My first thought is that the verbosity levels could use some constants in place of those 0,1,2,3s. Because that's the _only_ user unfriendly part I see in the whole API..

I'd also like to eventually see it accept a parameter in the use Log line with the filename etc.., where you can give it a custom subref with a single scalar as input and output, to reformat the data before output.

Looks good overall, nice job.