I'm not 100% sure if it's possible to immediately recognize whether a request cam from within a thread in Slack API but it's worth checking into, to make the responses consistent.

The Wikipedia API (REST and Action APIs) assume case sensitivity for titles. It automatically capitalizes the first character, but requires proper capitalization for the rest of the title.

As a result, for a page that exists: Edward Colston

• edward Colston works (first character is auto-capitalized)
• edward colston does not work (changed to Edward colston which does not exist)

This might make some sense inside Wikipedia since pages with different capitalizations are indeed different pages but it is making it very difficult for bot users to understand why requests are failing.

If a result list presented an entry "Edward Colston" and the user follows up with @wikibot summary edward colston then getting a "page not found" is confusing.

This ticket is about attempting to find a valid way to resolve this issue.

When a page is requested (and potentially for search results?) add links to the page's talk page and history page.

This might be the result of adding the fallback get command... But a direct usage of "search" command only returns the first result rather than a list, as intended.

The 'top' command fetches from the Wikimedia Analytics cluster, that works with UTC. As a result, asking for results for a certain date is only available if that day is "over" in UTC time (the calculations for 'top' seem to be bundled per day rather than rolling)

The problem starts with assuming a fetch for "today" (when asking for top without date parameters). If today's date, in the bot's timezone, is a day after the current UTC date (for SF, for example, this is basically true after 5pm) then the response is that "no results are found" but that is because the assumption of "today" is wrong.

I tried to fix it up by having a method comparison checking if the requested date is before UTC, but it isn't working very well. This should be fixed more in-depth.

The search action API returns the results with a "rank" property that dictates the order. This order isn't preserved in the actual object that's created, so the returned processed results should use the 'rank' to order themselves.

Use an empty image instead: https://commons.wikimedia.org/wiki/File:Transparent_square.svg

The wikipedia.go piece can be split to its own module, allowing to fetch data from the APIs -- splitting it out would mean that we can also look into allowing this "bot" to operate in other interfaces, like:

• WhatsApp bot
• Discord bot
• Telegram bot