This package provides both a CLI (command-line interface) and a library for interacting
with and deploying to RStudio Connect. The library is also used by the
rsconnect-jupyter
package to deploy
Jupyter notebooks via the Jupyter web console. Many types of content supported by RStudio
Connect may be deployed by this package, including WSGI-style APIs, Dash, Streamlit, and
Bokeh applications.
Content types not directly supported by the CLI may also be deployed if they include a
prepared manifest.json
file. See "Deploying R or Other
Content" for details.
RStudio Connect supports the deployment of Jupyter notebooks, Python APIs (such as
flask
-based) and apps (such as Dash, Streamlit, and Bokeh apps). Much like deploying R
content to RStudio Connect, there are some caveats to understand when replicating your
environment on the RStudio Connect server:
RStudio Connect insists on matching <MAJOR.MINOR>
versions of Python. For example,
a server with only Python 3.8 installed will fail to match content deployed with
Python 3.7. Your administrator may also enable exact Python version matching which
will be stricter and require matching major, minor, and patch versions. For more
information see the RStudio Connect Admin Guide chapter titled Python Version
Matching.
To install rsconnect-python
from PYPI, you may use any python package manager such as
pip:
pip install rsconnect-python
You may also build and install a wheel directly from a repository clone:
git clone https://github.com/rstudio/rsconnect-python.git
cd rsconnect-python
pip install pipenv
make deps dist
pip install ./dist/rsconnect_python-*.whl
Here's an example command that deploys a Jupyter notebook to RStudio Connect.
rsconnect deploy notebook \
--server https://connect.example.org:3939 \
--api-key my-api-key \
my-notebook.ipynb
Note: The examples here use long command line options, but there are short options (
-s
,-k
, etc.) available also. Runrsconnect deploy notebook --help
for details.
If you would like to use your shell's tab completion support with the rsconnect
command, use the command below for the shell you are using.
If you are using the bash
shell, use this to enable tab completion.
#~/.bashrc
eval "$(_RSCONNECT_COMPLETE=source rsconnect)"
If you are using the zsh
shell, use this to enable tab completion.
#~/.zshrc
eval "$(_RSCONNECT_COMPLETE=source_zsh rsconnect)"
If you get command not found: compdef
, you need to add the following lines to your
.zshrc
before the completion setup:
#~/.zshrc
autoload -Uz compinit
compinit
The information used by the rsconnect
command to communicate with an RStudio Connect
server can be tedious to repeat on every command. To help, the CLI supports the idea
of saving this information, making it usable by a simple nickname.
Important: One item of information saved is the API key used to authenticate with RStudio Connect. Although the file where this information is saved is marked as accessible by the owner only, it's important to remember that the key is present in the file as plain text so care must be taken to prevent any unauthorized access to the server information file.
Usually, an RStudio Connect server will be set up to be accessed in a secure manner,
using the https
protocol rather than simple http
. If RStudio Connect is set up
with a self-signed certificate, you will need to include the --insecure
flag on
all commands. If RStudio Connect is set up to require a client-side certificate chain,
you will need to include the --cacert
option that points to your certificate
authority (CA) trusted certificates file. Both of these options can be saved along
with the URL and API Key for a server.
Note: When certificate information is saved for the server, the specified file is read and its contents are saved under the server's nickname. If the CA file's contents are ever changed, you will need to add the server information again.
See the Network Options section for more details about these options.
Use the add
command to store information about an RStudio Connect server:
rsconnect add \
--api-key my-api-key \
--server https://connect.example.org:3939 \
--name myserver
Note: The
rsconnect
CLI will verify that the serve URL and API key are valid. If either is found not to be, no information will be saved.
If any of the access information for the server changes, simply rerun the
add
command with the new information and it will replace the original
information.
Once the server's information is saved, you can refer to it by its nickname:
rsconnect deploy notebook --name myserver my-notebook.ipynb
If there is information for only one server saved, this will work too:
rsconnect deploy notebook my-notebook.ipynb
You can see the list of saved server information with:
rsconnect list
You can remove information about a server with:
rsconnect remove --name myserver
Removing may be done by its nickname (--name
) or URL (--server
).
You can verify that a URL refers to a running instance of RStudio Connect by using
the details
command:
rsconnect details --server https://connect.example.org:3939
In this form, rsconnect
will only tell you whether the URL given does, in fact, refer
to a running RStudio Connect instance. If you include a valid API key:
rsconnect details --server https://connect.example.org:3939 --api-key my-api-key
the tool will provide the version of RStudio Connect (if the server is configured to divulge that information) and environmental information including versions of Python that are installed on the server.
You can also use nicknames with the details
command if you want to verify that the
stored information is still valid.
There are a variety of options available to you when deploying a Jupyter notebook to RStudio Connect.
You can include extra files in the deployment bundle to make them available when your notebook is run by the RStudio Connect server. Just specify them on the command line after the notebook file:
rsconnect deploy notebook my-notebook.ipynb data.csv
If a requirements.txt
file exists in the same directory as the notebook file, it will
be included in the bundle. It must specify the package dependencies needed to execute
the notebook. RStudio Connect will reconstruct the Python environment using the
specified package list.
If there is no requirements.txt
file or the --force-generate
option is specified,
the package dependencies will be determined from the current Python environment, or
from an alternative Python executable specified via the --python
option or via the
RETICULATE_PYTHON
environment variable:
rsconnect deploy notebook --python /path/to/python my-notebook.ipynb
You can see the packages list that will be included by running pip list --format=freeze
yourself,
ensuring that you use the same Python that you use to run your Jupyter Notebook:
/path/to/python -m pip list --format=freeze
By default, rsconnect
deploys the original notebook with all its source code. This
enables the RStudio Connect server to re-run the notebook upon request or on a schedule.
If you just want to publish an HTML snapshot of the notebook, you can use the --static
option. This will cause rsconnect
to execute your notebook locally to produce the HTML
file, then publish the HTML file to the RStudio Connect server:
rsconnect deploy notebook --static my-notebook.ipynb
You can create a manifest.json
file for a Jupyter Notebook, then use that manifest
in a later deployment. Use the write-manifest
command to do this.
The write-manifest
command will also create a requirements.txt
file, if it does
not already exist or the --force-generate
option is specified. It will contain the
package dependencies from the current Python environment, or from an alternative
Python executable specified in the --python
option or via the RETICULATE_PYTHON
environment variable.
Here is an example of the write-manifest
command:
rsconnect write-manifest notebook my-notebook.ipynb
Note: Manifests for static (pre-rendered) notebooks cannot be created.
There are a variety of options available to you when deploying a Python WSGI-style API,
Dash, Streamlit, or Bokeh application. All options below apply equally to api
,
dash
, streamlit
, and bokeh
sub-commands.
You can include extra files in the deployment bundle to make them available when your API or application is run by the RStudio Connect server. Just specify them on the command line after the API or application directory:
rsconnect deploy api flask-api/ data.csv
Since deploying an API or application starts at a directory level, there will be times
when some files under that directory subtree should not be included in the deployment
or manifest. Use the --exclude
option to specify files to exclude. An exclusion may
be a glob pattern and the --exclude
option may be repeated.
rsconnect deploy dash --exclude "workfiles/*" dash-app/ data.csv
You should always quote a glob pattern so that it will be passed to rsconnect
as-is
instead of letting the shell expand it. If a file is specifically listed as an extra
file that also matches an exclusion pattern, the file will still be included in the
deployment (i.e., extra files take precedence).
If a requirements.txt
file exists in the API/application directory, it will be
included in the bundle. It must specify the package dependencies needed to execute
the API or application. RStudio Connect will reconstruct the Python environment using
the specified package list.
If there is no requirements.txt
file or the --force-generate
option is specified,
the package dependencies will be determined from the current Python environment, or
from an alternative Python executable specified via the --python
option or via the
RETICULATE_PYTHON
environment variable:
rsconnect deploy api --python /path/to/python my-api/
You can see the packages list that will be included by running pip list --format=freeze
yourself,
ensuring that you use the same Python that you use to run your API or application:
/path/to/python -m pip list --format=freeze
You can create a manifest.json
file for an API or application, then use that
manifest in a later deployment. Use the write-manifest
command to do this.
The write-manifest
command will also create a requirements.txt
file, if it does
not already exist or the --force-generate
option is specified. It will contain
the package dependencies from the current Python environment, or from an alternative
Python executable specified in the --python
option or via the RETICULATE_PYTHON
environment variable.
Here is an example of the write-manifest
command:
rsconnect write-manifest api my-api/
You can deploy other content that has an existing RStudio Connect manifest.json
file. For example, if you download and unpack a source bundle from RStudio Connect,
you can deploy the resulting directory. The options are similar to notebook or
API/application deployment; see rsconnect deploy manifest --help
for details.
Here is an example of the deploy manifest
command:
rsconnect deploy manifest /path/to/manifest.json
Note: In this case, the existing content is deployed as-is. Python environment inspection and notebook pre-rendering, if needed, are assumed to be done already and represented in the manifest.
The argument to deploy manifest
may also be a directory so long as that directory
contains a manifest.json
file.
If you have R content but don't have a manifest.json
file, you can use the RStudio
IDE to create the manifest. See the help for the rsconnect::writeManifest
R function:
install.packages('rsconnect')
library(rsconnect)
?rsconnect::writeManifest
These options apply to any type of content deployment.
The title of the deployed content is, by default, derived from the filename. For
example, if you deploy my-notebook.ipynb
, the title will be my-notebook
. To change
this, use the --title
option:
rsconnect deploy notebook --title "My Notebook" my-notebook.ipynb
When using rsconnect deploy api
, rsconnect deploy dash
, rsconnect deploy streamlit
, or rsconnect deploy bokeh
, the title is derived from the directory
containing the API or application.
When using rsconnect deploy manifest
, the title is derived from the primary
filename referenced in the manifest.
When specifying information that rsconnect
needs to be able to interact with RStudio
Connect, you can tailor how transport layer security is performed.
RStudio Connect servers can be configured to use TLS/SSL. If your server's certificate
is trusted by your Jupyter Notebook server, API client or user's browser, then you
don't need to do anything special. You can test this out with the details
command:
rsconnect details --api-key my-api-key --server https://connect.example.org:3939
If this fails with a TLS Certificate Validation error, then you have two options.
-
Provide the Root CA certificate that is at the root of the signing chain for your RStudio Connect server. This will enable
rsconnect
to securely validate the server's TLS certificate.rsconnect details \ --api-key my-api-key \ --server https://connect.example.org:3939 \ --cacert /path/to/certificate.pem
-
RStudio Connect is in "insecure mode". This disables TLS certificate verification, which results in a less secure connection.
rsconnect add \ --api-key my-api-key \ --server https://connect.example.org:3939 \ --insecure
Once you work out the combination of options that allow you to successfully work with
an instance of RStudio Connect, you'll probably want to use the add
command to have
rsconnect
remember those options and allow you to just use a nickname.
If you deploy a file again to the same server, rsconnect
will update the previous
deployment. This means that you can keep running rsconnect deploy notebook my-notebook.ipynb
as you develop new versions of your notebook. The same applies to other Python content
types.
To bypass this behavior and force a new deployment, use the --new
option:
rsconnect deploy dash --new my-app/
If you want to update an existing deployment but don't have the saved deployment data, you can provide the app's numeric ID or GUID on the command line:
rsconnect deploy notebook --app-id 123456 my-notebook.ipynb
You must be the owner of the target deployment, or a collaborator with permission to change the content. The type of content (static notebook, notebook with source code, API, or application) must match the existing deployment.
Note: There is no confirmation required to update a deployment. If you do so accidentally, use the "Source Versions" dialog in the RStudio Connect dashboard to activate the previous version and remove the erroneous one.
The App ID associated with a piece of content you have previously deployed from the
rsconnect
command line interface can be found easily by querying the deployment
information using the info
command. For more information, see the
Showing the Deployment Information section.
If the content was deployed elsewhere or info
does not return the correct App ID,
but you can open the content on RStudio Connect, find the content and open it in a
browser. The URL in your browser's location bar will contain #/apps/NNN
where NNN
is your App ID. The GUID identifier for the app may be found on the Info tab for
the content in the RStudio Connect UI.
You can see the information that the rsconnect
command has saved for the most recent
deployment with the info
command:
rsconnect info my-notebook.ipynb
If you have deployed to multiple servers, the most recent deployment information for each server will be shown. This command also displays the path to the file where the deployment data is stored.
Stored information files are stored in a platform-specific directory:
Platform | Location |
---|---|
Mac | $HOME/Library/Application Support/rsconnect-python/ |
Linux | $HOME/.rsconnect-python/ or $XDG_CONFIG_HOME/rsconnect-python/ |
Windows | $APPDATA/rsconnect-python |
Remembered server information is stored in the servers.json
file in that directory.
After a deployment is completed, information about the deployment is saved
to enable later redeployment. This data is stored alongside the deployed file,
in an rsconnect-python
subdirectory, if possible. If that location is not writable
during deployment, then the deployment data will be stored in the global configuration
directory specified above.
rsconnect-python {{ rsconnect_python.version }}