If it means anything, I still have a branch open to make a https://topviews.toolforge.org -esque visualization of toolviews. See T87001: Provide basic page view metrics for individual tools on toolforge for the original task. I was going to work on it at the Wikimedia Hackathon 2020, but well, a pandemic happened instead.

This may be a good project for Inspiration Week! I basically just need to fork the Pageviews Analysis code and only keep the Topviews bits, and obviously use the toolviews API instead of pageviews (fortunately the response format is mostly identical). More complicated than it seems, but it can be done in less than a week.

That sounds terrific!

Can it tell us e.g. how many exports WS Export has done, or translations uploaded by SVG Translate (both things that have their own URL patterns)?

I'd worry about counting just hits to a tool's homepage, because that's going to be affected by bots etc. and isn't always a good indicator of how "well used" a tool is.

Can it tell us e.g. how many exports WS Export has done, or translations uploaded by SVG Translate (both things that have their own URL patterns)?

I'd worry about counting just hits to a tool's homepage, because that's going to be affected by bots etc. and isn't always a good indicator of how "well used" a tool is.

Toolviews apparently goes by the Nginx access logs (not the access.log that may or may not be enabled for the tool), so I'm guessing it picks up hits to any path. I still agree however it may not accurately reflect usage. I.e. https://siteviews.toolforge.org doesn't make any requests to the tool's backend, so a single session could include tons of unique requests but there will only be one registered use in toolviews.

At any rate, I was just thinking out loud in my effort to create a toolviews app. Don't let me stop any existing efforts built something for our team, specifically (as this task implies)!

Thank you both for the comments! 🙂 that sounds really interesting @MusikAnimal, I'll definitely take a look — I agree this wouldn't be the most accurate statistics, but at "least" this inaccuracy will be somewhat consistent between the monitored tools?

In T311613#8070621, @TheresNoTime wrote:

Thank you both for the comments! 🙂 that sounds really interesting @MusikAnimal, I'll definitely take a look — I agree this wouldn't be the most accurate statistics, but at "least" this inaccuracy will be somewhat consistent between the monitored tools?

For most of our tools, yes, toolviews should provide a good picture of usage. This of course doesn't include bots or things lacking a webservice, but that's okay. The bots we can infer usage from the edits it makes and/or pageviews to those pages.

Also I believe toolviews only applies to Toolforge. WS Export and a few others will still get some data from old links that now redirect to VPS, but just be aware if the numbers are low it's probably because most hits are directly to VPS.