Yeah, that looks good. We can say that a/b is "a kipisi pi b". Now all rationals are ℚ. And we can open up our space by using la for something like "nanpa pi la nanpa leko nanpa li a" for the square root of A and giving special names for irrationals such as "nanpa sike" for pi and "nanpa kasi" for the golden ratio (alluding to its occurrence in many plants)
That's not how you say "divided by" in toki pona using kipisi. Using your method, tu kipisi pi tu would mean "two divided by two," however you can't have only one word after pi.
Something like mi kipisi tu e san would be good for "I divide three in two," and then after stating that you can use ni to refer to it afterwards.
mi kipisi tu e san, ni en tu li mute kili. > "I divide three in two, this plus two is the amount of fruit." > "Three and a half is the amount of fruit."
I also suggest X pi kipisi Y, "X, of pieces of Y." Fr example, tu tu pi kipisi tu would be "four out of pieces of two," and when you make four out of pieces of two, you get two pieces in total. (This one may work worse)
kin la my name in toki pona, jan Tu Leko comes from me trying to translate "squared" directly, but I think this is more of a calque, and I reinterpreted it to mean actual square shapes.
I'm not a big fan of leko for squares and square roots. I suggest actually explaining what a square root is, like mi pana e leko. suli pi poka leko li tu. suli pi weka ni li suli wile: kiki ni (points to one corner) en kiki ni (points to the opposite corner).