amd-three.js is an example of how to structure an three.js application, with functionality split into different Require.js modules.
It is meant as a starting point for more involved projects involving three.js, where having all the code in one file can quickly get unwieldy.
It is also useful for prototyping, as a lot of the boilerplate is moved out of your way.
Live demo of this repository can be found at http://felixpalmer.github.io/amd-three.js/
Just host this directory with a webserver of your choice. You can also use the webserver.sh script included (provided you have Python) to set up a simple development server.
Then visit http://localhost:8000 in your browser.
Custom shaders are put in /js/shaders and are saved as .vert or .frag files, for ease of editing. To convert these into a require.js module, run /js/shaders/compile.py. This will create /js/app/shader.js, which will export a shader object with the relevant shaders attached. E.g. test.frag will be available as shader.fragment.test
The structure of the file is as per the recommendations in the docs for Require.js. Namely, all the Javascript lives in the js folder, with the Three.js library and extension code in js/lib, while the app specific code is in js/app. The idea is that you shouldn't ever need to modify the files in js/lib, only add third party code.
The common three.js constructs, e.g. camera, renderer are all put into their own modules within the js/app directory, and are injected into the js/app/app.js file where they are brought together. They sometimes include each other, e.g. the light module needs to know about the scene. By splitting these object into modules, we are in essence creating singletons, which means we don't have to worry about which order we create them in, as Require.js will figure out the dependencies for us.