UpSlide does not receive, transfer or process any data for the AI services: it is completely managed within your infrastructure. This article is a step-by-step guide to deploying UpSlide's AI features on your Azure tenant.
Introduction to UpSlide's AI features
An overview of the AI-powered UpSlide features can be found on our dedicated page.
A complete overview of how communication works within your Azure tenant can be found here.
Azure requirements
This feature requires to be set up on Azure. If your company doesn’t use it, the AI features will not be available.
UpSlide requirements
- Using UpSlide version 6.9.18 and above.
- Auto-update is turned ON, and users are using the latest version.
Set up the AI features
Components
To set up the Azure OpenAI service, you need to create the necessary components, existing or new, on the Azure tenant:
- A resource group to manage multiple Azure resources collectively.
- OpenAI cognitive service to manage AI models.
- OpenAI cognitive service deployment, the AI model itself. Eligible models are GPT-4, GPT-4o and GPT-4o mini
-
An UpSlide app registration, to register UpSlide as an application on your tenant
- App permission (delegated: user_impersonation) to allow the application to use AI resources in the user's name.
- Public client authentication with redirect URI, to allow a user to authenticate from within the application.
- Service principal to provide the application with an identity on Entra ID.
- A role assignment called Cognitive Services User, the role that will be assigned to users of an Azure OpenAI service to allow them to use the AI models.
Create a resource group
On the Azure Portal, navigate to Resource groups and click Create.
To complete the form, you will need to pick a subscription. You must be at least a contributor to the subscription to be able to assign it to a resource group.
The values you pick during this step will not influence how UpSlide's AI features work.
Create an Azure OpenAI service
In the Azure Portal, navigate to Azure OpenAI to create a new Azure OpenAI service.
- In Resource group select the group created in the previous step.
- The Region field refers to where the service will be hosted. Different models will be available depending on the region selected. It does not have to match the resource group’s region. We recommend using Sweden Central and East US for the widest range of available models. Please refer to this page for model availability by region.
Fill out the rest of the form by using the configurations that satisfy your security policies.
Create a deployment for the service
On the Azure OpenAI page, click on the service you just created and Go to Azure AI Foundry portal.
From the Deployments tab, deploy a new base model. UpSlide is compatible with GPT models. If you cannot find the model you want, then you might want to check if that model is available in your Azure OpenAI service's region. If not, you should start over at step 2 and choose another region.
We recommend using Global Standard as the Deployment type as it provides the highest rate limits. The actual capacity is determined by your subscription.
Edit your deployment to ensure the Tokens per Minute Rate Limit slider is at the max. Raising the limit is necessary to avoid throttling when using our AI features, but this is not a big risk in terms of safeguarding consumption.
Once the deployment is created, copy the Endpoint's Target URI and save it for later.
Setup an app registration
In the Azure Portal, navigate to App registrations and click New registration.
Set a name and click Register. Once the app registration is created, copy the Application (client) ID and the Directory (tenant) ID and save them for later.
Then, navigate to Manage > Authentication and click Add a platform. Select Mobile and desktop applications.
Add the following Custom redirect URIs and replace {clientID} with the Application (client) ID you copied during the app registration step:
ms-appx-web://Microsoft.AAD.BrokerPlugin/{clientID}
Create a role assignment
Still under the Manage menu, click API permissions, then Add a permission. In the Request API permissions menu, select the APIs my organization uses tab and search for Microsoft Cognitive Services.
There, select and add the user_impersonation permission.
Grant permissions to the App & Users to the OpenAI Resource
Now that the EntraID Application have the correct API Permissions, we need to grant the permission to the app on your deployed Azure OpenAI Service
In the Azure Portal, navigate to Azure OpenAI and select the OpenAI Resource previously created, then go to Access Control (IAM) -> Add -> Add Role Assignment
Select the Cognitive Services User role, then go to the Members Tab
Then go to Select members and search for the EntraID App created previously:
Click on Review + Assign, the rights are now set:
How to configure the link in our Portal
How to retrieve the endpoint URL
- please go to ai.azure.com/resource/deployments
- select the deployment
- The "Target URI" is what needs to be input in the portal. It should look like this : https://[yourOpenAIServicesName].openai.azure.com/openai/deployments/[yourDeploymentName]/chat/completions?api-version=[yourAPIversion]
Go to our Portal, and click on AI Settings under the Configuration pane on the left-hand side
Input your Tenant name, App ID and Endpoint URL
Here, you can also configure which AI features you would like to be available and not available to your users.
Provide the AI Glossary option to UpSlide users
When the AI Glossary option is available on your account, you will see it on your dashboard in the UpSlide portal.
You can manually assign the option to specific users in the Users tab or assign the option to a specific user group in the Options tab.