A web app to input the parameters needed to run scenarios through the New Hospital Programme (NHP) demand model.
The app is deployed to Posit Connect. You must have an account and sufficient permissions to view it.
Results can then be viewed in the outputs app, which is generated from the nhp_outputs repository.
You can find more information on the NHP model project information site, including a diagram of how the components of the modelling process fit together.
The app is built and maintained by members of the Strategy Unit’s Data Science team.
Technically there are two apps: the main app in the main
branch, and
the inputs selection app (where users start or edit a scenario) in the
inputs_selection_app
branch. Users arrive at the selection app before
being routed to the main app.
Both apps are built with Shiny and the main
app uses the the {golem}
package. Server and UI modules
can be found in R/
, configuration in inst/golem-config.yml
and
supporting data and text in inst/app/
.
Packages used in the app are listed in DESCRIPTION
, and can be
installed with devtools::install_deps(dependencies = TRUE)
.
Run the app locally on your machine to test that your changes work as expected.
To prepare, add an .Renviron
file to the project root that contains
the required environment variables. There is an .Renviron.example
file
which describes the required variables. You can get these from a member
of the Data Science team.
Then, from the main
branch (or your development branch of it), use
Git’s worktree
function to put the inputs_selection_app
branch (or
your development branch of it) in its own subfolder.
In the terminal:
git fetch origin inputs_selection_app
git worktree add inputs_selection_app inputs_selection_app
If you want a development version of the inputs selection app, you can
specify the branch name as the second of the arguments to
git worktree add
.
To run the app from RStudio, start up the selection app by opening the
dev/watch.R
script, go to the ‘Background Jobs’ tab of the console
pane and click the ‘Start Background Job’ button. In the ‘Run Script as
Background Job’ dialog box select the project root as the ‘Working
Directory’, then hit ‘Start’. When ready, the app will tell you to visit
http://127.0.0.1:9081/
in your browser.
Note that your selections in the app remain local to you and the
resulting json file of parameters for your scenario will live in your
local params/development/
directory. These scenarios will be
selectable and editable in future from your locally-run inputs selection
app.
If you need to remove the selection-app folder at any point, you can terminate the background job (if running) and in the terminal run:
git worktree remove inputs_selection_app
During pre-release QA we test both apps on the server. This helps us spot any issues that are server-specific and might be overlooked if running the apps locally.
To test the unreleased app in our development environment, first ensure
you’ve (a) merged any PRs you want to test into the main
branch and,
if you’ve made any changes to the inputs selection app, (b) run the
manual deploy()
call under the ‘development’ heading in the deploy.R
script in your selection app development branch.
When using the dev inputs selection app on Connect, make sure to set the ‘Model Version’ dropdown to ‘dev’ in the expandable ‘Advanced Options’ box. That way you’ll be taken to the dev inputs app when you hit ‘Start’.
The app displays trust-specific data to users. The data is processed via Databricks scripts in the nhp_data repository and stored in Azure storage.
Deployment is controlled by GitHub Actions, where:
- pushes to the
main
branch redeploy the app to /nhp/dev/inputs/ for purposes of quality assurance - tagged releases trigger a new deployment to /nhp/vX-Y/inputs/, where ‘vX-Y’ is the current version (note the hyphen)