Azure AI Services

This guide explains connecting to an Azure AI service running on Azure infrastructure from your DevBox.

Architecture Diagram

Azure AI Services Architecture

Azure AI Services is a suite of cloud-based artificial intelligence (AI) services that help developers and organizations create intelligent, cutting-edge, market-ready, and responsible applications with out-of-the-box and pre-built tools, APIs, and models.

This connection between the Azure AI Service Playground and your DevBox would occur by setting up a bastion host that advertises the private routes to your DevZero network so you can access the private service through network tunneling. You must also set up a DNS Private Resolver to access the AI Services's Private Endpoint from your DevBox.

Prerequisites

Before you begin, follow the Connecting to Azure guide to set up the Bastion Host to access your private Azure services.
Following the above criteria, follow the Setting up DNS Private Resolver guide to access the DNS Private Zones.

Existing Azure AI Services

To connect to an Azure AI Playground, ensure it is within the same Resource Group containing the Bastion Host.

Step 1: Creating a Recipe

First, we need to create a recipe for the workspace:

Go to the DevZero Dashboard > Recipes and click on New recipe.
Enter the recipe name and click on Create a recipe.
Now use the below-provided snippet to create a recipe for your workspace:

version: "3"
 
build:
  steps:
    - type: apt-get
      packages:
        [
          "build-essential",
          "curl",
          "git",
          "nano",
          "software-properties-common",
          "ssh",
          "sudo",
          "tar",
          "unza ip",
          "vim",
          "wget",
          "zip",
        ]
    - type: command
      command: |
        sudo curl -sL https://aka.ms/InstallAzureCLIDeb | sudo bash

Click on Save and Build and when the build is successful, move on to the next step.

Step 2: Creating a Workspace

Go to the Devzero Dashboard > Workspaces and click on New workspace.
Enter the workspace name and click on Select from recipe library.
Select the recipe you just created and click on Select.
Click on Launch, and your workspace will be ready shortly.

Step 3: Creating a Private Endpoint

To access the service privately, we need to restrict public access and assign it a private endpoint:

Go to Azure Portal and open your AI Service.
Go to the Networking section and click on Disabled under the Allow access from section.
Go to the Private endpoint connections tab and click on Private Endpoint.
After that, go to Networking > Private endpoint connections and click on Private endpoint.
Choose your Resource group and enter the instance name.
Click Resource and select your desired Target sub-resource for your private endpoint. Remember that if you have more than one sub-resource type, you need to create a separate endpoint for each one.
Click on Virtual Network and select your Virtual Network (VNET), which houses the bastion host and DNS Private Resolver.
Select a compatible subnet, then click on Next, and leave the rest of the settings as default.
Click on Create, and your Private Endpoint will be created.

Step 4: Accessing the AI Services

Now follow the below steps to access the AI Services on your DevBox:

Open Azure AI Studio and click on Create New Deployment.

Azure AI Studio

Choose a base or fine-tuned model and modify the configurations according to your needs.
Click on Deploy to create and deploy the model to your playground.
Select the Deployment you just created and click on View Code.

Azure AI Service Connection String

Copy the API Key and export it to your environmental variables using the following command:

export TOKEN=<api-key>

Below that, you will find code snippets for several languages and frameworks. For this guide, we will demonstrate the secure connection with curl.
Copy the code snippet and place it in your TOKEN environmental variable.
A successful connection will show the same output as below.

Azure AI Service Access

New AI Services

If you need to make a new AI Services and access it through DevBox, then follow the below steps:

Step 1: Creating an AI Services

Firstly, you must set up the DNS Private Resolver by following the Setting up DNS Private Resolver guide.

After the steps mentioned earlier are completed, you may go ahead and create a file share for AI Services by following these steps:

Go to Home > AI Services and click on Create.
In the Basics section, select the Resource group you previously selected for your VNET.
Then enter your Region, Name, and Pricing tier.
In the Networking section click on Disabled and then click on Add Private Endpoint.
Choose your Resource group and enter the instance name.
Click Resource and select your desired Target sub-resource for your private endpoint. Remember that if you have more than one sub-resource type, you need to create a separate endpoint for each one.
Click on Virtual Network and select your Virtual Network (VNET), which houses the bastion host and DNS Private Resolver.
Select a compatible subnet, then click on Next, and leave the rest of the settings as default.
Click on Create, and your Private Endpoint will be created.
After setting up the private endpoint, click on Review + Create and click on Create.

Azure AI Service Creation

Step 2: Creating a Recipe