What was reported:
Issue Azure/azure-sdk-for-python#34370 (closed, 8 comments): - **Package Name**: azure.ai.formrecognizer - **Package Version**: 3.3.2 - **Operating System**: Windows 10 - **Python Version**: 3.10.13 **Describe the bug** When training a custom document classifier, there is a `TrainingContentMissing` error even tough the `container_sas_url` works. The code is based on [this](https://github.com/Azure/azure-sdk-for-python/blob/azure-ai-formrecognizer_3.3.2/sdk/formrecognizer/azure-ai-formrecognizer/samples/v3.2_and_later/sample_build_classifier.py) sample. I found a similar [issue](https://learn.microsoft.com/en-us/answers/questions/1300120/azure-form-recog
What works:
If the custom document classifier build fails with TrainingContentMissing / InvalidContentSourceFormat on the OCR/re layout results, a working workaround is to bypass the parsed response and use the raw HTTP callback instead. Create the DocumentAnalysisClient, then pass a `cls` callback that captures `raw_response.http_response` and returns `response.json()` bytes. You get the full analysis payload without the SDK tripping over the content format.