Predict which DOM element a user will interact with next.

View the demo

You give it a list of elements and it will try to predict when a user is about to mouse over one of those elements.

It's pretty naive, it just looks at the velocity and position of the mouse and tries to find the element that you are probably moving towards based on that.

It uses a voronoi under the hood, but instead of looking at current mouse position it looks at expected mouse position. In my testing this seems strictly better than the voronoi / bubble cursor technique.

You could use it to perform some optimizations, maybe similar to instantclick.

The library has these premonitions, but they aren't always right.

$ npm install premonish

import Premonish from 'premonish';
const premonish = new Premonish({
  selectors: ['a', '.list-of' '.selectors', '.to', '#watch'],
  elements: [] // Alternatively, provide a list of DOM elements to watch
});

premonish.onIntent(({el, confidence}) => {
  console.log(el); // The DOM node we suspect the user is about to interact with.
  console.log(confidence); // How confident are we about the user's intention? Scale 0-1
});

var premonish = new Premonish({
  selectors: [], // list of selectors
  elements: [] // A list of DOM elements
});

At least one of selectors or elements is required.

The onIntent callback will be called when premonish thinks that a user is likely going to interact with one of the watched elements.

premonish.onIntent(({el, confidence}) => {
  // el is the expected DOM element
  // confidence is a score from 0-1 on how confident we are in this prediction.
});

The onMouseMove callback allows users to look at some of the internal calculations that premonish is making.

premonish.onMouseMove(({ position, velocity, expected }) => {
  // Each value is an object { x: number, y: number }.
  // `expected` is the approximate point premonish thinks the
  // user is moving the mouse to.
});

Call stop() when you are done using the library to cleanup the events it is using.

premonish.stop();

Matthew Conlen (github@mathisonian.com)

MIT