Air is an R package that lets you ask R related questions to OpenAI and get working code, or explanations of code.
You can ask “how-to” questions to get working code solutions:
howto("get a vector of the second element from a list of vectors")
# sapply(list_of_vectors, function(v) v[2])
howto("extract the second largest number from a vector")
# vec <- c(1, 5, 3, 9, 7)
# sort(unique(vec), decreasing = TRUE)[2]You can ask “what-is” questions to get detailed explanations of code you cannot understand:
whatis("paste0(vector1, vector2)")
# [Abbreviated output]
# Overall, the R function `paste0` concatenates vectors after
# converting them to characters. It combines corresponding
# elements of `vector1` and `vector2` without any separator.
#
# Sub-expressions:
# - `paste0`: The function used for concatenation without any separator.
#
# Example:
#
# ```R
# vector1 <- c("Apple", "Banana", "Cherry")
# vector2 <- c("Pie", "Bread", "Jam")
#
# result <- paste0(vector1, vector2)
# print(result)
# ```
#
# Output:
# ```
# [1] "ApplePie" "BananaBread" "CherryJam"
# ```Air requires your OpenAI API key (see Setup direcitons below), which it stores securely in your operating system’s keyring.
Future features under consideration:
tellme()).Feel free to leave issues or reach out to maintainers with any questions.
Air is currently only available from Github. Use your favorite installer tool as follows:
Using pak:
or, Using devtools:
Air needs your API key for OpenAI.
Create an OpenAI acount if you don’t have one at: https://platform.openai.com/signup
Ensure you have API credits by:
Please set your API credentials (key and model) either in your your R environment, or in your system’s secure keyring (suggested).
Your R environment can specify user level environment variables stored in a local .Renviron file. Please note that anyone with access to this file can steal your credentials – make sure not to expose it to any other processes.
Please read about .Renviron files if you are unfamiliar with them. Briefly, to set your OpenAI API credentials, we suggest using usethis::edit_r_environ() to open the correct .Renviron file for editing. You may create VAR=value pairs to set your API key and preferred model in this file - for example:
OPENAI_KEY="sk-my-api-key"
OPENAI_MODEL="gpt-4-1106-preview"You will have to restart the R session for these new environment variables to be loaded. You can confirm they are present in your R environment by using:
Alternatively, we recommend storing your key and model preference securely in your operating system’s keyring using Air’s built-in functions for accessing the keyring.
We recommend you use your system’s keyring to enter it safely in a system popup window. But you may choose to set it programatically at the console, but note that it will be stored in your .Rhistory file where it could get compromised:
# Safely set the key in a system popup
air::set_key()
# or programatically set it:
air::set_key("api-key-goes-here")You must also specify the OpenAI model you prefer to use:
# Set your preferred model
set_model("gpt-4-1106-preview")
# or use the default model (see documentation)
set_model()Note that your system may occasionally popup windows to get your login password to access the keyring for these credentials. We suggest using an ‘always accept’ option in such popups to reduce how often you see them.
You may later delete your key and model from your keyring using:
library(air)
howto("zip vectors foo and bar together, creating a list of vectors which are pairs of elements from the original two vectors")
# zip_list <- mapply(c, foo, bar, SIMPLIFY=FALSE)Please run tests using devtools::test()
Please first submit Github issues for bugs or feature suggestions, for discussion on whether and how to fix or implement them. We greatly welcome pull requests on open issues that are slated for development. We also welcome suggestions or fixes for documentation.