How to Use DataFrames for In-Memory Input and Output with Zingg Instead of File Storage?

Also, is there a way to keep the model output in memory in a DF instead of writing it to files, such as, in my case, to S3? Conversely, can I pass an input dataset from a DF? I guess this would provide greater flexibility for pre- or post-data processing in memory, rather than requiring intermediate reads/writes. I appreciate that Zingg is configured via config files that point to FS locations, so maybe not, but I'm just checking.