Coda AI Point to PDFs

Just a thought that popped up in my head, has anyone used the Coda AI to point to a PDF (think an organizations operations manual) and set up a coda AI column and an inquiry text column so you could press a button enter an inquiry and the AI will generate a response based on the PDF?

I think this could be a great use case for a custom Q&A section of a project doc.

Hi Juan,

I just tested something like this out and it seems to work.

My prompt was simply: Summarize @File

So it looks like Coda AI can read files, which is awesome!

hi @Samuel_Langford ,

I guess it depends on the content of the PDF, I got this friendly message:

Sorry, I cannot summarize a PDF document without further instructions or access to the file. Please provide more information or clarify your request.

I provided more detailed instructions, but still it failed.

Cheers, Christiaan

yeah im getting the same result as you Christiaan,

I’m thinking the summary it generates is directly tied to the name of the file. My original prompt was Summarize @file (the file name is “Employee Policy Manual”) and the result was, “this is a manual for employees”.

Then i asked it to summarize the first few pages of @file and got the result in the screenshot above

Maybe Coda employees could elaborate on this?

Agree this would be great to understand more.
FYI I tested the public Bitcoin paper https://bitcoin.org/bitcoin.pdf and it worked in a table.

When I downloaded this PDF, and put it in a google drive, that is viewable by anyone, Coda AI failed.
I can share the Coda test file if that is useful.

I’d love to be able to read a PDF from a Google drive link and summarize information using Coda AI. Can anyone from the coda team provide info on whether this is possible?

I have plans to create a PDF file reader as a Pack, but unfortunately it has proven imposible without sending the file content to a server and will take more time :frowning:

Got it. I’m glad that someone is working on this challenge. Let me know if you need a beta tester at some point.

Oh, I came to the same conclusion! I’m afraid I beat you to it, spent the whole day getting this up:

In the example above you can do ReadPDFs("https://bitcoin.org/bitcoin.pdf") to get

Awesome! Good to know there will be multiple alternatives :slight_smile: