Robin is your AI pull request reviewer in the form of a super fast lightweight github action.
Full Changelog: https://github.com/Integral-Healthcare/robin-ai-reviewer/compare/v1.6.1...v1.7.0
This change includes a prompt escape character fix which was causing formatting issues with Robin's responses, as well as a slimmed docker image and some better logging strategies. Huge thanks to a new contributor @RytisBa who spearheaded these fixes! 🥳 🎉
Thanks to our first external contributor @RockyNiu !! 🎉
This release includes an updated prompt to better follow the conventions of good prompt engineering as well as a format change to nest the details of the code review score in a dropdown.
Despite tons of tweaks and parameterization, the original prompt seems to perform the best.
Code base language is now being inferred and there are more explicit callouts with respect to the inputs and outputs to be given. yet to be tested extensively with gpt4 model
This was a regression in performances in most cases, so we're ripping it out for now.
The initial prompt that feeds into GPT's API is now parameterized! You can set:
These new variables are optional and so no action is necessary and the logic should behave the same. Consider trying to update the coding_principles to be something like "efficiency, code complexity, or edge case handling".
This updates the default model option to be the more stable 3.5-turbo
This update includes a simplified prompt and stricter instructions on the code block.
Introducing a new parameter files_to_ignore which is a space delimited string of file paths. Now you can skip over config and package file changes.