Well, have you read the http://sonia.pitaru.com/reference/LiveOutput_start_.html page? Sorry for asking, but sometime people forget there is a documentation...
Since I just searched Sonia and found this page, I can hardly give more information than what is given in this page...

I saw the page, but I guess my question is a little vague. Sorry about that. I get the idea of the two parameters but have a hard time pi ecing together the details. For example, the documentation states:

The population of the buffer is the line LiveOutput.data[i] = sin(sinData).  The whole section of code is a for loop that runs from 0 to the length of the sound buffer changing the contents of the buffer into a sin wave at a frequency specified by the mouse position.  That buffer will then go to the sound card which will read back one value at a time at a rate of 1 sample every 1/44100 of a second (or whatever the sample rate is).

The sound is buffered because real computers with other tasks running at the same time aren't very good at reliably servicing a callback exactly every 1/44100th of a second so instead you get 2000 samples (or whatever the buffer size is) at a longer interval.

Thanks. That was a very clear and concise explanation. In this case, may I ask what is the stream size parameter for? Does that define how much data is being passed to the sound-card at once?

That's my reading of it, although the documentation is a little unclear.  Larger buffer sizes generally mean a lower chance of the sound being interupted (because you're raising generating callbacks) but also higher latency (the sound responds later to user input because you've already supplied the soundcard with a greater amount of 'future sound').