Show HN: Sensus – Constructive PR Comments

Show HN: Sensus – Constructive PR Comments(chrome.google.com)

7 points by sneas 2 years ago | 5 comments

Being rude and dismissive in code reviews is as destructive as shouting at your kid — you both will suffer.

We all have good intentions, but even the best have bad days. We all can be grumpy, impatient, or insecure.

“Am I helping to resolve this PR?” is a valuable question to ask yourself before posting a comment.

My friend and I came up with Sensus - a tiny AI helper that asks this question and provides an answer as a score between 1 and 5. It incorporates constructiveness, politeness, and agreeableness into the calculations.

Sensus is free. It is available as a Chrome extension that works in GitHub comments.

At this stage, we are figuring out how many people may find this extension worth using.

If we see enough interest in this project, we will add it to other platforms.

Feel free to leave us any feedback in the comments.

phgn 2 years ago |

So it's like Grammarly but for the tone of your message? Why stop at code review comments -- I would love to know how my Slack messages, emails, etc sound to other people.

Also I wish some colleagues would install this...

bradknowles 2 years ago | |

And I wouldn't use it for the same reason I don't use Grammarly --I don't want all the text I type being sent to some third party server for analysis and storage and whatever else they want to do with that text.

sneas 2 years ago | |

It can be helpful in any text input. We decided to validate the idea on GH for the starter because most users can try it as a Chrome extension. If the idea gets some traction, it can be extended further.

paulrchds 2 years ago |

Seems interesting. What language model are you using? Are what info are you sending to the model? Privacy is obvs a concerns with this type of thing.

sneas 2 years ago | |

It is a ChatGPT wrapper behind AWS Lambda. We use gpt-3.5-turbo model.

The extension sends the text of the commit message to Lambda. Lambda asks ChatGPT to grade the text for politeness, constructiveness, and agreeableness. The score is the average of those 3.