PyBaMM integration in co-simulation

[Matia-Torlini]

Hi everyone, I would like to run a simple pybamm DFN model coupled with an external simulator.
I am thinking about different solutions (with '.solve()' and with '.step()' methods) :

CASE .solve():

model = pybamm.lithium_ion.DFN()
parameter_values = pybamm.ParameterValues("Chen2020")
parameter_values["Ambient temperature [K]"] = "[input]"
sim = pybamm.Simulation(model, parameter_values=parameter_values)

values to manage the number of simulation cycles I need

total_time = 10
start_time=0
stop_time=1

while stop_time < total_time:
receive External Temperature
solution = sim.solve(t_eval=[stop_time - start_time], inputs={"Ambient temperature [K]" : env_temp})
publish solution["Cell temperature [K]"].entries[-1]
start_time = stop_time
stop_time = stop_time + 1

CASE .step()

model = pybamm.lithium_ion.DFN()
parameter_values = pybamm.ParameterValues("Chen2020")
sim = pybamm.Simulation(model, parameter_values=parameter_values)

values to manage the number of simulation cycles I need

total_time = 10
start_time=0
stop_time=1

while stop_time < total_time:
receive External Temperature
parameter_values["Ambient temperature [K]"] = env_temp
sim = pybamm.Simulation(model, parameter_values=parameter_values)
solution = sim.step(dt=1)
publish solution["Cell temperature [K]"].entries[-1]
start_time = stop_time
stop_time = stop_time + 1

So basically I want to run for a simulation cycle, update a parameter of the model, and restart to run a new cycle.
The solve() function accepts 'inputs' as an argument, but solving for t_eval[0,1], t_eval[1,2] etc. actually solves that time window or they are just 'logical' values? Does the method have any memory of what has been computed during the last cycle?

I saw there's the possibility to use the 'starting_solution' argument with the .step() function, but it is valid if in an experiment only, since I am in a co-simulation I do not know the operating conditions at time t of the battery, which depend on the state of coupled simulators.
Does the step() method retain memory about the last cycle? Can I restart from where I left?

Looking at the source code I saw that the step() method solves starting from self._solution . In case the step() method does not have memory, can I set the simulation object ._solution to be the last solution computed and call the step() method again?

Any clarification or suggestion will be appreciated.
Thanks.