commonists / fastcci Goto Github PK

Summary
The FastCCI server seems to be down.
Link: https://fastcci1.wmflabs.org/
Message: 502 Bad Gateway.

Ideas on how this could be done:

Certain subcategory types could be filtered out or become filterable such as "xyz in art" subcats (e.g. "Maps of the world in art" underneath Maps of the world") and there could be a lower default subcategory-depth level so that it only goes through the highest 5 subcategories (the latter would be a temporary fix or one could make it so that one could load more images deeper down on request/when scrolling [infinite scrolling]).

I've also mentioned this here. I came across this issue when lots of nonmicroscopic images (and entirely unrelated) showed up when running it on Microscopic images relating to biology.

This is a really great and underrated tool but it usually never works (I commented underneath a closed issue about this). Here I made a relevant proposal which maybe could make use of this tool or parts of the code.

The tool has a feature that shows why a specific image is located somewhere underneath a specific category (the category where the Quality images button was clicked) as shown in the screenshot.

As noted here, it would be great if there was a tool where could enter a file and a category and it would show the path/s why it's in that cat (or a gadget by which one can do so via a dropdown at the respective file or category page).

For example, I wanted to find out why chart Life_expectancy_in_Albania.svg was somewhere in the category tree of Category:Maps of the world.

I guess close this early or move it to Discussions if there are some for this repo since it's not about the gadget itself, but it would be great if you could do this since the code for this seems to largely already exist here.

The server for FastCCI has been down since at least 2020-12-11. See here: https://fastcci1.wmflabs.org/status

The max cat in fastcci_build_db.cc needs to be at least 50000000 as the program now gives off a segmentation fault now.

Currently the websocket transport is disabled due to a bug (either in libonion or the way I use it). While everything works fine with HTTP transport I'd like to get websockets up again for improved UX (live reporting of searched images etc.).

Please see https://commons.wikimedia.org/wiki/Help_talk:FastCCI#It does not work

I often use the tool, but most of the time I get:

Waiting in line. 2 ahead of us.

... or similar.
When that happens, the wait last for a very long time, so I always give up and search manually.

That's a shame, it is a wonderful tool when it works. It is essential if Commons really wants to become the place to go when looking for a quality image.

I know it is easier said than done, but how about making fastcci more scalable?

• Run several requests in parallel threads?
• Install the whole thing on several servers and do some load balancing?
• Set up a time limit? (for instance only return the results found within 1 minute)

Please see https://commons.wikimedia.org/wiki/Help_talk:FastCCI#Down_again?

Steps to reproduce: Go to a category like https://commons.wikimedia.org/wiki/Category:Mycalesis_junonia . Click "Good pictures" near the top right corner. Wait. Forever.

See also https://commons.wikimedia.org/wiki/Commons:Village_pump#Connecting...

Directed here by https://phabricator.wikimedia.org/T179122#3714112 .

The tool seems to fail to update for a few months now. No recent photos are visible.