What it did was take kanji/kana information from a certain field and instantly print out possible readings in another section.
Like:
Stardard Japanese deck
F1 (i forget what it's called as i edit them)
あなたの方が好きです。
F2
Whatever
F3
あなたの {ほう|かた} が {す|こう|ゆ|い|この} すで。
You have to set "Reading Destinastion" flag in the model properties menu. Start a new deck and look through and you'll see it.
I think it is very annoying. I turned it off as it actually made entering sentences more time consuming. I'd rather just type the original sentence all over than fish through these.
If you are talking about a different kakasi then I have no idea
Also, I'd like to know if there is a chance for kakasi to work like this website: http://nihongo.j-talk.com/parser/?demo=1 I'm concerned only with kanji->kana conversion.