Optimize on two correlated arrays in ORTools - or-tools

Question

Is there some general rule on how to model soft constraints of correlation between two arrays of dependent variables in OR-Tools?
I am trying to solve a bit of a complex shift scheduling problem and I cannot wrap my head around it. The staff is split into two teams and the general rule is that only people from one team work except when there is a need to cover vacation or sick days. I imagine something like this:
day 1
day 2
day 3
day 4
day 5
team 1
x
x
-
-
x
team 2
-
-
x
x
-
x - work
'- - rest
and then each worker works primarily on the days where their team works except when there is a need to cover for somebody from the other team:
day 1
day 2
day 3
day 4
day 5
worker 1, team 1
x
x
-
-
x
worker 2, team 1
x
x
-
-
-
worker 3, team 2
-
-
x
x
x
worker 4, team 2
-
-
x
x
-
Notes:
day 5 is an example where worker 2 had to take day off and worker 3 from the other team covers.
there are also other complications (shifts, skills, etc.) for workers, omited for simplicity
Now if I have the below arrays, how do I tell or-tools to assign workers in the working team first and cover with others only when not possible to meet some of the other constraints, i.e. model the soft constraint between workers and teams?
team_assignments = {}
for d in range(num_days):
for t in range(num_teams):
team_assignments[d,t] = model.NewBoolVar(f'day_team:{d},{t}')
work = {}
for d in range(num_days):
for e in range(num_employees):
work[d,e] = model.NewBoolVar(f'emp_day:{e}_{d}')
In other words how do I express the penalty if workers 3 and 4 are working on days 1. 2. and 5?
Note this is doing a hard constraint where I need a soft one:
for d in range(num_days):
for e in range(num_emp):
model.Add(work[d,e] == team_assignments[d,emp_team[e]])

Answer 1

This seems to do the trick:
for d in range(num_days):
for e in range(num_emp):
t = emp_team[e]
works_with_other_team = model.NewBoolVar(f'emp_team:{e}_{t}')
model.Add(works_with_other_team == True).OnlyEnforceIf(work[d,e], team_assignments[d,t].Not())
obj_bool_vars.append(works_with_other_team)
obj_bool_coeffs.append(3) # pick appropriate penalty
and then minimize over var*coeff
model.Minimize(
sum(obj_bool_vars[i] * obj_bool_coeffs[i] for i in range(len(obj_bool_vars)))
)
Initially I was trying to do something like the below and it was failing
model.Add(works_with_other_team == work[d,e] & team_assignments[d,t].Not())
and when this failed with NotImplementedError: calling and on a linear expression is not supported, please use CpModel.AddBoolAnd I went on trying to implement it with AddBoolAnd or AddBoolOr but the much more elegant solution is to use OnlyEnforceIf