Error

This model cannot read images

Why an attached image is refused, and how to get a model that can read it.

What it means

A model can only look at a picture if it has a second file beside it, called an image projector. Draggy downloads that file with the model when the repository has one, and starts the model with it. When the engine cannot load the projector, Draggy starts the model without it so text chat keeps working. From then on the model cannot take images, and sending one is refused with this message.

Why the projector may not load

The engine understands a fixed set of projector types. A projector of a type it does not know is refused. A few models ship one that only a modified build of llama.cpp can read, and say so on their own page. Draggy runs the standard llama.cpp release, so those models can be used for text but not for images.

What to do

• Send the message again without the image. The model still works for text.
• Choose a different vision model from the model list in Settings → Models.
• Restarting Draggy, or downloading the model again, makes Draggy try the projector once more. That is worth doing after Draggy has been updated.

Still stuck

Settings → Data → Open log folder has the logs, and app.log is safe to share. Open an issue with the message you saw and that log, or read Troubleshooting.