Talk:Community Wishlist/W379

Abuse filter stats

Hi @Matěj Suchánek

It looks like some of what you're asking for exists (or used to exist) on this grafana dashboard https://grafana-rw.wikimedia.org/d/000000393/mediawiki-abusefilter-profiling As you can see the Graphite datasource has been deprecated, but we could probably move it to use Prometheus. I think that would cover the first 2 bullet points in your wish (but please let me know if I'm wrong!)

For the 3rd bullet point - there's a StaleFilters report on enwiki, looks like the bot that powers it has stopped working, but if we could get the maintainer to resurrect the bot would that do what you need?

And on your 4th point - we can't really expose the sensitive data to the replicas. Can you get all the data you need from the apis?

And more generally - it'd really help us figure out how to implement what you're asking for if you gave us a few concrete use-cases. Like - "I want to do X, and I think this would be useful for everyone who administers Abuse Filters, so I'd like a page where I can see/do Y"

CParle (WMF) (talk) 09:19, 6 May 2025 (UTC)

Hi @CParle (WMF),

Thanks for your feedback. In general, the main idea of the wish is that all the pieces of information are centralized and provided directly by the AbuseFilter. This would be useful for users as they wouldn't have to create bots for something that could otherwise be pulled out from the database. Also, access to potentially sensitive data tables can be bound to the ordinary user rights managed on-wiki (which is not possible in database replicas).

As for the performance stats, I am actually aware of that dashboard, but I find it most useful to system operators and developers (even though some per-wiki stats can also be of interest for filter maintainers). The stats there are rather aggregate, they do not include for instance per-filter breakdown, which is available only in textual form on individual filter pages.

So one particular story could be "I want to see all data about filter performance in a single table, so that I can order filters by performance, see which filters perform the worst, and act on it."

Another one could be "I want AbuseFilter to produce a table like w:User:MusikBot/StaleFilters/Report, so that filter maintainers on any wiki can benefit from this functionality."

In fact, there is a proof-of-concept patch for these two use-cases: [1][2].

I imagine many more use-cases exist, but this would also need some input from other filter maintainers.

--Matěj Suchánek (talk) 13:40, 7 May 2025 (UTC)

Ok cool, thanks for that @Matěj Suchánek ... we'll have a chat about it on the team (though be warned it'll probably take us a week or two to get back to you) CParle (WMF) (talk) 13:52, 7 May 2025 (UTC)