#holosdiscover
19 posts · Last used 5d
With #HolosDiscover we checked multiple criteria before indexing: "indexable" enabled, account not locked, no #nobot or #noindex in bio, not in opted-out list, only public posts. Every deletion, edit or block was processed instantly via #ActivityPub.
Google uses that same "indexable" flag but ignores everything else, keeps deleted content cached for weeks.
We shut it down after pushback. Was that the right call? Don't hesitate to share, this concerns the whole Fediverse.
Contrary to what some people say, #HolosDiscover is not a scraper. It speaks #ActivityPub, and every deleted or edited post is immediately updated in the database.
It relies on the #Fediverse indexable flag, with stricter rules: accounts with #NoBot or #NoIndex in their profile, and locked accounts, are excluded. Only public posts are indexed.
I recently fixed a bug where the #NoIndex tag in bios was not respected. Sorry about that.
https://discover.holos.social/
Boosted by @fedicat@pc.cafe
I fixed a privacy bug in #HolosDiscover. If you asked not to appear in search through your bio, it did not always respect that because the #NoIndex tag in bios had a bug and the bot did not read profile fields.
In some cases discovery also fell back to the discoverable flag, which is now removed (that flag is about accounts, not posts).
Everything is fixed and I also cleaned up what was already stored. The database is clean. Sorry for that.
Replying to on mastodon.bsd.cafe
@stefano@mastodon.bsd.cafe - for #LittleFedi have you thought about using the same technique or even the say relays as #HolosSocial and #HolosDiscover to have the micro single users server speak to and see the larger Fediverse better?
Boosted by @fedicat@pc.cafe
I will investigate the issue with #HolosDiscover. Since another bot was launched on holm.community, I have been getting several complaints:
- NoIndex/nobot in bio not respected
- Locked accounts followed
- AP flag not honored
I contacted the maintainer to ask whether the code was changed. This should never happen. These complaints only started after that bot appeared.
I am taking this very seriously. If the problem is in my code, I will fix it immediately. Sorry about that.
Replying to on atomicpoet.org
@atomicpoet@atomicpoet.org Am a huge fan of #holosdiscover and actually the entire #holossocial project.
Boosted by @fedicat@pc.cafe
#HolosDiscover has an issue with accents that I'll fix ASAP. For instance, searching for "Androïd" returns far fewer results than "Android".
In case you can't see the difference, the letter "i" has a diaeresis (tréma) in the first word.
RE: https://toot.fedilab.app/@apps/116970784475112417
#HolosDiscover is a kind of search engine for the #Fediverse, fully working with #ActivityPub. That means it doesn't scrape content, it receives it organically.
As soon as a post is deleted, it's removed from the index. The same goes for edits. If an account opts out, everything is removed from the index.
It only indexes public posts from followed accounts that have the indexable flag turned on (locked accounts are excluded).
More: https://discover.holos.social
Boosted by @trending@homestead.social
#HolosDiscover is a search engine dedicated to the #Fediverse. It does not scrape content, because it speaks #ActivityPub and respects each user's choice to opt in or out.
The most respectful alternative, open to all and full of content to explore.
https://discover.holos.social
Boosted by @fedicat@pc.cafe
#HolosDiscover now displays #PeerTube videos correctly in search results.
https://discover.holos.social/
RE: https://toot.fedilab.app/@apps/116777926060900347
#HolosDiscover continues to evolve into the privacy respecting search engine the #Fediverse has needed for some while now....
Quoting
#HolosDiscover now indexes #PeerTube videos. You can find them in search and timelines, but the way they look still needs some work.
A video is not the same kind of #ActivityPub object as a normal post. A post is a short text made to be read in the timeline. A video has a title, a long description, and a media file, so it should be shown as a video card with a thumbnail and a link, not as plain text. That part is coming next.
Open quoted postBoosted by @tchambers@indieweb.social
#HolosDiscover now indexes #PeerTube videos. You can find them in search and timelines, but the way they look still needs some work.
A video is not the same kind of #ActivityPub object as a normal post. A post is a short text made to be read in the timeline. A video has a title, a long description, and a media file, so it should be shown as a video card with a thumbnail and a link, not as plain text. That part is coming next.
Boosted by @tchambers@indieweb.social
RE: https://toot.fedilab.app/@apps/116722285369696704
The most common answer is people asking what #HolosDiscover is. It's a service that indexes messages across the #Fediverse and lets you search through the data. What makes HolosDiscover unique is that it's not a scraper. It has an account, several opt-outs that automatically remove data, and most importantly indexed messages are immediately removed or updated as soon as the owner makes a change.
Have a look at https://discover.holos.social
Boosted by @tchambers@indieweb.social
#HolosDiscover 1.7.4 is available. This release fixes an issue where common words returned very few results when sorting by relevance, and it improves search quality for those words. It also makes searches faster for both rare and frequent terms.
https://discover.holos.social/
Boosted by @tchambers@indieweb.social
Concerning #Fedilab, with close to 8 million messages indexed, I think #HolosDiscover now has enough content to be worth integrating.
Full API support would allow pinning tailored timelines and interacting with new content. This would be particularly useful for small instances.
More about HolosDiscover: https://discover.holos.social/how-it-works
If you're a dev: https://discover.holos.social/api-docs
Boosted by @tchambers@indieweb.social
Holos Discover added a couple of new filters to its API search - minimum likes and minimum boosts. Additionally, you can pass contentType=image|video|audio|text|all|none to the /api/search endpoint and it works.
Interesting combinations, which I took advantage of for the Open Web Image Gallery.
Replying to
@apps@toot.fedilab.app Can you describe to me more about your planned support in #HolosDiscover, and what you have already done to support this in #FediLab?
#HolosDiscover now allows filtering results by multiple languages (for instance French + English). You can create timelines with up to 20 hashtags. You might be surprised to discover small accounts sharing hobbies you enjoy.
The #HolosSocial app already uses it. If you are an app developer, don't hesitate to use the API:
https://discover.holos.social/api-docs
#Holos 1.0.0-rc-3 has been published. This release now integrates #HolosDiscover for searches and to display trending. You can also block domains and sync remote blocking lists of domains. You can also import your list of blocked accounts. Account deletion is now possible through the app.
More about this release: https://codeberg.org/tom79/Holos-App/releases/tag/1.0.0-rc-3
Download: https://framadrive.org/s/b68BB42Rf5ogW8o
#HolosSocial
You've seen all posts