Prerequisite
Installation of InoWiz Prediction Builder
Active Microsoft Azure subscription. (Required to download and deploy the InoWiz Prediction Builder Azure application.)
Install Preditc4Dynamics App: Microsoft Appsource Link
Get InoWiz Prediction Builder app.
1. Start at the Predict4Dynamics Home Page
Open the Predict4Dynamics application home page in your browser.
From the home page, locate and click the Install button. This will redirect you to the InoWiz Prediction Builder Azure application page.
2. Download the InoWiz Prediction Builder (Azure Managed Application)
Once you click on Install it will navigate you to the InoWiz Prediction Builder page, click Get it now to begin the deployment process.
A confirmation prompt appears — click Get it now again to confirm.
3. Select Subscription and Create
Select your active Azure subscription from the subscription dropdown. (You must have an active subscription to proceed.)
Click Create to continue to the deployment configuration UI.
4. Basics Tab — Resource Group & Names
In the Basics tab of the deployment UI:
Create a new Resource Group OR select an existing Resource Group.
Important: If you select an existing resource group, it must be empty. If it contains resources, create a new resource group instead.
Enter the Container App name and Managed Application name.
5. OpenAI Service Configuration Tab
Move to the Open AI service configuration tab. You will be prompted to choose an option (Yes / No).
If you select Yes:
Assuming you already have existing Azure Open AI service with the deployed model (note the model should be of type chat completion)
After successful deployment of the InoWiz Prediction Builder, open the container apps.
Go to the settings > service connector, click on create in the basics tab > and keep the container the default selected one.
Service type select OpenAI service keeps the connection name as is don’t make any changes there
In the OpenAI dropdown, all your existing Azure OpenAI services will be listed. If you have an existing service, it will appear in that drop-down, so select it from there.
After that select the client type as python and then click next
In the Authentication tab, keep everything as it is, and then click Next.
In the Networking tabs also, keep the default settings and click on Review + Create
Navigate to the Review + Create tab, check all the filled details and settings, and click on the create button. This will deploy the existing Azure OpenAI service to the container app.
After successful deployment, the Azure OpenAI service is now shown in the container apps settings > service connector. Now, with the help of this connection, the container app can connect to the Azure OpenAI service.
If you select No:
The InoWiz Prediction Builder will create an Azure AI service for you and deploy the model into that service automatically.
From the dropdown, select the model you want to deploy and select/set the token for that model.
Set the token value based on your requirements and the tokens available in your Azure subscription. (Ensure you have sufficient tokens/quota for the chosen model.)
Click Next to proceed.
6. Tags / Environment / Owner
In the subsequent tab(s): Leave the default Tags settings as they appear (no change needed per documentation). Confirm Environment Type is set to Production (default).
In the Owner field, enter the name of the owner for the application being created. This identifies the responsible person or team for this deployed instance.
7. Finalize Deployment
Review all configuration selections (subscription, resource group, container app name, OpenAI configuration, model and token, environment, owner).
Complete the deployment (click Create / Deploy as applicable in the Azure UI).
This will going to create a managed application in the selected Azure subscription. After successful deployment user must open the managed application.
Open the Managed Resource Group
Inside the managed application, go to the Managed Resource Group. This resource group contains all the resources that were created by the managed application.
On the managed resource group overview you can:
View all created resources directly, or
Click Resource Visualizer to get a visual map of the created resources.
Inside the Container App, copy the Application URL — this is the endpoint URL that you will use on the Predict4Dynamics home page to connect the app to the container. Save this URL for later.
9. Container App: Cool-Down Period (Optional Adjustment)
(Optional) In the Container App settings, you can change the cool-down period. By default, it’s 900 seconds.
The cool-down period is the idle timeout after which the container app becomes idle to save costs; a request sent after this idle period may require a few seconds for the app to spin up. Adjust this value based on usage and cost tradeoffs.
10. Add Dataverse (D365) Credentials to Container App Secrets
In the Container App navigate to Security → Secrets. You must add values for the following three secret keys:
d365-client-id
d365-client-secret
d365-tenant-id
These values connect the container app to your Dataverse (Dynamics 365) environment.
11. Create an App Registration in Microsoft Entra ID (to obtain the three keys)
Open the Azure portal for the Microsoft Entra ID (Azure AD) tenant that hosts the Dataverse environment (this may be a different tenant than where the container app is deployed).
Navigate to Manage → App registrations and click New registration.
Enter the application name and select the first (default) option for supported account types, then click Register.
Copy the Directory (tenant) ID → this becomes d365-tenant-id.
After the app registration is created:
Open the application registration and copy the Application (client) ID → this becomes d365-client-id and Copy the Directory (tenant) ID → this becomes d365-tenant-id.
12. Grant API Permissions for Dynamics CRM
In the app registration, go to Manage → API permissions → Add a permission. Select Dynamics CRM from the API list.
Choose Delegated permissions and then check user_impersonation, then click Add permissions.
Click Grant admin consent to apply the permission for the tenant. This ensures the app can call Dynamics CRM APIs as the signed-in user.
13. Create a Client Secret (d365-client-secret)
In the app registration, go to Manage → Certificates & secrets → New client secret. Create a new secret and optionally set its expiry. Click Add.
Immediately copy the secret value — this value is shown only once. This copied value is the d365-client-secret you will add to the Container App secrets. If you leave the page, you cannot retrieve this value later.
14. Fill Container App Secrets with the Three Keys
Return to the Container App → Security → Secrets. For each secret key:
Click the secret (e.g., d365-client-id) → click the Edit icon.
Paste the corresponding value (client ID/tenant ID/client secret) → check the confirmation box → Save. Repeat for d365-client-secret and d365-tenant-id.
15. Copy the Container App API Key
In the Container App secrets list, copy the api-key value. This API key is the secured authentication key used by Predict4Dynamics to talk to the container app. Save this key
16. Validate Endpoint & API Key in Predict4Dynamics App
Open the Predict4Dynamics home page. Enter the Endpoint URL (container app URL) and the API Key you copied, then click Validate.
On successful validation you will receive a success notification on the form — then click Save to persist the connection.
17. Troubleshooting (Common Issues)
Insufficient subscription / missing permissions: Confirm the Azure subscription is active and you have permissions to create resources (Owner or Contributor role).
Model/token selection errors: Check that the selected model and token are available in your Azure tenant; if not, choose a supported model/token or provision required resources.
Managed Resource Group: Use the resource visualizer to confirm all resources were created correctly.
Cool-Down Period: Default 900s — adjust if frequent cold starts are a problem.
App Registration Secret Visibility: The client secret value is available only once at creation. Copy and securely store it immediately.
Tenant Differences: The Azure AD tenant that hosts Dataverse may differ from the tenant where the container app is deployed — ensure you register the app in the correct tenant.
Permission Scope: Ensure user_impersonation permission for Dynamics CRM is added and admin consent granted; otherwise, API calls will fail.
Validation Failures: If Predict4Dynamics validation fails during the endpoint/API key step, check the container app running state, the secrets accuracy, and that the container app endpoint is reachable from your network.
For any queries, reach out to us at [email protected]
Last updated
Was this helpful?