Don’t know if this is the right comm. Let me know if there’s a better one.
Inthe last few days I noticed that the results of some public searxng instances has gotten way worse (mostly cyrillic/chinese text when searching english text with :en).
Does anyone else have that problem? Die google start the next anti-consumer tactic to push their AI garbage?
This might be Bing, i noticed that Bing always returns the worst thing possible.
I configure the engines to use brave, duckduckgo, google, qwant and yahoo, always give me good results, might be the law of large numbers in action.
Second that. I’m self hosting my own instance. If I only get some kanji (or sometimes, but very rare) cyrillic results, they’re usually by bing
DDG is just Bing results. I thiiink Qwant might be as well.
They do some treatment to the results too + internal data from them, it’s not copy pasted from Bing, so the results turn out different for duckduckgo and other Bing downstream search engines.
I run my own SearX instance (https://searx.andmc.ca/) and really haven’t noticed “worse” results. Granted I haven’t really touched the configuration in several weeks but it’s been working fine for me. There’s nothing special about my instance either it’s just straight up default settings for almost everything.
Maybe it’s just whatever instances you’re using? I don’t know. I mean feel free to give mine a shot to see if it’s any better.
That one is indeed better. Interesting how my default ones have these problems.
Maybe I should spin up an instance on my VPS as well, when I find the time.
edit: Thanks for sharing, btw! ^^
it’s very easy to setup. for me it took me all of like 5minutes via docker and like I said I didn’t change much at all, if anything really.
The “easy things to set up” are just currently piling up. ;)
I’ve noticed this too.
My assumption is that it might be mostly Chinese instances being hosted, configured to return such results, but it’s just a hunch.
How do you use SearXNG btw?
I use instance randomizers,
so each query goes through a random SearXNG instance, e.g:GimmeASearX could help here,
since you can white/blacklist instances you want to use, but that can become tedious.Searloc + Neocities are great for mobile devices and other places where you can’t self host GimmeASearX.
Those instance randomizers seem like a good idea, thanks!
It is an aggregator… So Bing and Google are just being enshitified
Also what servers do you use?
I don’t know if I’m doing them a favour by increasing their traffic, so I’d rather not say. ;)
Can confirm, still getting results in English, but most of my preferred instances (granted I have a bias for ones with shorter URLs) have been returning unusable results. Google’s rate-limiting mechanism is probably hitting them hard.
Hopefully the SearXNG devs can find a workaround soon, but it might be time for me to consider self-hosting or taking a look at kagi. I personally can’t put up with plain DDG since its autocorrect is too aggressive.
Yeah nothing but random results. I had to switch to ddg for now.
I think it’s pretty obvious that it’s not their doing and rather downstream search APIs. I’d reach out to their development team.
Yeah, that’s what I figured. I wanted to ask in a “lower stakes” forum first, though. To find out if the issue is already known.
Just not to annoy the devs with an issue report that’s already been reported (and which report I overlooked)
Yeah I think it’s similar to YouTube where specific JavaScript functionality is required to execute in order to collect results from Google. This impacts others like Whoogle too unfortunately.
Yeah, exactly that - see: https://github.com/searxng/searxng/issues/5286
