Skip to main content

So I am trying to extract relevant details from research articles about intellectual disability and output information about the studies to a google scheets spreadsheet. Since it’s quite inneffective to automate data extraction by simply prompting, I decided that I needed a workflow to extract several studies at the same time. Also I have noticed that it’s impossiblke to upload files to ChatGPT’s chat.

 

Anyway, here is my workflow:

 

And here is the prompt and the other settings that I’m using. I try to use named variables for the variables that I want to extract from each article:

 

However, I received this message from ChatGPT:

I don’t know how to procceed with this or why ChatGPT cannot access my pdfs. I have also tried with txt files but had no better luck with it. Has anyone tried to do something similar with ChatGPT and Zapier? Why do you think that ChatGPT cannot access the file? Do I need to change the workflow?

Hi @Marcus Allgulin 

Looks like you’ve only provided the Title of the PDF, not the actual file.

 

Help article about using files in Zaps: https://help.zapier.com/hc/en-us/articles/8496288813453

 

Try using a different ChatGPT action that allows you to use files.

 


Hi, I’ve made significant updates to my workflow and added the ChatGPT upload action. I read the guide that you sent to me.

 Now ChatGPT does not seem to accept my request. I have tried to provide the link to the file instead. However, I got the error “Unauthorized”. Proabbly because these files are on my personal drive which requires a login. Anyway, this is the error I have received:

And this is how I have configured the ChatGPT assistant step:

 

 


@Marcus Allgulin 

Try these:

  1. Remove the mapped variable from the field: New Files
  2. Map the File ID from the previous step to the Existing Files field.

 


Hmm, now I have added the file ID but the same error still keeps showing up

I also tried mapping the file ID to New files field but I got the same error. I’m very confused.


@Marcus Allgulin 

 

Make sure the File with the ID being used to test still exists in ChatGPT.

 

Check the supported file types.

 


I checked the file ID for numerous steps. I went all back to the the trigger step. Here are the results. The first ID is from the trigger step when I refreshed my records. The ID looks like it has changed from the upload step to the conversation step so maybe they don’t match up. However, I would assume that more file ID: s would show up if that was the case. I made sure to refresh every step to make sure that the file ID was updated.

I’m a bit confused now

 

Then I checked the upload step:

 

 


@Marcus Allgulin 

You should not have any variable mapped to the field: New Files

 


I also checked the supported file types and it seems like the OpenAI models support pdf files, which is the kind of file I’m trying to give to ChatGPT.

 

A big thank you for your patience so far!


I removed the variable from Existing Files field but I still get the request error

 


@Marcus Allgulin 

Is the GDrive File that is being uploaded a publicly available file?

You may need to change the GDrive Folder Sharing Preferences or add a GDrive step to change the File Sharing Preferences.

 

 


I have made the Gdrive public now (with editing rights) and I still get the invalid request error. However, I can provide links to ChatGPT now. I also tried to provide the link to the file instead. I tried to upload the embedded link and I still got the request error. Same with the alternative link.

 

I logged in on my OpenAI account and can see that it has the file object. My account balance is also full. It seems like ChatGPT has the file object but cannot work with it. The problem seems to lie with the assistant.


@Marcus Allgulin 

Links would need to be the direct download link to the publicly available file.

Tool: https://sites.google.com/site/gdocs2direct/