OpenAI file attachment

[edwardzev]

Your Task History ID

IjU3NjUwNTZmMDYzNjA0MzM1MjZmNTUzMTUxMzE1MTZiNTQzOTBmMzAi_pc

I am trying to build workflow that is triggered by an email parser, receive attachment url (invoice - which is pdf file with open text). Fetch details from invoice using openai, and stores it in airtable or any other spreadsheet.

I built an assistant. tested it in playground and the assistant analyze quite efficiently the uploaded file and returns needed details (invoice num, sum, etc)

Step1 - trigger=>attachment url
Step2 - upload file to assistant=>file id
Step3 - create thread=> thread id
Step4 - create message with attachment - here i have a problem. i can't get message created with file id, that contains pdf file, although it is obviously uploaded since I received file id.
Step5 - create run => status queued. Wait a bit
Step6 - list messages=>I get response that indicate that the assistant didn't really checked the uploaded file.

I am aware that the easier option is to retrieve text with the help of third party pdf parser like pdf.co, but since openai assistant easily retrieve data from the open base pdf, i want to make it available through the direct uploading of the file, and not by feeding the extracted text.

 

[ArshilAhmad]

Hi @edwardzev,

In the 'Create Message' action step, please enable mapping, then map the File ID enclosed within double quotes and square brackets, and see if that fixes your issue.

 

[edwardzev]

thank you for a prompt reply.
but it didn't really help.

 

[ArshilAhmad]

Please allow us some time to look into this concern. We will get back to you with an update soon.

 

[Preeti Paryani]

Hello @edwardzev,

Please try using the Create Message V2 (Beta) action instead of Create Meaasage and let us know if it works for you.

 

[edwardzev]

Thank you.
I tried. It returns "Invalid type for 'attachments[0].file_id': expected a string, but got an array instead."
My guess is that in TYPE field I chose TEXT (not image, not image url obviously) but seems TEXT also wrong. My guess.
Pls check screenshot.

 

[ArshilAhmad]

Could you please click on 'Re-Capture Email Parser Response' and capture a new file in the Email Parser trigger step? Once done, please let us know.

 

[edwardzev]

done

 

[ArshilAhmad]

Thank you for uploading a new file. Please allow us some time to review this concern. We will get back to you with an update shortly.

 

[Preeti Paryani]

Hello @edwardzev ,

It seems there have been changes made to your workflow, and as a result, we are unable to locate the OpenAI steps. We kindly request you to add these steps back so we can proceed with testing the issue.

Thank you for your cooperation!

 

[quantum.je]

Hi @edwardzev & @Preeti Paryani !

I follow exactly the same workflow and have tested it with both PNG and PDF files. However, I encounter an error message at the last step:

"Files were not found."

Could you let me know how I can help in return? I’m happy to conduct additional tests if needed.

Also, has this type of workflow (e.g., document analysis via OpenAI) already been implemented on Pabbly's side?

Thank you in advance for your support!

Access to my workflow : here