View a web page or an image and answer a question about it.
Pass a page url (screenshotted in a real browser) OR a direct imageUrl
(viewed as-is), plus a query. A vision model answers in text based only on
what it can see. Costs 1 data credit per successful call. This is the same
unified capability the AI-employee take_website_screenshot / view_image
tools use (app/utils/visual_qa.py).
Authorizations
Company API key - scoped to a specific company. Generate from the Integrations page in your dashboard.
Body
Visual question-answering parameters
Company ID whose data credits are billed for the call
What to answer about what is seen, e.g. "Describe the visual design and layout" or "What text is in this graphic?"
Full http(s) URL of a web PAGE to screenshot and view. Provide exactly one of url or imageUrl.
Full http(s) URL (or data: URI) of a direct IMAGE file to view. Provide exactly one of url or imageUrl.
For a page url: view the entire scrollable page (default true) or just the visible viewport (false). Ignored for imageUrl.

