rsconnect-jupyter is a plugin for Jupyter Notebook that enables publishing notebooks to RStudio Connect.
- Python 2.7.9 or Python 3.5.0 and higher
- Jupyter Notebook 5.x
- pip
- wheel
- RStudio Connect v1.7.0 or higher, configured with Python support.
If using conda
, pip
and wheel
should already be installed.
If you are installing rsconnect-jupyter
for use in Jupyterhub, please see the
Jupyterhub section below.
We recommend working within a virtualenv
. If you
are unfamiliar, these commands create and activate a virtualenv
at /my/path
:
pip install virtualenv
virtualenv /my/path
source /my/path/bin/activate
Install Jupyter inside the virtualenv
:
pip install jupyter
Note: be sure to run Jupyter from the virtual environment, not from a global installation.
Install the rsconnect-jupyter
package with the following command:
pip install rsconnect_jupyter
Enable the rsconnect-jupyter
extension with the following commands:
# Install `rsconnect-jupyter` as a jupyter extension
jupyter-nbextension install --sys-prefix --py rsconnect_jupyter
# Enable JavaScript extension
jupyter-nbextension enable --sys-prefix --py rsconnect_jupyter
# Enable Python extension
jupyter-serverextension enable --sys-prefix --py rsconnect_jupyter
Note: The above commands only need to be run once when installing
rsconnect_jupyter
.
Note: If you run into an issue during installation please let us know by filing a bug here.
First disable and remove the rsconnect-jupyter
notebook extension:
# Disable Python extensions found in `rsconnect-jupyter`
jupyter-serverextension disable --sys-prefix --py rsconnect_jupyter
# Remove JavaScript extension
jupyter-nbextension uninstall --sys-prefix --py rsconnect_jupyter
Finally, uninstall the rsconnect-jupyter
python package:
pip uninstall rsconnect_jupyter
To upgrade rsconnect-jupyter
, first uninstall the extension and then
re-install it.
Open a notebook and click the blue "Publish to RStudio Connect" icon to publish the current notebook to RStudio Connect.
If this is your first time publishing a notebook, you will be prompted to enter the location and a nickname for the RStudio Connect server.
You will also be prompted to enter your API Key. See the RStudio Connect User Guide for instructions on generating API Keys for your user.
When you click the Add Server button, rsconnect-jupyter
will send a request to the RStudio Connect server to verify that it can be reached via the requested URL and that the API key is valid.
If your RStudio Connect server was configured with a self-signed certificate (or other certificate that computer hosting your Jupyter notebook server does not trust), the attempt to contact RStudio Connect will fail with a TLS-related error. In this case, please contact your system administrator for instructions on how to configure your computer to trust the certificate they have provided. If trusting the certificate is not an option, you can check the Disable TLS Certificate Verification
box, which will bypass the certificate check. Your connection to the server will still be encrypted, though the server identity will not be verified.
There are two different publication modes. Selecting "Publish finished document only" will publish an HTML snapshot of the notebook to RStudio Connect. HTML snapshots are static and cannot be scheduled or re-run on the RStudio Connect server.
If you select "Publish document with source code", the notebook file and a list of the Python packages installed in your environment will be sent to RStudio Connect. This enables RStudio Connect to recreate the environment and re-run the notebook at a later time.
The list of packages sent along with the notebook comes from the python
environment where the notebook kernel is running. In order for environment
inspection to work, the rsconnect-jupyter
package must be installed in the
kernel environment; that is, the environment where the ipykernel
package is
installed. In most cases that will be the same as the notebook server
environment where jupyter
is installed.
The command pip freeze
will be used to inspect the environment. The output
of pip freeze
lists all packages currently installed, as well as their
versions, which enables RStudio Connect to recreate the same environment.
If content that matches your notebook's title is found on RStudio Connect, you may choose to overwrite the existing content or create new content.
Choosing "New location" will create a new document in RStudio Connect. You can choose either publication mode - an HTML snapshot or a document with source code.
Updating an existing document will not change its publication mode.
Upon successful publishing of the document a notification will be shown in toolbar. Clicking the notification will open the published document in the RStudio Connect server you selected in the previous dialog.
To collaborate with others add them as collaborators in RStudio Connect. During publishing they should provide their API key and will be able to choose a content location to publish to if the notebook title is the same.
You may share notebooks if appropriate.
In JupyterHub, follow the directions above to install the
rsconnect-jupyter
package into the Python environment where the Jupyter
notebook server and kernel are installed. Typically those will be the same
environment. If you've configured separate kernel environments, install the
rsconnect-jupyter
package in the notebook server environment as well as each
kernel environment.
The exact install location depends on your Jupyterhub configuration.
This section presents a simple working example of a Jupyterhub configuration
with rsconnect-jupyter
installed.
This example uses Docker, but you can install the rsconnect-jupyter
package in
any Jupyterhub installation. Docker is not required.
Example Dockerfile:
FROM jupyterhub/jupyterhub:0.9.4
# Install Jupyter notebook into the existing base conda environment
RUN conda install notebook
# Download and install rsconnect-jupyter in the same environment
# Update this to specify the desired version of the rsconnect-jupyter package,
# or pass `--build-arg VERSION=...` to docker build.
ARG VERSION=RSCONNECT_VERSION
ARG REPOSITORY=https://s3.amazonaws.com/rstudio-rsconnect-jupyter
RUN wget ${REPOSITORY}/rsconnect_jupyter-${VERSION}-py2.py3-none-any.whl
RUN pip install rsconnect_jupyter-${VERSION}-py2.py3-none-any.whl && \
jupyter-nbextension install --sys-prefix --py rsconnect_jupyter && \
jupyter-nbextension enable --sys-prefix --py rsconnect_jupyter && \
jupyter-serverextension enable --sys-prefix --py rsconnect_jupyter
# create test users
RUN useradd -m -s /bin/bash user1 && \
useradd -m -s /bin/bash user2 && \
useradd -m -s /bin/bash user3 && \
bash -c 'echo -en "password\npassword" | passwd user1' && \
bash -c 'echo -en "password\npassword" | passwd user2' && \
bash -c 'echo -en "password\npassword" | passwd user3'
CMD ["jupyterhub"]
Run these commands to build and start the container:
docker build -t jupyterhub:rsconnect-jupyter .
docker run --rm -p 8000:8000 --name jupyterhub jupyterhub:rsconnect-jupyter
Connect to Jupyterhub on http://localhost:8000 and log in as one of the test users. From there, you can create a notebook and publish it to RStudio Connect. Note that the current Jupyterhub docker image uses Python 3.6.5, so you will need a compatible Python version installed on your RStudio Connect server.