functai: Functions Whose Body a Language Model Writes, for the Tidyverse

Write a function's inputs, its answer's type and a sentence saying what it does; a language model writes the body. The functions are vectorised, so they work as columns in 'dplyr' verbs; the answer comes back as the type you declared (a factor, a number, a tibble of fields). The same functions are 'parsnip' models (ai_model()), so they fit, predict, resample and tune with 'tidymodels'. Measure how often a function is right with a confidence interval ('broom'-style tidy(), glance() and augment()), log every call, rate calls, improve functions with worked examples or rewritten instructions, and run functions saved in 'Python', 'TypeScript' or 'Julia'. Implements the same contract as the 'Python', 'TypeScript' and 'Julia' packages.

See also

Useful links:

• https://maximerivest.github.io/functai/r/manual/

• https://maximerivest.github.io/functai/r/index.html

• https://github.com/MaximeRivest/functai

• Report bugs at https://github.com/MaximeRivest/functai/issues

Author

Maintainer: Maxime Rivest maxime.rivest@gmail.com [copyright holder]

Authors:

• Maxime Rivest maxime.rivest@gmail.com [copyright holder]