app.R

# API app -----------------------------------------------------------------

# Shiny App for providing data input to cars plumber API
library(shiny)
library(httr)

base_url <- config::get("base_url")

ui <- fluidPage(

    # Application title
    titlePanel("Cars MPG Predictor"),

    # Sidebar with a slider input for number of bins 
    sidebarLayout(
        sidebarPanel(
            sliderInput("hp",
                        "Horsepower",
                        min = min(mtcars$hp),
                        max = max(mtcars$hp),
                        value = median(mtcars$hp)),
            selectInput("cyl",
                        "Cylinder",
                        choices = sort(unique(mtcars$cyl)),
                        selected = sort(unique(mtcars$cyl))[1]),
            fluidRow(
              actionButton("add",
                           "Add"),
              actionButton("remove",
                           "Remove"),
              actionButton("predict",
                           "Predict")
            )
        ),

        # Show a plot of the generated distribution
        mainPanel(
          tableOutput("data"),
          wellPanel(
            textOutput("raw_results")
          )
        )
    )
)

server <- function(input, output) {
  # Create reactive_values
  reactive_values <- reactiveValues(data = data.frame(),
                                    predicted_values = NULL)

  # Update user data
  observeEvent(input$add, {
    # Reset predicted_values
    reactive_values$predicted_values <- NULL
    
    # Add to data
    data <- reactive_values$data
    # Remove predicted column if present
    reactive_values$data <- rbind(data[!names(data) %in% "predicted_mpg"],
                                  data.frame(hp = as.numeric(input$hp), cyl = as.numeric(input$cyl)))
  })

  observeEvent(input$remove, {
    # Reset predicted_values
    reactive_values$predicted_values <- NULL
    
    # Set aside existing data
    data <- reactive_values$data
    
    # Remove rows that match current input
    reactive_values$data <- dplyr::anti_join(data[!names(data) %in% "predicted_mpg"],
                                             data.frame(hp = as.numeric(input$hp), cyl = as.numeric(input$cyl)))
  })

  observeEvent(input$predict, {
    # Make API request
    api_res <- httr::POST(url = paste0(base_url, "/predict"),
                          body = reactive_values$data,
                          encode = "json")
    
    # Extract JSON from API response
    reactive_values$predicted_values <- httr::content(api_res, as = "text", encoding = "UTF-8")
    
    # Add predicted values to data
    if (!"predicted_mpg" %in% names(reactive_values$data)) {
      reactive_values$data <- cbind(reactive_values$data, 
                                    predicted_mpg = as.numeric(jsonlite::fromJSON(reactive_values$predicted_values)))
    }
  })
  
  output$data <- renderTable(reactive_values$data)
  output$raw_results <- renderPrint({
    if (is.null(reactive_values$predicted_values)) {
      "No predictions"
    } else {
      jsonlite::prettify(reactive_values$predicted_values)
    }
  })
}

shinyApp(ui = ui, server = server)