Create your first Cortana skill

Once you have set up your environment, you can dive in and build your first Cortana skill. As you learned in the previous module, you can develop a Cortana skill on either a Mac or a PC running Windows 10 Anniversary Update (build 1607) or later, using your choice of development tools. For more information on setting up your development environment, see Set up your development environment.

This module shows the basics of creating a Cortana skill using the Azure Portal and Azure Bot Service. First, you'll create a bot, and then register it and connect it to the Cortana channel to make it a Cortana skill.

For more information on creating your first skill, see Create your first skill.

Step 1 - Create a bot using the Azure Bot Service

To create a bot using the Azure Bot Service, first open the Microsoft Azure Portal and log in with your Microsoft account.

  1. Click on Create a resource.

  2. In the Azure Marketplace pane, click AI + Machine Learning.

  3. Click on Web App Bot in the Featured column.

Azure Marketplace

Enter a name for your bot. The bot service uses the bot name as the default name for the resource group name and app name.

Specify your subscription plan, location, app name, and the bot template you want to use, and then click Create. To create a simple guided conversation bot that pretends to book a trip for you, use the Basic template.

Create Pane

Note

The process described automatically create an Azure App Registration, providing an App ID and password. There are times you want to create these manually (or have an IT administrator create them). In these cases, please review the App Registration documentation. Cortana requires that access tokens be version 2, and that the skill (app) be multi-tenant. You can verify the manifest under App Registrations, and check "accessTokenAcceptedVersion": 2 and
"signInAudience": "AzureADandPersonalMicrosoftAccount" are set.

It may take a while to provision and initialize your bot. When the process is complete, you'll be taken to a screen that lists your resource. From here, you can test your bot. If it's not listed under Recent resources, click on See all your resources.

Resource pane

In the All resources pane, you should see your skill name listed. There will probably be more than one entry on the list. Click on the one whose type is Web App Bot.

Response pane

This will bring up a summary screen showing the options for your bot. The screenshot below highlights the two testing options, Test in Web Chat and Test your bot online in Web Chat. You can click on either one. They both take you to the same test screen.

Bot summary screen

In the test screen, try typing something over the Type your message... prompt at the bottom of the screen. It doesn't matter what you enter, you'll see the screen shown below.

Test Bot

Again, when you see the What can I help you with today? query, it doesn't matter what your response is. The test will always ask Where would you like to travel to?. This is the start of a guided conversation about booking a trip. The bot isn't actually doing anything on the backend. You can book any trip you like, as this sample shows.

Test trip booking

Step 2 - Explore the code

The dxtBasic bot uses .NET C# code that is part of the Basic bot template. To view the code behind the bot:

  1. Click Build under Bot management.

  2. For purposes of this example, click on the Open online code editor link to open the App Service Editor.

  3. If you'd prefer to work in a local editor, you can click on Download Bot source code in the Build section, but this example won't cover that.

Find the code

When the online editor opens, it will display the readme file (readme.md).

Online editor readme example

As you can see by looking at the sidebar, there's C# code in the Bots, Controllers, and Dialogs sections. For this example, the bot's confirmation output line has been changed in MainDialog.cs (highlighted below). For the changes to take effect, you need to redeploy the bot. Click the Open Console icon to open the console window.

Changed line

After the console opens, enter build.cmd.

Console Window

You'll see a lot of status lines printed to the console. Once the process is complete, you can return to the Test pane to test your bot. In this case, after going through the guided conversation again, you'll see a different confirmation line.

New booking confirmation message

Step 3 - Connect your Azure Bot to the Cortana Channel

Once you have created a bot, connect it to the Cortana channel to make it a Cortana skill.

On the bot's Overview page, click on the Connect your bot to channels. link.

Bot overview page

On the Configure Cortana page, scroll down to the Discovery and Management section and click Manage. Click the Publish menu tab and then Publish to self to open the channel configuration settings, where you can specify an optional icon for the skill, a display name, and the invocation name. Cortana uses the invocation name you specify to invoke the skill.

Configure Cortana Skill

On the Channels tab of the Azure Portal, you can now see Cortana listed as an mva-hello-world bot service channel.

Bot Channel

To return to the Configure Cortana page, where you can update the skill's configuration settings, click Edit. To test your new skill, direct Cortana to invoke the skill using the invocation name you specified. For example, "Cortana, tell Hello World 'hi there'."