coderabbitai / ai-pr-reviewer

AI-based Pull Request Summarizer and Reviewer with Chat Capabilities.
https://coderabbit.ai
MIT License
1.41k stars 256 forks source link

PR comments support #495

Open Tradunsky opened 8 months ago

Tradunsky commented 8 months ago

I like the idea of AI Agent supported code review! Giving a try to code rabbit ai now.

Generally want to reduce time when code review is triggered and focus it only on times, when humans also asked for code review.

One of the use cases highlighted is sometimes users create a PR and iteratively keep adding to it until it is fully ready for code review and even after code review changing code as well.

In that use case, the preference is to have AI code review just once. No need for summaries, release notes, nothing except code review and non trivial suggestions.

Right now using PR level comment trigger the action is skipped with the message:

Warning: Skipped: this action only works on push events or pull_request

Would it be possible add the support for PR level comments that would simply scan all changes in PR?

Here is the full trigger of the workflow we have:

on:
  issue_comment:
    types: [created]
  pull_request_review_comment:
    types: [created]
guritfaq commented 7 months ago

CodeRabbit Pro supports on-demand reviews but this doesn’t exist in OSS yet. In Pro, automatic reviews can be disabled and reviews can be triggered manually using review commands.

Tradunsky commented 7 months ago

Thank you for the explanation @guritfaq 🙏🏻

The review commands in code place do work as expected. However, when I run a review via PR level comment the Github Action application prevents this from being triggered. Even though the Pro version might indeed be more comprehensive, do you think it is still possible to let the open source variant to be triggered when it is best for community users?

guritfaq commented 7 months ago

@Tradunsky We are a small team and have to prioritize requests for our Pro version at this stage. However, we will add this to our backlog and address it as soon as possible.