Hi all,

During last week’s biweekly community meeting, we briefly discussed
the idea of introducing AI-assisted code review. I did some follow-up
research and wanted to share findings and get more input from the
community.

>From what I can see, there are some constraints in Apache repositories
that may not align with our initial expectations. In particular, two
popular AI code review tools—GitHub Copilot and CodeRabbit (which is
free for open source)—currently cannot be enabled at the repository
level under ASF policies:

* CodeRabbit requires write access to the repository, which conflicts
with ASF’s rule that only committers have write permissions. Relevant
discussions:
 - https://issues.apache.org/jira/browse/INFRA-27676
 - https://issues.apache.org/jira/browse/INFRA-26837
* GitHub Copilot is not free in this context. Based on Infra discussions:
 - https://issues.apache.org/jira/browse/INFRA-26698
 - https://issues.apache.org/jira/browse/INFRA-27132

It seems unlikely that ASF Infra will enable this by default for now
due to budget constraints. However, individuals with Copilot Pro can
still use it when reviewing pull requests.

As a possible workaround, one option is to fetch pull requests
locally, run an AI code review tool on the changes in your local
environment, and then manually copy the suggestions or comments back
into the corresponding PR on GitHub. While not fully automated, this
could still provide value.

Would love to hear if others have explored similar approaches or if
there are alternative tools or workflows that better fit within ASF
constraints.

Looking forward to your thoughts.

Best,
Dianjin Wang

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to