The current implementation for world information is a mess:
-Entries seemingly stay for way too long in context as keys are triggered by the entire text feed to the AI without range limitations, flooding the context and removing precious space for more important entries and the story itself. I suggest adding an option for the activation range "Short" (<= 4 lines or <= 500 last characters), "Medium" (<= 10 lines or <= 2000 last characters), "Long" (entire context).
-Entries are added in an "unordered" way, based on the creation date of the entry, but not by priority or trigger distance. The obvious solution would be a Priority system and a secondary ordering by the distance of the triggered key.
-Entries are currently seemingly added in a rather expensive way with "[key1,key2,key3: Content of WI]\n". Listing all individual keys eats up the token budget even further and can also lead to reduced output quality. Most other services reduce that to just "[Content of WI]\n" (Possible correction: Backend might handle wi entries differently)
-Managing a large number of entries is currently a pain. Entries cannot be ordered in different ways or organized in folders.
-Advanced key triggering based on regular expressions is a powerful tool and gives the users that want it much more control while everyone else is not bothered by it. On most modern CPUs, running regexp on the last 10k characters is not particularly expensive.
-A Context viewer to inspect and test world info triggering is completely missing. The only way to see what is potentially added to the context is to use Web Developer Tools in desktop browsers.