How to access to direct and indirect effects using GAM with PiecewiseSEM

[74RoMaIn]

Hello,

I am running a SEM with the piecewiseSEM package by using GAMs (cuz this is useful to choose the topology of the DAG).
However, as explained in the documentation of this package, no estimates are given in the summary of the SEM because we are fitting a truly non-linear curve and so we don't have a single estimated parameter.
However, I was wondering if there was a method to access these direct and indirect effects? Through the R^2 for example, as a variance partition analysis? Or even another method that could avoid translating our model into parametric ?

Thanks in advance,
Romain.