Hello all, this was supposed to be resolved, "unable to read/access the files in "configure tab"

Hello all, getting tired of this message “I’m unable to access/read the documents” it is becoming increasingly frustrating when ChatGPT, is unable to use the documents…is this a limitation or what?

Context (im using basic stuff):

  • The files are PDF’s and Word documents…
  • there is text and images on them
  • is for fantasy writing the purpose.

This topic was the top search, and it is marked as RESOLVED, which clearly needs some attention again. GPT suddenly can't read files in its knowledge base - #4 by chieffy99

Making ChatGPT more efficient with documents is all it takes. It’s easiest to read text, so what you should do is reduce the following factors in your document, such as:

  • Many languages available There is frequent switching.
  • The layout is not fixed. Whether it’s aligned to the left, middle, right edge, divided into two columns.
  • Contains text that is not related to the main content, such as speech bubbles, techniques, and tricks that are easily understood by humans.
  • There is another line of text that is dropped at the end.
  • Contains emoji, characters, symbols, pictures
  • Many colors, sizes, fonts, or handwriting
  • Tables, charts, pictures
  • The background has a watermark. If it’s in text, it’s even worse.
  • If you think of an example file, it would be a student summary sheet.
  • If you try importing your document into Googledrive and translating it into text, your document may not be able to fit into text content if it contains images. Or there are a lot of obstacles that you need to edit into easy-to-read text. Luckily, GPTs have to endure this forever. Even though GPTs are smart, the number of objects that get in the way of data retrieval makes it take a long time. or may not be able to pull
  • I currently use .text as knowledge for GPTs, even if I don’t need line breaks. It still works well. Because it doesn’t use the eyes You try to think that if it were you, you would definitely get a headache if you had to come across this picture. It encountered an obstacle that it didn’t know what it was.

Adding to this thread to say this is one of the biggest challenges I’m facing with building solutions using Open AI’s Assistants API. Without a reliable way to ensure that the file is accessed, automations fail at an alarming rate.

This issue is far from resolved.