Programming language R: meaning of 'weights' parameter in library method 'loess' -

i use library method loess of r programming language non parametric data fitting. dataset two-dimensional. have not found proper documentation of method parameter weights.

my data points distributed random variables, , have estimate of respective standard deviations. wondering whether parameter weights allows me supply r details of standard deviations. in other words: wonder whether individual weights in weights (relative) measures of data quality, fit can improved if measure of data uncertainty supplied via parameter weights.

edit: suspect entries in weights used weights in weighted least squares regressions of local datasets in loess procedure (maybe additional weight prefactors (position dependent) kernel functions?). suggest case of data points independent distributed random variables, still have different noise levels (i.e. different standard deviations) (as in case), weights should chosen 1/\sigma_{i}^2, \sigma_{i} standard deviation of respective random variable/data point. if knows sure, nice know.

this page confirms suspicion:

https://docs.tibco.com/pub/enterprise-runtime-for-r/3.1.0/doc/html/language_reference/stats/loess.html

regarding parameter weights of loess says:

an optional expression weights give individual observations in sum of squared residuals forms local fitting criterion. default, unweighted fit carried out. if supplied, weights treated expression evaluate in same data frame model formula. should evaluate non-negative numeric vector. if different observations have nonequal variances, weights should inversely proportional variances.