GPTs halding of user uploaded file

I have an API server that converts certain file types from one format to another.
I’m working on creating a GPT that can allow users to upload files and have them sent to my server for conversion.

The server API is very well defined and described to great extent in the openapi.json file uploaded to the GPT during configuration.

When users upload a file to the GPT, it indeed makes an attempt to upload the image for conversion to my API. however, no matter what, it doesn’t actually include the file in the POST request.

How can I make it send the file in the POST request?
How can I learn more about how GPT constructs those API requests other than the fact that it has access to the openapi.json file?


Hi and welcome to the community!

I did take a look around the forum and the existing plug-ins and it appears there is no easy way to handle this via Custom GPTs.

The chat with pdf plug-ins all request a URL where the file is stored and from there it will be uploaded to the server of the plug-in provider.

In some cases a file could be converted to base 64 and send in chunks to the server just to resemble them there.

If your target audience are everyday users then maybe you can create a action that will assign some webspace on the fly to the user plus a link where to upload the file with the most basic user interface ever. Then upon returning to the conversation the user confirms the file upload and the conversation can begin.

1 Like

Thank you @vb. This really helps!

1 Like

I have a similar question. There should be APIs like Adobe PDF tools. I would also like to build a custom GPT action but can’t seem to figure it out. I feel it is generally a relatively simple request, but I am unsure if custom bots support it or not or how to implement it properly.