Skip to content

Server that runs and renders Jupyter notebooks as interactive dashboards

License

Notifications You must be signed in to change notification settings

maqetta/dashboards_server

 
 

Repository files navigation

Build Status Google Group

Jupyter Dashboards Server

A NodeJS application that can display Jupyter notebooks as dynamic dashboards outside of the Jupyter Notebook server.

Dashboards server screenshot

The Jupyter Incubator Dashboards effort covers:

  1. Arranging notebook outputs in a grid- or report-like layout
  2. Bundling notebooks and associated assets for deployment as dashboards
  3. Serving notebook-defined dashboards as standalone web apps

This repository focuses on (3) above, while jupyter-incubator/dashboards handles (1) and jupyter-incubator/dashboards_bundlers implements (2).

See https://github.com/jupyter-incubator/dashboards/wiki for an overview of the entire dashboards effort.

What it Gives You

  • Ability to run a Jupyter Notebook with layout metadata as a standalone dashboard application
  • Ability to navigate a list of multiple notebooks and select one to run as a dashboards
  • Optional shared login to secure access to the dashboard server
  • Ability to add custom authentication mechanisms using the Passport middleware for Node.js
  • API for POSTing notebooks to the server at runtime with optional authentication (/_api/notebooks)

The behavior of the application is similar to that of Thebe, but with some key technical differences:

  • The notebook code is visible only in the NodeJS application backend.
  • The NodeJS backend is the only actor that can send notebook code to a kernel for execution.
  • The browser client can only send Jupyter comm messages to kernels (not arbitrary code).
  • The application uses the jupyter-js-services and jupyter-js-widgets libraries for communication with kernels.

The following libraries are known to work with the dashboard server:

  • jupyter_dashboards 0.6.x
  • jupyter_dashboards_bundlers 0.8.x
  • ipywidgets 5.1.5+
  • jupyter_declarativewidgets 0.6.x
  • Bokeh 0.11.x
  • Plotly 1.9.x

Install it

You can install the dashboard server using npm.

npm install -g jupyter-dashboards-server

You can then run the dashboard server from the command line. See the next section about how to install and configure the other prerequisite components.

# shows a list of all nconf options
jupyter-dashboards-server --help

# runs the server pointing to a public kernel gateway
jupyter-dashboards-server --KERNEL_GATEWAY_URL=http://my.gateway.com/

# runs the server pointing to a kernel gateway that requires token auth
export KG_AUTH_TOKEN='somesecretinenvironment'
jupyter-dashboards-server --KERNEL_GATEWAY_URL=http://my.gateway.com/

Run It

The dashboard server is meant to enable the following workflow:

  1. Alice authors a notebook document using Jupyter Notebook.
  2. Alice adds a dashboard layout to her notebook using the jupyter_dashboards extension.
  3. Alice associates required frontend assets with her notebook.
  4. Alice one-click deploys her notebook and associated assets to a jupyter_dashboards_server using jupyter_dashboards_bundlers.
  5. Bob visits the dashboards server.
  6. Bob interacts with Alice's notebook as a dashboard.
  7. Alice updates her notebook and redeploys it to the dashboards server.

This workflow requires multiple components working in concert.

Minimal dashboard app deployment diagram

To bring all of these pieces together, you can start with the recipes in the jupyter-incubator/dashboards_setup repo. (We'll gladly take PRs that reduce the complexity of getting everything set up!)

Alternatively, you can clone this git repository and build the Docker images we use for development in order to run the demos in etc/notebooks. After setting up Docker (e.g. using docker-machine), run the following and then visit http://<your docker host ip>:3000.

make build
make examples
make demo-container

Develop It

To setup a development environment, install these minimum versions on your host machine.

  • Node 5.5.0
  • npm 3.5.3
  • gulp 3.9.0
  • Docker 1.9.1
  • Docker Machine 0.5.6

With these installed, you can use the make dev-* targets. Run make help to see the full gamut of targets and options. See the next few sections for the most common patterns.

Setup

# re-run if the Dockerfile.kernel changes
make kernel-gateway-image
# re-run if package.json changes
make dev-install
# run if you want to try the preliminary jupyter-incubator/declarativewidgets support
make examples

Dashboard Server w/ Auto Restart

# uses gulp:watch to restart on any changes
make dev
# mac shortcut for visiting URL in a browser
open http://127.0.0.1:3000

Dashboard Server w/ Auto Restart and Debug Console Logging

make dev-logging
# mac shortcut for visiting URL in a browser
open http://127.0.0.1:3000

Dashboard Server w/ Auto Restart and Remote Debugging

npm install -g node-inspector
make dev-debug
# a browser tab should open with the debugger visible
# refresh if it errors: the server might not be running yet

Dashboard Server w/ Auto Restart and Form Auth

make dev USERNAME=admin PASSWORD=password
# mac shortcut for visiting URL in a browser
open http://127.0.0.1:3000

See the Authentication wiki page for information about configuring alternative authentication mechanisms.

Dashboard Server w/ Auto Restart and Self-Signed HTTPS Certificate

make certs
make dev HTTPS_KEY_FILE=certs/server.pem HTTPS_CERT_FILE=certs/server.pem
# mac shortcut for visiting URL in a browser
open https://127.0.0.1:3001

Dashboard Server Tests

# unit tests
make test
# backend integration tests
make integration-test
# installation tests
make install-test

Detailed Developer Documentation

  • Server API - server endpoints
    • Also contains information about bundled dashboards (allowing specification of external resources).
  • Authentication - examples of integrating 3rd-party authentication strategies

About

Server that runs and renders Jupyter notebooks as interactive dashboards

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Jupyter Notebook 47.7%
  • JavaScript 44.2%
  • Makefile 3.0%
  • CSS 2.8%
  • HTML 2.2%
  • Shell 0.1%