Defining conditional observation and action spaces for Sudoku in openAI Gym

Question

I am having a difficult time trying to identify the observation and action spaces to initialize my custom Gym environment with. I have functionality built out so that a 9x9