Image File Processing

[lickylips]

Hi there,
For a few weeks now, I've been having gemini CLI read my obsidian vault (A note app that works with local markdown files). I have been taking photos of my handwritten notes and attaching them to my obsidian note. I would then ask gemini cli to find the images that were linked and transcribe them and append the text to the bottom of the note. This has been working well and it has been able to do things like spell check and identify headders from my handwriting.

In the last two days, Gemini CLI has been saying that it is not able to process image files so can't do this. I've tried rephraising my request and the same behavour has been happening on my install on windows and linux. Gemini web interface can still do this but it is limited to 10 images at a time.

Has there been an intentional change here? Is this just a bug?

Update 2025-10-29
I've updated to the latest version, now when I ask it to do the same thing it is refusing saying that the 4 images it will send will exceed it's context window. Claiming that these 4 images represent 3.8 million tokens. That seems a bit much for just some images. I've tried a few different times in a few different ways and getting the same issue

Thanks
Seán