Episcleral plaques are commonly used for the treatment of ocular tumours such as choroidal melanoma. Treatment planning involves the assignment of seeds to slots on the plaque to achieve a desired dose rate distribution. Seed assignment is rather straightforward if seeds are ordered on demand. However, the assignment task becomes tedious and laborious if the seeds have to be chosen from an existing stock of seeds with different activities. To date, this task has usually been performed by a human planner through trial and error. An algorithm has been developed to automate the task of seed assignment using a mixed-integer programming method. We also explore ways to simplify the problem such that the method becomes practical in most facilities. We have tested the method on three randomly chosen clinical cases from our past records, to show that the algorithm could yield solutions within a shorter time frame and with less deviation from the desired dose rate distributions, as compared with the solutions from a human counterpart.