Declension of -i and -u stems

I've been realizing for some time that Seadi as originally designed has an artificially small number of declension classes. Back in 2005 I figured out consonant-stem nominals, and I knew that I was going to have to open up -i and -u eventually; apparently "eventually" meant about three years later, and here's my first draft as of this morning's BART ride.

Stems in -i, Singular

A turr-i-Ø > turri
E turr-i-nis > turris
G turr-i-v > (turriu)
I turr-i-hin > turrin
S turr-i-še > turrie
P turr-i-da > turria
V turr-i-t > turri

Stems in -i, Plural

A turr-i-i-Ø > turrī
E turr-i-i-nis > turrīs/turries?
G turr-i-i-v > (turriju)
I turr-i-i-hin > turrīn/turrien?
S turr-i-i-še > turrīs
P turr-i-i-da > turrija
V turr-i-i-t > turrī

Stems in -u, Singular

A karh-u-Ø > karhu
E karh-u-nus > karhus
G karh-u-v > (karhū)
I karh-u-hun > karhun
S karh-u-še > karhui
P karh-u-da > karhua
V karh-u-t > karhu

Stems in -u, Plural

A karh-u-i-Ø > karhui
E karh-u-i-nis > karhūs/karhīs
G karh-u-i-v > (karhuju)
I karh-u-i-hin > karhūn/karhīn
S karh-u-i-še > karhuis
P karh-u-i-da > karhuja
V karh-u-i-t > karhū

Now for the interesting part: I need to decide how paradigmatic leveling is going to affect not only these paradigms, but nominal inflection in general. Given that Seadi is inflectional and not agglutinative, there really ought to be a lot less consistency in the declensional endings.