This package provides an R library to instrument prediction code that lets you capture inputs to the model, predictions, prediction properties, and other metadata.

Setup

• Install R
• You can use R dependency management to your projects to manage local r environments
• Use below commands to get started

make deps
make test

• If you are able to run tests means your local environment is properly setup

Important make rules

Here are the available make rules that will help you in easing your development work

make all                 -> run check and clean
make clean               -> Remove intermediate files
make lint                -> Run lint
make test                -> Run test
make deps                -> Install dev dependencies
make install             -> Install package
make docs                -> Generate docs
make coverage            -> Run coverage
make check               -> Build as cran and run checks
make build               -> Run build

Releasing a package

• For releasing and packaging related help please refer to CRAN Releasing a package
• You need to run the below command from the RStudio console to push the release

devtools::release()

• The release process will prompt necessary checks, please read carefully and answer to continue

How to create new environment in Domino

• Create a personal access token
• Make sure you enable cerebrotech SSO for the pat that you have generated by following the steps mentioned in the above link
• Use the above generated personal access token in the following commands

RUN R --no-save -e "install.packages(c('devtools'))"

RUN R --no-save -e "devtools::install_github('cerebrotech/r-prediction-logging', auth_token = '<github pat>')"

How to use

library("DominoPredictionLogging")
prediction_client <- PredictionClient(feature_names=c("min","max"),predict_names=c("prediction"))
predictionClient$record(c(1,100), c("2"))

Example

# This is a sample R model
# You can publish a model API by clicking on "Publish" and selecting
# "Model APIs" in your quick-start project.
 
# Load dependencies
library("jsonlite")
library("DominoPredictionLogging")
prediction_client <- PredictionClient(
    feature_names=c("min","max"),
    predict_names=c("prediction")
)
 
# Define a function to create an API
# To call model use: {"data": {"min": 1, "max": 100}}
my_model <- function(min, max) {
  random_number <- runif(1, min, max)
  predictionClient$record(c(min,max), c(random_number))
  return(list(number=random_number))
}