How does Loess decide how many sub-intervals to use for local regression?



I want to do local regression at each and every input point. For example:

using Loess
xs = sort(rand(100))   #100 input points
p = 1
ys = exp(p*tpoints) + 0.1 * rand(100)
model = loess(xs,ys)

when I do
I only get 2-D array of 33×3 dimension. I want to get b’s for all points in xs (i.e 100×3 dimension). Is there a way to do it?