Deployment styles

A published notebook can show its code to the person running it, hide the code behind the widgets, or hide it for good. You choose the style once, in the tool's invoke script, and it applies to everyone who launches the tool.

The three styles

Style	What the user sees	start_jupyter flags
Notebook	Every code cell, and the full notebook interface. The default.	none
App	Widgets, markdown and output only. An Edit App button reveals the code.	-A
Tool	Widgets, markdown and output only. No way to reveal the code.	-A -t

Notebook style suits a shared analysis, where the point is the code and the reader is expected to change it. App style suits teaching and demonstrations: the tool works out of the box, and the curious can still look inside. Tool style suits a finished application whose code would only distract, or whose author does not want it read in the browser.

Whichever style you pick, changes a user makes while running your published tool are not saved back to it. Each launch starts from the notebook as deployed.

Choosing a style

App and Tool style both rely on the Appmode extension in the hub's Jupyter installation. Where the hub has it, an Appmode button appears in the notebook toolbar, and pressing it shows the notebook as App style will show it; Edit App switches back. Use that button to check the layout before you deploy, since a notebook that reads well with its code visible often needs rearranging without it.

If the button is not there, ask the hub's administrators which Jupyter version and extensions they deploy before committing to App or Tool style.

Setting the style

The flags in the table go on the start_jupyter call inside the invoke script, not on invoke_app. Both commands happen to use -t, and they mean different things:

• invoke_app -t names the tool.
• start_jupyter -t hides the code cells for good.

Invoke scripts for Jupyter notebooks gives the full script for each of the three styles, and the rest of the arguments.