Images

Every read returns image_urls. With include_images, Frankensurf also downloads them and checks each one.

• Each image gets a SHA256 hash, its real format and size, and its own error if it failed. A file extension is not trusted.
• One bad image never fails the read.
• Download exact URLs on their own with download_images.

Result cards

On a search or listing page, card_images pairs each result with its own thumbnail in the same load, so checking results by eye or with a vision model doesn’t cost a page load per item. Each card is {url, title, image}.

• Order: entries from the page’s JSON-LD ItemList come first, then linked cards that have an image, then the rest. Menu links never crowd out results.
• Which image: a card’s image is the one inside its link, or inside the nearest container that holds no other result. Neighbouring cards never share one, and logos and icons are skipped.
• Lazy images: lazy-loaded images are found through data-src and srcset, and browsers scroll first so they load.
• Unlinked cards: some sites render cards without links and list the URLs only in JSON-LD. Those cards come back with their URL but no image, because matching them up is site knowledge.

page = await web.read("https://shop.example.com/search?q=lamp",
                      policy_overrides={"card_images": True})
for card in page["cards"]:
    print(card["title"], card["image"])

Usage

Python

page = await web.read(url, policy_overrides={"include_images": True, "max_images": 10})

images = await web.download_images(["https://example.com/a.jpg", "https://example.com/b.png"])

CLI

frankensurf read https://example.com/p/2231 --images --max-images 10
frankensurf images https://example.com/a.jpg https://example.com/b.png

MCP

{ "tool": "images", "arguments": { "urls": ["https://example.com/a.jpg"], "max_images": 10 } }

Options

Option	Default	Description
include_images	false	Download the page’s images.
max_images	50	Most images to download. It caps the count; it doesn’t mean the gallery is complete.
card_images	false	Add cards: each result link with its title and its own thumbnail URL.
scroll_screens	0	Screens to scroll in a browser before capture so lazy images load. card_images alone scrolls 3.
max_image_bytes	20 MB	Largest image accepted.