The images category is still text-to-image search. There is no way to ask which public pages contain or resemble a supplied image.
A public puzzle carrier only had an Imgur URL and no useful metadata. Filename, hash, and text queries found nothing, so tracing its source required manually driving Yandex image search, extracting 80 thumbnails, and ranking them locally with perceptual hashes. That is a natural search capability, but it sits outside askweb's current search(query) contract.
Could askweb add a separate provider capability such as searchByImage({ url, maxResults })? It should not overload the text query API. A normalized result could carry the source page URL, matched image URL, dimensions, provider, and similarity/rank metadata. URL input is enough for a first version; byte uploads can come later.
SerpAPI Google Lens, Yandex, or Bing could implement it independently, with the CLI and agent tools exposing the capability only when the selected provider supports it.
The
imagescategory is still text-to-image search. There is no way to ask which public pages contain or resemble a supplied image.A public puzzle carrier only had an Imgur URL and no useful metadata. Filename, hash, and text queries found nothing, so tracing its source required manually driving Yandex image search, extracting 80 thumbnails, and ranking them locally with perceptual hashes. That is a natural search capability, but it sits outside askweb's current
search(query)contract.Could askweb add a separate provider capability such as
searchByImage({ url, maxResults })? It should not overload the text query API. A normalized result could carry the source page URL, matched image URL, dimensions, provider, and similarity/rank metadata. URL input is enough for a first version; byte uploads can come later.SerpAPI Google Lens, Yandex, or Bing could implement it independently, with the CLI and agent tools exposing the capability only when the selected provider supports it.