Question

A retail team wants a lightweight vision app built in Foundry that accepts two product shelf photos in the same prompt and returns which shelf appears better stocked. They ask the developer how to pass both images to the deployed multimodal model. What is the correct approach?

Accepted Answer

Include both images as separate image content items within a single user message alongside the text instruction.

Answer

Send two separate chat completion requests, one per image, and merge the responses in the app code.

Answer

Concatenate the two images into one file first, because a multimodal model can only accept a single image per request.

Answer

Place the images in the system message so the model treats them as fixed instructions.