Using the Bokeh Server¶

Overview¶

The Bokeh server is an optional component that can be used to provide additional capabilities, such as:

• hosting and publishing Bokeh plots for wider audiences
• streaming data to plots so that they automatically update
• interactively visualizing very large datasets by employing downsampling and abstract rendering
• building and deploying dashboards and apps with sophisticated user interactions

The Bokeh server is built on top of Flask. Bokeh ships with a standalone executable bokeh-server that you can easily run, and you can also embed the Bokeh server functionality inside another Flask server using the Bokeh Server Flask Blueprint.

The basic task of the Bokeh Server is to be a mediator between the original data and plot models created by a user, and the reflected data and plot models in the BokehJS client:

Here you can see illustrated the most useful and compelling of the Bokeh server: full two-way communication between the original code and the BokehJS plot. Plots are published by sending them to the server. The data for the plot can be updated on the server, and the client will respond and update the plot. Users can interact with the plot through tools and widgets in the browser, then the results of these interactions can be pulled back to the original code to inform some further query or analysis (possibly resulting in updates pushed back the plot).

We will explore the capabilities afforded by the Bokeh server in detail below.

Authentication¶

Bokeh defaults to single user mode - where every request is automatically logged in as a user with username defaultuser. However for teams, and for plot publishing, it makes more sense to add an authentication layer. This way users won’t be able to overwrite each other’s plots.

To do enable multi user mode, You need to turn on the multi_user bokeh server setting. This can be done using the command line parameter -m. Once this is done, all scripts that use the bokeh server must authenticate with the bokeh server. A user can be created with:

session = Session(root_url=url)
session.register(username, password)

Or logged in with:

session = Session(root_url=url)
session.login(username, password)

The bokeh client library will store authentication keys in the ~/.bokeh directory, so that in subsequent invocations, logging in is not necessary

Plot Hosting¶

For plot hosting, we generally recommend running the server in multi user mode. See the Authentication section for instructions on how to do this.

Once a plot is created, you have to publish it. Assuming you have a session that is authenticated against a document:

session = Session()
session.use_doc('mydocument')
#do some plots here
session.publish()

Or if you are using output_server:

output_server('myplot')
#make some plots
cursession().publish()

A public link to a plot on the bokeh server page can be viewed by append ?public=true To the url - for example if you have the url to a plot http://localhost:5006/bokeh/doc/some-doc-id/some-plot-id, You can generate a public link to the published plot using http://localhost:5006/bokeh/doc/some-doc-id/some-plot-id?public=true. In addition, the autoload_server function call in bokeh.embed also takes a public=true keyword argument, which will generate an embeddable html snippet that will load the public version of a given plot

Streaming Data¶

Large Data¶

Widgets and Dashboards¶

Command Line Configuration¶

$ bokeh-server --backend=memory