has been established successfully you will see "Connection: SUCCESS". Connection is tested every 5 minutes. You can test the admin connection at any time by pressing the Test button.

In the Configuration Section/Tab you will see the currently deployed configuration on the given Edge Deployment.

Here you can configure the following settings:

• Which version of the Voicegain platform should be deployed
• What specific functional configuration should be deployed. -- In addition to several generic configurations you will see here some custom configurations prepared specifically for you account. Custom configurations deploy only the necessary functionality and and make optimal use of your GPUs.
• What IP or FQDN should be used to access Local Edge Web Console. That may be different from the IP of the Kubernetes cluster if you use routers or load balancers,
• External document database - enter values here if you do not want to use Mongo deployed within Kubernetes.
• External object storage - enter values here if you do not want to use the Minio S3 storage deployed withing Kubernetes. Here are instructions for creating S3 bucket and the Key.

After you change any configuration items you can apply them to the Edge cluster by clicking the (Re)Build cluster button. This will use the configured kubeconfig,yaml to authenticate with your Kubernetes cluster and will apply the chosen target configuration to your Kubernetes which will then pull necessary docker images from public and Voicegain repositories and launch them with the appropriate configuration.

Rebuild may take about 10 minutes to complete. You can check the rebuild status by refreshing the page.

You can check the Licensing status of your server in the last Section/Tab