Thanks for the feedback! Right now, we're extracting data from New Relic/Librato, so the code expects JSON format. However, the plan is to write plugins to convert arbitrary data formats to something that the core ML stack can operate on.