An Azure service that turns documents into usable data. Previously known as Azure Form Recognizer.
Hello,
First of all, I apologize for the late reply. I missed it.
Thank you for your response and for the suggested troubleshooting steps.
I checked Azure Service Health, but I did not find any active incident or service degradation related to this issue.
The issue started suddenly from one day to the next. The same setup and the same type of invoice documents had been working normally before.
There was only a very minor change on our side: we corrected a typo in the description of the specific field that the model extracts. This was only a description correction and should not have changed the document structure, field definition, or extraction logic.
I would also like to clarify that when using the API, the latest build before this update still works normally with the same documents. This suggests that the issue is unlikely to be caused by file upload, PDF corruption, encryption, or the general document format.
I apologize, but I have not yet had time to test other prebuilt models such as prebuilt-layout or prebuilt-read, and I have not performed cross-region testing yet. However, for our use case we specifically need this custom invoice model, so testing other models or regions would mainly help isolate the issue, but would not be a practical replacement for our production workflow.
Could you advise whether correcting a typo in a field description could impact the model analysis stage, or whether there is a way to restore/continue using the previous working build?
Thank you.
Best regards,
Martin