hearchco / agent

Agent for Hearchco built using Go.
https://hearch.co
GNU Affero General Public License v3.0
24 stars 3 forks source link

[FEAT] Search snippets support #361

Open GGLVXD opened 2 months ago

GGLVXD commented 2 months ago

Describe your feature request add "people also ask" and whatever the other one is called (screenshots) i think these would be pretty challenging to add but it would make search experience 10x better and easier

Screenshots image image

aleksasiriski commented 2 months ago

The first screenshot was already dicussed, it should be possible to have that extract from Wikipedia like how DDG does it.

As for the other one, we should find out what it is on other search engines and then it could be scraped maybe.

@matijakljajic and me will work on the Wikopedia extracts, as for the "people also ask" PRs are welcome :)

matijakljajic commented 2 months ago

As for the other one, we should find out what it is on other search engines and then it could be scraped maybe.

Remember how you've done suggestion "scraping", I believe the solution has to be something similar. @aleksasiriski

aleksasiriski commented 2 months ago

As for the other one, we should find out what it is on other search engines and then it could be scraped maybe.

Remember how you've done suggestion scraping, I believe the solution has to be something similar. @aleksasiriski

Suggestions aren't scaped, they have a proper standard API used by browsers. I don't think "people also ask" is used anywhere besides the SE's webpage.

matijakljajic commented 2 months ago

I added the much needed quotation marks in my previous comment. My point was that there's probably an API for some of those features. (It was just a guess)

Features requested in this issue are SERP features, more specifically rich and featured snippets.

SerpApi does exist for example, but I don't think it's viable for what we're trying to achieve. I believe it's worth looking at for possible ideas though. Here's their GitHub.

aleksasiriski commented 1 month ago

If this is a standardized API from Wikipedia we should consider moving this to pure frontend impl since I don't see real benefit of having it go through the backend (single source, no use in an alternative client?)