JarCastingkdgrass 
KDBush with flat API, which turns out to be even faster.
const kdgrass = require('kdgrass')
let grass = kdgrass(points);
let ids1 = grass.range(10, 10, 20, 20); // bbox search - minX, minY, maxX, maxY
let ids2 = grass.within(10, 10, 5); // radius search - x, y, radius
API
let grass = kdgrass(points, nodeSize?)
Creates an index from the given points.
points: Input array of points in[x, y, x, y, ...]form.nodeSize: Size of the KD-tree node,64by default. Higher means faster indexing but slower search, and vise versa.
let index = kdgrass(points, 64);
grass.range(minX, minY, maxX, maxY)
Finds all items within the given bounding box and returns an array of indices that refer to the items in the original points input array.
let results = index.range(10, 10, 20, 20).map((id) => points[id]);
grass.within(x, y, radius)
Finds all items within a given radius from the query point and returns an array of indices.
let results = index.within(10, 10, 5).map((id) => points[id]);