When opening a new Kagi Translate window, source language defaults to "Detect", which usually works great (it does sometimes have problems differentiating between Danish and Norwegian, for example, but it gets it right often enough).
The target language, however, switches randomly to languages I never use, and I have no idea what heuristic is even attempted, since there is no obvious reason why I should want to translate a particular text into e.g. Hungarian. I never use the language, and the source text doesn't contain any hints.
A better choice would be to default to whatever target language I last used. Kagi remembers these, as they are listed at the top of the target language dropdown. So it should be a small improvement to select one of these before a completely random target.
I should rather have that it doesn't translate anything until I choose the target, than it almost always chooses wrong a target for me, since I get stressed out about activating the translation, wasting tokens on something every single time I use the app. This doesn't impact me negatively, but I like to believe that if Kagi doesn't waste tokens, I get more value somewhere else. Surely the optimization here carries some non-trivial saving.
Thanks for your great work!
When a user types in something, the source language should default to "Detect" as it already does. And the target language should either default to the most recently used language, the most frequently used language, or no default at all, not activating translation until a target language is chosen. I just don't need automatic translation into Hungarian. Very few people do, mostly Hungarians.
