Form Recognizer action returns 'Bad Request' but the input file is not corrupt or unsupported

[_mpa]

Dear reader,

We have been trying to set up a Power Automate flow to use our trained AI model which we have built in Azure Document Intelligence Studio (so not in Power Automate AI Hub using AI Builder). This means that the AI Model is running within our Azure cloud. We use the connector called 'Form Recognizer' and are connected through an endpoint and an API key (which is retrieved from our Cloud environment). The status of the connector is 'Connected'.

The error in the flow, which is the reason for this post, is the following:

The error notification indicates that the file is either corrupted or the format is not supported. The file input is a PDF file, which is not corrupted at all. Trust me, we have tested this in all possible ways. We have checked the microsoft documentation and PDF files are a supported format for the Form Recognizer actions. 

Description of the flow:

The flow triggers when a new e-mail is received in a folder in a mailbox. Then, the attachment is placed on a SharePoint location in the 'Create file' step. Next, the file properties are retrieved to get the URL. The URL functions as the input to the last step, which is 'Analyze Document for Prebuilt or Custom models'. However, this last step generates an error. We have been trying to figure out what is causing the error and how to fix it, but still no succes. 

Some extra information: the file is succesfully created on the sharepoint location. Also, the properties are succesfully retrieved from this file. 

The Form Recognizer step is configured as follows:

Side information:

The complete https URL to the file is 247 characters, which I assume is within limits of the parameters.

We have tried for 6 hours to fix this but so far no luck. We have tried rebuilding the flow and all its connections, but this does not seem to matter. Please note that my Power Automate is running on a trial license and there are no RPA/AI credits assigned to the environment. So far, I have not found this to be a problem anywhere online. 

Whenever I run the flow, and then copy the inputURL of the Form Recognizer action (the last step) and paste it into a new tab in my browser, I am directed to the pdf file on the sharepoint. To me, this means that the URL is correct. But the flow keeps throwing the error 'Bad Request'. We are totally lost for solutions. 

My request:

Can someone please try and help us fix our flow? Or please give any advice on how to try to fix it. Any suggestion is welcome. 

I will try to reply as soon as possible and help provide information. 

Thanks in advance for your help and response.

Kind regards,

Bram

[Antrod]

Hi @_mpa,

It's possible that you get signature image from the email which could appear as a corrupted file. Try including this condition:

And only extract information in the "yes" case.

If it doesn't solve your problem, I would recommend you to ask the question in the Azure Document Intelligence Q&A page: Azure AI Document Intelligence - Microsoft Q&A

[_mpa]

Hi Antrod,

Thanks for your suggestion. During our first test of the Power Automate flow, we got an error thrown for a .png it had found in the e-mail. Due to that, we have removed all content from the e-mails we have tested our flow with. 

Thanks for your suggestion to post my question on another Q&A Forum as well. I have done so and also received a reply there. Do you have any other suggestions to fix this flow? Many thanks in advance.

[_mpa]

All other suggestions are welcome as well! If you have any idea, please reply to this post

[_mpa · Accepted solution]

For the answer on this question, please refer to the following link:
Unable to fix BadRequest error - Microsoft Q&A

I have posted my question on another forum as well, and provided the solution over there. Hopefully this will help you out.