[RETIRED] Jupyter Dashboards Server
This project has been retired. See the proposal to move the project to jupyter-attic, announcement of the proposal on the mailing list, and Steering Council vote on the proposal PR for more information.
A NodeJS application that can display Jupyter notebooks as dynamic dashboards outside of the Jupyter Notebook server.
The Jupyter Incubator Dashboards effort covers:
- Arranging notebook outputs in a grid- or report-like layout
- Bundling notebooks and associated assets for deployment as dashboards
- 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 dashboard incubation effort.
What it Gives You
- Ability to run some Jupyter notebooks as standalone dashboard applications [1]
- Ability to navigate a list of multiple notebooks and select one to run as a dashboard
- Optional shared login to secure access to the dashboard server
- Ability to add custom authentication mechanisms using the Passport middleware for Node.js
- An API for POSTing notebooks to the server at runtime with optional
authentication (
/_api/notebooks
)
The qualification in [1] stems from the fact that supporting one-click deploy of notebooks with arbitrary JavaScript and kernel dependencies is a "Really Hard Problem." We've invested effort in getting these dashboard, visualization, and widget libraries working in the dashboard server.
- jupyter_dashboards 0.6.x
- jupyter_dashboards_bundlers 0.8.x
- ipywidgets 5.2.x
- jupyter_declarativewidgets 0.6.x
- matplotlib 1.5.x
- Bokeh 0.11.x
- Plotly 1.9.x
If you try another library and find that it does not work in the dashboard server, see the wiki page about Widget Support below for steps you might take to resolve the problem.
Install it
Install Node 5.x and npm 3.5.x. Use npm
to install the
jupyter-dashboards-server
package.
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 support the layout-bundler-deploy workflow described on the project overview page. This workflow requires multiple components working in concert.
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
Run Behind a Proxy
The dashboards server can be run behind a reverse proxy. In order to do so, you will need to set the following options as command line args or in environment vars.
TRUST_PROXY
- The simple option is to just set this totrue
. However, if you require further configuration on which requests to trust, this option can also take values as specified by the Express documentation.BASE_URL
- Specify the base URL (prefix) at which the dashboards server will run. The server supports two options here: passing the prefix along with the request or stripping the prefix off the request.
For example:
# allow proxying of "http://proxy_host/db/..." to "http://dashboards_host/db/..."
jupyter-dashboards-server --TRUST_PROXY=true --BASE_URL=/db --KERNEL_GATEWAY_URL=http://my.gateway.com/
# allow proxying of "http://proxy_host/db/..." to "http://dashboards_host/..."
jupyter-dashboards-server --TRUST_PROXY=true --BASE_URL='[/db]' --KERNEL_GATEWAY_URL=http://my.gateway.com/
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
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
Technical Details
See the wiki attached to this project for additional technical details including the server API, authentication plugins, adding support for new widgets, and more.