For the past couple months, I’v been getting ChatGPT to help me analyze a daily journal that I keep in an Excel spreadsheet. Each day is on a new tab. At the end of each day, I upload the file to ChatGPT, it finds the correct tab (named after the date) and we go through the entries.
Lately, however, ChatGPT has been having issues reading the file and I’ve been having to resort to taking screen shots. It started at the end of August when I asked it to do a summary of August and it told me it couldn’t access Aug 5 to 8:
I searched the available prior-file context, but the August journal workbook itself isn't available there in a form I can reliably read. Given all the date mistakes we've had, I don't want to reconstruct Aug 5–8 from memory or infer them from later discussions. Please re-upload the workbook, and I'll specifically inspect the Aug 5, 6, 7, and 8 worksheets.
So I tried re-uploading.
I see the re-upload notification, but unfortunately the workbook still isn't actually present in the accessible file system, so I can't read the Aug 5–8 worksheets yet.
So then I tried creating a whole new Excel file. I copied the relevant sheets to the new file, named the tabs appropriately (Aug 5 to Aug 8), and uploaded.
That would be perfect, but unfortunately the new file still isn't mounted in the Python environment. I tried to open the specifically provided upload path, and it isn't there.
By "Python environment," I meant an internal workspace I sometimes use to open and process uploaded files. That's confusing and unnecessary terminology on my part.
More importantly, I shouldn't have told you that the file "isn't mounted" as though you needed to do something differently. The system is telling me that you've uploaded:
august 5 - 8.xlsx
but the particular file-reading mechanism I was trying to use wasn't able to access its contents.
So the problem isn't that you uploaded it incorrectly. It's an access problem on my side.
I tried again using the specific sandbox path the system provided, and it still isn't actually available to the file-processing environment.
So unfortunately, I still can't read this particular upload. The important difference this time is that I verified the path rather than assuming the file was unavailable.
So I’ve been resorting to screen shots. To this day, it still has issues with file uploads and I still have to use screen shots.
Is this a known bug? Is anyone at OpenAI on top of it? Is there something I can do on my end?
Thank you.