No, the chart in the manual page was from a programming job. The inputs were a list of indicators that were particularly required by the client. The 3 layers with 20 neurons in the example was just a random setting for demonstration.

adviseLong() and adviseShort() do exactly the same, they just train two separate models. So you can indeed use only one advise function for both long and short prediction. Or for example one advise function for predicting and the other one for filtering with a different method.