What is it
It's an attempt to do a well done real-time search for Mastodon. Something I miss from Twitter is to be able to search what people are talking about breaking news and other heated events
This whole thing is provided as is.
How it works
- It listens to some instances local timelines via their API.
- It collects posts and keep them for a short time (currently 24 hours).
- It only collects posts from authors that opt-in to search indexes (honors noindex flag).
- It sends User-Agent header with following value when calling instance APIs: Mozilla/5.0 (compatible; Masto.Kukei.eu-Bot/0.2; +https://masto.kukei.eu)
Who's behind
It's created and maintained by Szymon Nowicki, you can also check all the source code on GitHub
You can always contact me on Mastodon or hey@nowicki.io
Stats
At the moment it talks to social.josko.org, pol.social, 101010.pl, mastodon.social, mastodon.online, fosstodon.org, gruene.social, hachyderm.io, troet.cafe, mastodon.art, ioc.exchange, nrw.social, norden.social, sueden.social, social.cologne, muenchen.social, ruhr.social, hessen.social, chaos.social, lewacki.space, mas.to, social.vivaldi.net, front-end.social, ard.social, zdf.social, edi.social, hachyderm.io, mastodon.au, kind.social, cyberplace.social, mstdn.social, mastodon.com.pl, defcon.social, flipboard.social, mastodon.scot, toot.community, ohai.social and has
Browse Disclaimer
/browse is an experimental browse page. You can browse by categories and/or languages.
It is all a result of LLM processing using local Ollama models on my computer, Ollama.com cloud and/or using OpenAI compatible providers. Some posts are filtered out as I'm a chicken and I don't want risk showing something illegal here. Ergo, some posts that model considers "harmful" are being flagged as "banned" and not shown.
To see how this works exactly, go to the source code to /scripts/make-categories.js
Summaries Disclaimer
Few times a day I take last 1000 posts in a previously assigned category (news, technology, programming), put it through AI to try to make a "summary of what people are talking about".
I believe this is one of the "good things" about LLM, it can quickly make summaries, that are mostly correct, from a lot of content.
I'm experimenting with it for some time and since I take all posts (in all languages), I find the biggest value from discovering topics that are out of my language bubbles.
As the "browse" page, this is all experimental and I might sunset it without a notice.
Privacy policy
This website is hosted on my living room PC. It doesn't use any extensive tracking, only Plausible.io powered, backend proxied tracking to further anonymize visitors IP
I do log queries due to abuse (some creeps are using this search to search for creepy stuff. It's not even here, but I keep banning them.
I sometimes check logged queries, then I remove them.