Jupyter APIs

I read this blog article from an engineer at IBM that exposes his jupyter cells as rest API endpoints. I've adapted this to my own collection of NodeJS notebooks. I can run cells as individual functions or upload a collection of functions as more of a monolithic service.

Recently, I've been enjoying the idea of picking out one cool feature from a project, and only reimplementing that single feature. For example, I read Fuse.js search uses levenstien, so I made a simple mockup of a search algorithm using another levenschtein implementation I found online. This is a lot like the search algorithm code I blogged about earlier.

levSearch code

Now, any function in the collection of notebooks can be accessed remotely by name, or by the markup that comes before it. I could even execute test cells to turn features on and off automatically based on the code that is available. Why not inject apps directly in to the output of a notebook?

Documentation RPC screenshot

The screenshot above is another example of a simple proof-of-concept project. I like the idea of automatically generating documentation, so I created a simple template and I give it a Github repository and it generates documentation for every API call.