How do I combine the posterior distribution of the parameters with new data to get predictions in R-INLA?

Question

I fit a model using R-INLA, with a mix of linear and non-linear effects (but no spatial), and I understand I can generate samples of the posterior distribution of the parameters by the inla.posterior.sample() function, but I can't work out how to combine this model fit with new data in order to make predictions over this new data? Any help is greatly appreciated.

Thanks, Erin