Document Intelligence Studio | deploy / import models with CI/CD

Richard Jackson 71

I am looking to deploy a Document Intelligence Studio resource and I wish to be able to import a pre-trained Custom Extraction Model into the newly deployed resource as part of an automated CI/CD pipeline.

How would I do this?

Thanks!

kothapally Snigdha 1,620 Reputation points Microsoft Vendor

2025-02-13T23:06:17.2566667+00:00
Hi Richard Jackson,

To deploy a Document Intelligence Studio resource and import a pre-trained Custom Extraction Model as part of an automated CI/CD pipeline, you can follow these steps:

Ensure you have an Azure subscription and create a Document Intelligence instance in the Azure portal.

Import the Custom Extraction Model:

Navigate to the Document Intelligence Studio.

Select the Custom extraction models tile under the custom models' section.

Click on the Import button on the custom extraction model's page.

In the import project dialog, paste the project token of the pre-trained model that you wish to import and select import.

I hope these helps you. Thank you!
Richard Jackson 71 Reputation points

2025-02-14T09:03:23.4833333+00:00

[accidental duplicate - comment to be deleted]
Richard Jackson 71 Reputation points

2025-02-14T09:35:34.7833333+00:00
Hi @kothapally Snigdha Thanks for your response. Those are the two steps required, and you've described the manual processes. I am looking to automate them as part of a CI/CD pipeline.

Ensure you have an Azure subscription and create a Document Intelligence instance in the Azure portal.

I have used a Bicep file within an Azure DevOps pipeline to do this.

Import the Custom Extraction Model

This the focus of my question - how would one automate these steps? Even if I could provide the Project Token of the Custom Extraction Model as an environment variable, how could I automate the "Import" step?

Thanks!

Rich

kothapally Snigdha 1,620 Microsoft Vendor

Hi Richard Jackson,

Sorry for Delay

To automate the import of your pre-trained Custom Extraction Model into Document Intelligence Studio as part of your CI/CD pipeline, here's are some steps.

You've already deployed the Document Intelligence Studio resource using your Bicep file, which is great!
For automating the import of the pre-trained model, you'll need to use Azure CLI and the Form Recognizer REST API in your pipeline. Here's a simple breakdown:
First, store the Project Token of the model as an environment variable in your pipeline.
Then, use Azure CLI to authenticate and get an access token.
Finally, make a POST request using curl to import the model into your Document Intelligence Studio resource.

I’ve put together a quick pipeline script example to help you automate this:

jobs:
- job: DeployAndImportModel
  pool:
    vmImage: 'ubuntu-latest'

  steps:
  - task: AzureCLI@2
    inputs:
      azureSubscription: '<AzureSubscriptionName>'  # Your Azure subscription name
      scriptType: 'bash'
      scriptLocation: 'inlineScript'
      inlineScript: |
        # Get Azure Access Token
        ACCESS_TOKEN=$(az account get-access-token --query accessToken --output tsv)

        # Set your resource name and the project token (from environment variable)
        RESOURCE_NAME="<YourResourceName>"  # Your Document Intelligence resource name
        PROJECT_TOKEN="$PRETRAINED_PROJECT_TOKEN"  # The pre-trained model's token

        # Make the API call to import the model
        curl -X POST "https://$RESOURCE_NAME.cognitiveservices.azure.com/formrecognizer/v2.0/custom/models/import" \
            -H "Authorization: Bearer $ACCESS_TOKEN" \
            -H "Content-Type: application/json" \
            -d "{
                    \"projectToken\": \"$PROJECT_TOKEN\"
                }"

This script will authenticate to Azure, grab the access token, and then automatically import the model into your resource using the Project Token you've stored in your pipeline. I hope these helps you. Thank you!

Richard Jackson 71 Reputation points

2025-02-19T14:35:30.63+00:00

Hi @kothapally Snigdha

That looks perfect, I'll check this out and report back. Thank you so much for the direction!

Rich
Richard Jackson 71 Reputation points

2025-02-19T14:46:18.9466667+00:00

@kothapally Snigdha
You mention the "https://$RESOURCE_NAME.cognitiveservices.azure.com/formrecognizer/v2.0/custom/models/import" endpoint but I can't find any documentation about this. Would you mind providing a source for this please?
kothapally Snigdha 1,620 Reputation points Microsoft Vendor

2025-02-21T17:54:43.8066667+00:00
Hi Richard Jackson,

This URL is for Microsoft's Form Recognizer service, part of Azure Cognitive Services, and it's used to import custom models.

https: Secure HTTPS protocol.

$RESOURCE_NAME**:** Placeholder for your Azure resource name.

cognitiveservices.azure.com: Domain for Azure Cognitive Services.

form recognizer/v2.0/custom/models/import: Form Recognizer service, version 2.0, custom models, and the import function.

Just replace $RESOURCE_NAME with your actual resource name to access the endpoint for importing custom models into the Form Recognizer service.

I hope these helps you. Thank you!
kothapally Snigdha 1,620 Reputation points Microsoft Vendor

2025-02-24T17:04:42.36+00:00

Hi Richard Jackson,

Following up to see the response was helpful.
Richard Jackson 71 Reputation points

2025-02-25T23:41:39.51+00:00

Hi @kothapally Snigdha

Sorry for the delay, I'll be trying this out next week - however, crucially, you have yet to link me with any documentation to back-up your claims. Without sources, you could have just copied/pasted from generative AI. Please could you link me to where in the documentation that URL is mentioned?

Thank you

Share via

Document Intelligence Studio | deploy / import models with CI/CD

Your answer