Mozilla has been experimenting with the usage of machine learning (in particular, large language models) to aid Firefox developers with code review.

This talk will explain the architecture of the software, and the path that led there, including iterative improvements to the prompts, retrieval augmented generation, integration of pre-existing tools such as code search.

In addition, we will delve into the results of the initial experiments, our plans for the future of the tool, and the potential for its adoption in other open source projects.

The project source code is part of the bugbug repository, which is a platform for Machine Learning projects on Software Engineering used by Mozilla for various tasks (for example, bug triage and test selection).