I envision loopback's datasources & models as a useful tool for consuming an API, not just for automatically creating a REST API.
Pretend I'm using Spotify's API. I want a command line application that looks up user information on Spotify. I could write loopback models to consume say /v1/albums.
I'm hoping I can create an Album model and use it like Album.find('Thriller');.
What I don't want is to create a REST API. I just want a better language to consume other people's APIs such as Facebook or Instagram.
A LoopBack model is just a config file and a collection of functions in a model file. All of the endpoints are generated by LoopBack itself, and as such not very helpful by themselves. That said, you can generate a swagger spec for each of your models with the explorer component (installed by default when you use the scaffolding CLI: slc loopback).
Simply start up your LoopBack application, then navigate to:
http://localhost:3000/explorer/resources/MyModels
You could then use the swagger spec in any framework that supports that standard.