Model Settings
Configure a Multimodal Model API
Connect any OpenAI-compatible endpoint. Configuration stays on the server, and your API key is never returned to the browser or exposed in page source.
Connection Status
Once saved, both image workflows will use the new configuration immediately.
Compatibility Requirements
The endpoint must support POST /chat/completions and OpenAI's image_url message format. “Test Vision Connection” sends a small built-in image to verify the full image-input path, not just a text request.