Which package is best for adding a random effect to multinomial regression model?

In my experiment, nine toads were housed in one of two enclosure types "Basic" or "Enriched". In either enclosure type they could be in one of four postures "Low", "Medium" "MediumHigh", and "High". I want to know if Enclosure type could predict the posture a toad is in.

The data format has a column for ID, a column for enclosure type (Basic or Enriched) and a column for Posture (Low, Medium, MediumHigh, High).

I started off with some basic code that worked just fine: multinom(Posture ~ EnclosureType, data=ungrouped.drop)

However, I need to account for individual variation among toads (ie a random effect). It doesn't seem that there is any way to add random effects using "multinom". (Please correct me if I'm wrong)

So I turned to the mlogit package and attempted to use the npar aregument to specify the random effect: mlogit(Posture ~ BasicEnriched, rpar=c(ID="n"), data= ungrouped.drop)

I got the following error: Error in dfidx::dfidx(data = data, dfa$idx, drop.index = dfa$drop.index, : The choice variable must have exactly two modalities

Since my choice variable (I believe it's "Posture" but correct me if I'm wrong) has more than two modalities I decided perhaps mlogit wasn't right for my scenario and turned to mclogit instead. (Please correct me if I'm wrong): mclogit(Posture ~ BasicEnriched, random = ~1|ID, data=ungrouped.drop)

And I got this error: Error in mclogit(Posture ~ BasicEnriched, random = ~1 | ID, data = ungrouped.drop) : need response counts and choice set indicators

I think this may be a problem with my data format which would require mlogit.data to reformat. However, the mandatory choice argument converts my non-binary "Posture" variable into a binary "True" or "False" which doesn't seem right. I'm also not sure if my data format is wide, long, vertical, or something else.

Your help is much appreciated!