Wed, Apr 14
Tue, Apr 13
Thu, Apr 8
Wed, Apr 7
Tue, Apr 6
Fri, Apr 2
Does T277190: Return results in a randomized deterministic way make this task obsolete?
Do you have a preference for which way we present it?
Wed, Mar 31
Per Cormac on T275816: Provide which wiki an image suggestion is found on, we are okay to change the format without bumping the version. We should consider the changes from both that task and this one when making the format changes, so we change as few times as possible.
Tue, Mar 30
Thu, Mar 25
I like the last one better as well. Considering the discussion in T277190: Return results in a randomized deterministic way, I'd actually prefer to wrap the entire response in a containing object, so we'd have somewhere to put the seed value, and any other fields we think of in the future. So something like:
Wed, Mar 24
We continue to have some confusion surrounding the word "source". We are currently using it to mean both Algorithm vs MediaSearch and to specify how the Algorithm identified a suggestion.
Subscribing to follow along. May have implications for image recommendations.
Tue, Mar 23
Moving this back to "Ready" and unassigning myself, as it is not being actively worked on at the moment. I'm happy to pick it back up when priorities allow, but would not object if someone else got to it first.
Further investigation did not yield any short-term easy gains. Caching seems to be our best immediate solution for a speedup. Should we choose to add caching (and I can't imagine that we won't at some point), it would best be implemented under its own task.
@AMooney to assign a reviewer
"Bad last article" fix merged, service restarted, affected .json files regenerated.
(also - the format of the result for https://image-suggestion-api.toolforge.org/image-suggestions/v0/wikipedia/ceb/pages?offset=113476 looks wrong - there's no project or page in the response)
Fri, Mar 19
Thinking about this a bit more - we'll want to wrap the existing array response in a containing object, so that we have a place to put the returned seed. That's probably a good idea anyway. We may think of more top level data we want to return in the future. And it is a safe and trivial change on the service side.
It does, yes. @BPirkle should you guys return a seed in the response? Or just allow a user to supply their own seed? If it's the latter we'll need to specify the seed format I guess
Mar 18 2021
Returns fully randomized results, almost certain to be different every time you call it:
Thanks for the clarifications.
Mar 16 2021
At least for this phase of the project, the Image Suggestion API is implemented as a nodejs service. So MultiHttpClient doesn't apply, because we're doing this outside MediaWiki. But we are making requests asynchronously per the usual node patterns. So it isn't like we're iteratively making multiple requests to MediaSearch one after the other.
Mar 15 2021
Mar 13 2021
@BPirkle Considering what we learned yesterday in our meeting with Search team, could we leverage the pages indexed with the "has image rec" flag and cache that instead? Maybe this is what you were recommending but just trying to consider other paths.
@BPirkle Do you imagine us providing the capability for both randomized and paginated results? It sounds that bot-writers need not want randomized results
Mar 11 2021
I did a bit of profiling today.
See T277091: Consider how page sorting in .tsv files will affect client results and user acceptance for some thoughts and potential challenges regarding iteration. Basically, at least in some languages, both ends of the list in the current sort order are kinda funky from a human perception perspective. Maybe for the bot use case we don't care, though.
Mar 10 2021
Note that for the PoC, one solution might be for the client to simply start requesting results at some offset into the file to get past the "weird" results on either end. However, we should discuss all this with the client teams to confirm their needs.
This is a surprisingly tricky implementation detail.
Mar 9 2021
@hnowlan what is the impact? Is this blocking anything at the moment?
@hnowlan , are you planning to deploy this yourself, or does something else need to be done?
No objections from Platform Team.
In what way is this broken and what impact is it having? Platform Team is unsure how to help. Untagging for now, but please retag if there's work for us.