Question

A content team uses a deployed multimodal model in the Foundry portal to describe product photos. They now want to test the model's behavior on a single new image quickly, adjusting the prompt wording several times before writing any code. Which approach in Microsoft Foundry lets them do this fastest?

Accepted Answer

Use the model playground in the Foundry portal to upload the image and iterate on the prompt interactively

Answer

Write a lightweight vision client with the Foundry SDK and rerun the script after each prompt edit

Answer

Create a single-agent solution with a file search tool that indexes the image

Answer

Configure an Azure Content Understanding analyzer with an image field schema