Custom Model Vision

Give a custom text model image-recognition support through a separate vision-capable model.

Before you begin

If your custom model does not support multimodal input, Breezell 1.2.8 lets it obtain image-recognition support through another model. Configure the vision-capable model first, then assign it to the custom model that needs image understanding.

Open the Models settings

1. In the editor's top-right toolbar, click the settings gear.
2. In the menu, select Breezell Settings.
3. In the settings sidebar, select Models.

If you have not configured the custom model yet, complete the Custom API setup first.

Open Vision Proxy settings

1. Find the custom model that needs image recognition in the model list.
2. Click the small eye icon on that model to open Vision Proxy.

Choose an image-recognition model

1. In the Vision Proxy dialog, open the Image recognition model selector.
2. Select a configured model that supports image recognition.
3. Click Save to apply the selection.

Start a conversation

Select the custom model in a conversation, attach an image, and send your message. Breezell uses the selected image-recognition model to help the custom model handle the image.