Python Content Understanding Library

[docelmo]

I am working to classify documents in python. When I try calling the classify function I am getting the following

❌ Error classifying document: 404 Resource Not Found for url: https://xxx.services.ai.azure.com/contentunderstanding/classifiers/xxx:classify?api-version=2025-11-01 Error Code: 404 Error Message: Resource not found

I am looking at the API and do not see an example to classify. From what I was able to find it appears that the analyzers now do classification and extraction. Is this true?

[chienyuanchang]

Hi @docelmo, thank you for bringing up this question about the classify function. Yes, there are some changes in the latest GA with API version 2025-11-01. Classification is now integrated with the analyzer API, now supporting classification with the contentCategories property. Now the number of supported categories expands from 50 to 200, enabling precise classification and routing of diverse file types within a single analyzer—no separate classifier required.

We have a sample notebook of classify function for our latest GA API version: 2025-11-01: classifier.ipynb. Below is the example analyzer:

content_analyzer = {
    "baseAnalyzerId": "prebuilt-document",
    "description": f"Custom classifier for URL classification demo: {analyzer_id}",
    "config": {
        "returnDetails": True,
        "enableSegment": True,
        "contentCategories": {
            "Loan application": {
                "description": "Documents submitted by individuals or businesses to request funding, typically including personal or business details, financial history, loan amount, purpose, and supporting documentation."
            },
            "Invoice": {
                "description": "Billing documents issued by sellers or service providers to request payment for goods or services, detailing items, prices, taxes, totals, and payment terms."
            },
            "Bank_Statement": {
                "description": "Official statements issued by banks that summarize account activity over a period, including deposits, withdrawals, fees, and balances."
            }
        }
    },
    "models": {"completion": "gpt-4.1"},
    "tags": {"demo_type": "url_classification"}
}

To do the classificaiton, you will need to define contentCategories within the analyzer's configuration, specifying up to 200 category names and descriptions that the service will use to categorize your input files. Please set enableSegment parameter which controls how the classifier handles multi-document files to true. It will automatically split and classify different document types within a single file. Compared to preview API (2025-05-01-preview), you will also need to modify the baseAnalyzerId from previous analyzer, deploy LLM models in AI service, and add models configuration as reference.

Please let us know if you encounter any issues or have any questions.

[docelmo]

Thank you for that. I have been playing with the example. The content understanding library in this project references API's that keep telling me that the objects do not exist. I do not know how to overcome this and resolve the API issues with the lack of documentation. the API endpoint code
def _get_classify_url(self, endpoint: str, api_version: str, classifier_id: str) -> str: return f"{endpoint}/contentunderstanding/classifiers/{classifier_id}:classify?api-version={api_version}"
doesnt actually work. It tells me it cannot find the analyzer but when I list I can see them. Also this method in the rest API isnt documented as far as I can tell.

[chienyuanchang]

Hi @docelmo, thank you so much for pointing out this function. These classifier functions (including begin_create_classifier, begin_classify, ...etc.) were created for classifier in preview API like 2025-05-01-preview. We may update to remove these functions from our sample code in the future. If you have created classifier analyzers in preview version, you would need to use 2025-05-01-preview as api_version to use the old classifier. If you are creating a new classifier, I would recommend to use our GA version by following this classifier sample notebook.

[docelmo]

I have doen that. I am using the 2025-11-01 API version. When I try to call the /contentunderstanding/classifiers endpoint I get 404 when trying to call those endpoints. I also looked at the documentation for the REST API and I cannot find the classifiers endpoint documented in it anywhere. Should I be using the analyzer endpoints for classification?

[chienyuanchang]

Hi @docelmo, yes, classification feature is now integrated with the analyzer API in the latest GA API version 2025-11-01 and is using analyzer endpoints. In our sample code, we're using begin_create_analyzer and begin_analyze_binary to create and analyze for classification now. Sorry for the confusion, and thank you for your finding and sharing, I will work on removing codes for deprecated features.

The analyzer schema for classification will be like below in our classifier sample: classifier.ipynb.

content_analyzer = {
    "baseAnalyzerId": "prebuilt-document",
    "description": f"Custom classifier for URL classification demo: {analyzer_id}",
    "config": {
        "returnDetails": True,
        "enableSegment": True,
        "contentCategories": {
            "Loan application": {
                "description": "Documents submitted by individuals or businesses to request funding, typically including personal or business details, financial history, loan amount, purpose, and supporting documentation."
            },
            "Invoice": {
                "description": "Billing documents issued by sellers or service providers to request payment for goods or services, detailing items, prices, taxes, totals, and payment terms."
            },
            "Bank_Statement": {
                "description": "Official statements issued by banks that summarize account activity over a period, including deposits, withdrawals, fees, and balances."
            }
        }
    },
    "models": {"completion": "gpt-4.1"},
    "tags": {"demo_type": "url_classification"}
}

To do the classification, you will need to define contentCategories within the analyzer's configuration, specifying up to 200 category names and descriptions that the service will use to categorize your input files. Please set enableSegment parameter which controls how the classifier handles multi-document files to true. It will automatically split and classify different document types within a single file. You may need to deploy LLM models in AI service and add models configuration as reference if you haven't.