Hi Paul,

You can't transplant the kernel per se. You can publish your notebook as a 
standalone web application on the dashboard server.

The text and diagrams on the dashboards wiki here illustrate the concept: 
https://github.com/jupyter-incubator/dashboards/wiki

> If there's such a workflow, how would the dashboards server replicate not 
only Python virtualenv, but also environment variables that the notebook 
may depend on?

There's no magic here. If your notebook depends on pandas, some special env 
var, some local JPEG images, etc., you need to ensure the kernel gateway 
providing the kernel for the dashboard has pandas, that you start it with 
the same environment variable, that you bundle your JPEG images with your 
notebook, etc. Some of this is automated by the dashboard bundling process. 
But other parts, namely resolving kernel-side dependencies, is still a 
manual process.

That said, using the same Docker image between notebook server and kernel 
gateway and putting one-off pip/conda commands right in the notebook for 
eval on first run can take you a long way without much hassle in my 
experience.


On Friday, July 8, 2016 at 5:41:39 AM UTC-4, Paul A wrote:
>
> Is it possible to "transplant" the kernel onto the server?
>
> I.e. you develop a notebook on your laptop, click Deploy as Dashboard on 
> Jupyter Dashboards Server, close your laptop lid, and somehow the dashboard 
> keep running.
>
> If there's such a workflow, how would the dashboards server replicate not 
> only Python virtualenv, but also environment variables that the notebook 
> may depend on?
>

-- 
You received this message because you are subscribed to the Google Groups 
"Project Jupyter" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jupyter/c4c56152-1c69-432b-a7c4-5d781c030a1c%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to