Input and preprocess of novae.Novae for re-training a new model and cluetering evaluation

[liangxs7015]

Thank you for developing this amazing multi-module tool. I have two questions when using novae.

I'm trying to retrain a model for my HD visium data. I merged my 30 samples into one big anndata and use it as input. However, I found no no parameter to refer the obs column that stores sample batch.

So the first question is that can the preprocess inside novae.Novae handle the batch effects of 30 samples? Or should I input of list of anndatas, instead of a merged anndata without any batch correction?

The second question is that is there any approach or function to evaluate the best domain numbers? I'm wondering if I could chose the best niche domains based on object evaluation, instead of subjective manual selection.

Thanks again for your work, looking forward