PRODUCTION SCENARIO
An insurance app sends a claim photo to a deployed model and asks it to describe the visible damage.
What must the deployment support?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Image input in the same request as the text prompt
Describing a photo needs a vision-enabled model and a request that carries the image alongside the text, as an image_url part. A gpt-image model produces pictures from text and cannot read one you send it. A vector index retrieves stored content and gives the model no ability to see this photo. Content Safety image filtering screens what passes through the deployment; it does not let the model interpret the image.