Describe the feature
The router currently uses text embedding or classification models for routing. Support of multimodal prompt (text, image, audio) is missing.
Why do you need this feature?
We now have this tri-modal embedding model to start with.
Additional context
No response
