I had this all working and now all i get are variations of the following error
“It seems there is an issue retrieving data from the uploaded document. Please retry uploading the file or check if it’s in the correct format, and I’ll attempt to process it again.”
I confirmed the file IS in the vector store i am using for the current assistant via the playground.
Same here… It worked fine, but since a while (~25.9) this stopped working and I can’t figure out, what is going wrong besides the fact, that the assistant can’t “read” the content of the PDFs anymore.
When trying to give the assistant a PDF in the playground I get an error saying: “Error streaming run: Message content must be non-empty.”
I previously had everything functioning properly, but now I consistently encounter the following error message:
“It seems there is an issue retrieving data from the uploaded document. Please retry uploading the file or check if it’s in the correct format, and I’ll attempt to process it again.”
I have verified that the file has been correctly uploaded to the vector store.
I have give my all scenario detail on: Why am I unable to pass a file to my assistant message Module? - #8 by Benjamin_from_Make
Hi,
I would suggest to open a ticket at support so that they have a look; there may have been a change that breaks it
Benjamin
There isn’t a technical issue; this appears to be user-related. Instead of selecting the LLM in the Make window, you need to configure that directly on the OpenAI dashboard. This is how my setup looks: I’m able to extract specific information and apply logical conditions with error handling. I used the 4.0 mini model, which should be quite cost-effective at this point.
@TheLjohnson.com Could you kindly share only this specific module? Have you utilized the assistant created in Playground with a specific prompt? Is the code interpreter accurately extracting all the required information? How does the output of the code interpreter differ from that of the file search?
it does not work on my end
To start, make sure you’re adding to the right vectorstore if you have more than one, and that you’re using the correct agent with the right API key. It matters, and it has confused me a few times.
I did try it in the playground a few times before I switched back to using Make. I set the LLM as gpt4omini, and when I went to Make, I left the selection blank and just chose the chat assistant name. I’m not going to post my specific instructions, but they start with:
‘Extract data from the vector-stored document mentioned in the prompt…’
Note: It became clear to me during the rewrite that the instructions treated certain words with specific meanings, like DOCUMENT vs. IMAGE. Also, if I wanted to apply logic, I would use DETERMINE instead of EXTRACT with the document. That tied into my custom instructions and i used it later in the instructions.
I keep my temp low because i have only 1 doc in my vectorstore at a time.
@TheLjohnson.com Thank you for your message. I am currently working on the same task. The file uploads to the vector store correctly, and the appropriate agent is being used. However, when I choose the code interpreter, it functions properly. But when using file search, it presents the same issue: it cannot access the uploaded file.
i came across this video which explains the whole process. https://www.youtube.com/watch?v=YVQLw1XKOPE
Just stumbled on this thread trying to do the same thing.
For anyone struggling to get it to work, there’s a dedicated OpenAI ‘Add files to a vector store’ module now, which makes the process probably easier.
Not sure how long this dedicated module has been there, but it worked seamlessly.
I’m keep having problems with the output data of my PDF. On the end ChatGPT respond is:
“It seems there is an issue accessing the uploaded file. Could you please try uploading the document again?”
This is how my scenario looks like:
I already tried a lot of different PDFs, but its not working. I just need to have the text output from the PDF, so I can prompt ChatGPT to extract some elements.
Please help me
Current flow:
@Beth1401 Do you used in “Upload a file” the Map to choose the file from Google drive or the specific file?
Are you sure that the problem is in this step, because when I automate some file-upload with chatgpt than I had to add the file to a vector space and not only to upload it?
