I guess different people do it differently. Here is my view:
In each rackspace, you are using a mixer (or maybe just a gain/balance if you are only using one instrument in the rackspace).
Then you would set up a widgets connected to the mixer channels you are using.
There is a great six channel widget panel that I now always just set up for each rackspace. (Thank you Joey!). It has all the meters and all the faders and knobs you would typically ever need. I love it!
Here it is:
If you use setlist mode (like me) you might often use the “snapshot” feature to try to get the right levels within each song and across all songs in a setlist so you do not need to manually ride the volume as much. The “snapshot” feature basically allows you to modify any widget in the variation on a song and song-part basis. It is extremely useful, especially to get all the relative volumes lined up within a song and across a setlist.
I may be a bit weird in that I have my audio interface sitting on my keyboard and use that as a master controller for GP. I don’t think that is the norm.
If I am using internal keyboard sounds, I just use the volume controller on the keyboard. But, again I am a little weird (or let’s just say “in the minority”) in that I do not run the audio from my hardware keyboards back through GP. That is probably a better approach, but, for a number of reasons (laziness?!) I have not gone that route.
I hope this helps.