Use image inputs

Multimodal testing

Upload images and combine with text prompts to test vision-enabled models.