One interesting use case for the Eval API (reppo.xyz/evaluation-api) we are seeing as robotics moves into deployment and out of controlled environments inside labs, deployers as well as intermediaries are looking for failure data ie how often does a robot fail at the task it was designed to do. We are working with @eastworlds_io to showcase the use of the Eval API by a robot actively accomplishing a task. This would be a first where robots can access almost real-time RL ⛽️