How come gemini studio can use unsupported mimetype?

Gemini API
, ,
1

I’m creating an app which a user can upload a file (PDF). I’m trying this to Gemini Studio and it works perfect. I get the code in the studio and it accept “application/pdf”.

image
image854×496 48.9 KB

I check the prompting_with_media, there’s no “application/pdf”.

I double check it in the app and it also doesn’t work.

“[GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash:generateContent: [400 Bad Request] Unsupported MIME type: application/pdf”

How come gemini studio can use unsupported mimetype?

2

I’m stuck in exactly the same place.
It looks like the Vertex AI API supports it too, but the Gemini API doesn’t.

3

I’ve come up with a solution to convert the PDF into images and add them to the prompt. Since they have a lot of tokens, you can either use gemini-1.5-flash or gemini-1.5-pro.

4

Same issue here.
As Vertex AI seems to support PDF and Gemini Studio as well I guess it should hopefully soon be available through the API?
Especially considering that the code generated by Studio doesn’t work as of now.

5

What is the solution?

6

This is a huge issue. What is the business use case for a multi-modal model if it can’t read the core business document format: The PDF file.

Can we get some clarification from Google on this?