Skip to content

Latest commit

 

History

History
147 lines (91 loc) · 5.65 KB

README.md

File metadata and controls

147 lines (91 loc) · 5.65 KB

jhsingle-native-proxy

Wrap an arbitrary webapp so it can be used in place of jupyter-singleuser in a JupyterHub setting.

Within JupyterHub this allows similar operation to jupyter-server-proxy except it also removes the Jupyter notebook itself, so is working directly with the arbitrary web service.

OAuth authentication is enforced based on JUPYTERHUB_* environment variables.

This project is used in ContainDS Dashboards, which is a user-friendly way to launch Jupyter notebooks as shareable dashboards inside JupyterHub. Also works with Streamlit and other visualization frameworks.

Install and Run

Install using pip.

pip install jhsingle-native-proxy

The process to start is specified on the command line, for example a streamlit web app:

jhsingle-native-proxy streamlit hello

By default the jhsingle-native-proxy server will listen on port 8888, forwarding to port 8500.

But you will normally need to tell jhsingle-native-proxy which port the end process will run in, and maybe tell the end process which port you want it to use (which you can do with the substitution variable {port}).

Note the use of -- to signal the end of command line options to jhsingle-native-proxy. Then the third party command line itself can contain options starting with dashes. An alternative is to use the substitution {--}

jhsingle-native-proxy -- streamlit hello --server.port {port} --server.headless True --server.enableCORS False

To run jhsingle-native-proxy itself listening on a different port use:

jhsingle-native-proxy --port 8000 streamlit hello

To run jhsingle-native-proxy on port 8000, and the end process on 8505:

jhsingle-native-proxy --port 8000 --destport 8505 -- streamlit hello --server.port {port} --server.headless True --server.enableCORS False

Use the JUPYTERHUB_SERVICE_PREFIX env var to specify the first part of the URL to listen to (and then strip before forwarding). E.g. JUPYTERHUB_SERVICE_PREFIX=/user/dan will mean requests on http://localhost:8888/user/dan/something will forward to http://localhost:8500/something

You can also specify --ip 0.0.0.0 for the address to listen on.

Below we use the substitution {--} for the command to run, allowing us to specify --ip to jhsingle-native-proxy instead of the command being run.

jhsingle-native-proxy --port 8000 --destport 8505 streamlit hello {--}server.port {port} {--}server.headless True {--}server.enableCORS False --ip 0.0.0.0 

Similarly, use e.g. {-}m to represent -m in the final command.

Voila example:

Running voila at the subfolder URL e.g. /user/dan/:

python -m jhsingle_native_proxy.main --destport 0 voila ./Presentation.ipynb {--}port={port} {--}no-browser {--}Voila.server_url=/ {--}Voila.base_url={base_url}/ {--}debug

'destport 0' above instructs jhsingle-native-proxy to choose a random free port on which to run the sub-process (Voila), and of course substitutes that as {port} in the Voila command line so it knows which port to listen on. destport 0 is the default anyway.

Or specify presentation_path as a substitution instead of hard-coding, which is sometimes easier in your wrapper code:

python -m jhsingle_native_proxy.main --destport 0 voila {presentation_path} {--}port={port} {--}no-browser {--}Voila.server_url=/ {--}Voila.base_url={base_url}/ {--}debug --presentation_path=./Presentation.ipynb

In addition, if presentation_path is provided, two further substitution variables are available: presentation_dirname and presentation_basename. These are computed using Python's os.path.dirname and os.path.basename functions on presentation_path.

Authentication

The above examples all assume OAuth will be enforced, as per the JUPYTERHUB_* env vars.

Alternatives can be specified via the authtype flag:

Same as default:

jhsingle-native-proxy --authtype=oauth streamlit hello

No auth required at all:

jhsingle-native-proxy --authtype=none streamlit hello

Specifying Authorized Users

The env vars JUPYTERHUB_USER and JUPYTERHUB_GROUP can be used, as typical for any JupyterHub single server, to specify user/groups of JupyterHub that should be allowed access via OAuth. There is an additional bespoke env var called JUPYTERHUB_ANYONE which can be set to 1 to allow any authenticated user access. (i.e. anyone who has an account on the JupyterHub)

Extra Arguments

--request-timeout=300 specifies the timeout in seconds that it waits for the underlying subprocess to return. Default is 300.

{origin_host} in the command argument will be replaced with the first 'host' seen in any request to the jhsingle-native-proxy server.

Changelog

v0.3.1 released 18 June 2020

  • Defaults presentation_path to empty str ('') if not supplied, avoiding error

v0.3.0 released 17 June 2020

  • presentation_path can be provided as a command line argument to become a substitution variable.
  • presentation_basename and presentation_dirname are also available when presentation_path is supplied.

v0.2.0 released 11 June 2020

  • Better websocket handling (subprotocols)
  • {origin_host} variable added

v0.1.3 released 1 June 2020

  • request-timeout added to the proxy call, and the default set to 300 (20 seconds was the httpclient's default previously)

v0.1.2 released 29 May 2020

  • Now allows single-dash placeholder, e.g. {-}m translates to -m in the final subprocess command.

Development install

git clone https://github.com/ideonate/jhsingle-native-proxy.git
cd jhsingle-native-proxy

pip install -e .

To run directly in python: python -m jhsingle_native_proxy.main <rest of command line>