I'm doing this in my own fork anyway but an arg to have tempo python lib load series data to / from pandas time series objects would be stellar. Our data is pretty much touched by pandas anytime read.