Untitled Language

Here’s a sample of a very incomplete conlang I’m working on.

declension for raŋut (“tree”) (Middle period)

This isn’t even a skeleton, it’s a couple of little bones.
Of what I’ve come up with so far (there’s not a lot but there is quite a bit more than what’s here), the early form’s pretty isolating (and thus analytic) and the middle form has a lot of agglutination going on. I know that it’s more accurate to use these labels to classify features than to classify languages, but it is possible for a language’s set of features to skew way more one way than another, and creating fusional inflection from scratch (i.e. without making the morphemes that it’s fused from) seems way less naturalistic than creating free morphemes from scratch, so I think it makes sense to start with a language with no fusional inflection.
I ~~pulled some click consonants out of my ass~~ derived click consonants from sequences of other consonants despite having no way of knowing for sure if or how they develop other than being borrowed.
I want the late form to have a vowel harmony system of /i y ɨ u ɛ œ ɐ ɔ/, but though I understand that vowel harmony comes from coärticulation, I still need to research what specific traits would incline a language’s speakers to coarticulate vowels and then expand that pattern by analogy.
I’m on the fence about /q/.