R Plumber API on Heroku

Deploy a R Plumber API on Heroku

Deploying and self-hosting a R API using Plumber is difficult. I’ve been trying to use DigitalOcean and Docker services, but have had trouble with uptime. I recently came across a blog article by Magnus Furugard and realized Heroku could be an ideal solution. In short, Heroku has been a life saver! My goal in this post is to walk you through steps to implement a R Plumber API yourself.

R and Plumber

The first step is to have a working R Plumber API. Using RStudio, the process is trivial. Create a Plumber API project and set up your files as follows.

  |- init.R
  |- app.R
  |- plumber.R

The init.R file is used to install needed packages.

my_packages = c("plumber","randomForestSRC")
install_if_missing = function(p) {
  if (p %in% rownames(installed.packages()) == FALSE) {
    install.packages(p, dependencies = TRUE)
  else {
    cat(paste("Skipping already installed package:", p, "\n"))
invisible(sapply(my_packages, install_if_missing))

The run.R file is what’s used to establish the API. The critical lines are the host needs to be and port is set as the main Heroku port.

port <- Sys.getenv('PORT')
server <- plumb("plumber.R")
	host = '',
	port = as.numeric(port),

Finally, the plumber.R file is the main API file. Here is where you establish the API endpoints specific to your application. Since the goal of this application was to predict heart failure risk, I used a /predict endpoint. Of note, since the API will be accessed from various websites and applications, CORS is required.

# Load model
rf_model <- readRDS("model.Rds")
#* @filter cors
cors <- function(req, res) {
  res$setHeader("Access-Control-Allow-Origin", "*")
  if (req$REQUEST_METHOD == "OPTIONS") {
    res$setHeader("Access-Control-Allow-Headers", req$HTTP_ACCESS_CONTROL_REQUEST_HEADERS)
    res$status <- 200
  } else {
#* Submit data and get a prediction in return
#* @post /predict
function(req, res) {
  data <- tryCatch(jsonlite::parse_json(req$postBody, simplifyVector = TRUE),
                   error = function(e) NULL)
  if (is.null(data)) {
    res$status <- 400
    list(error = "No data submitted")
  data <- data.frame(data)
  ret <- predict(rf_model, data)
  ret <- ret$chf[,length(ret$time.interest)]


First, make sure the Heroku CLI is installed on your machine. Next, make sure the Heroku R Buildpack is specified in the application settings.

Once applied, navigate to the root of your application folder. Finally, push the application files to the stack and enjoy!

git init
git add .
git commit -m 'initial'
git push heroku master