How Linux Foundation Used AI Code Reviews to reduce manual bottlenecks in OSS

At a Glance:

Vertical: Open Source Technology
Primary contact: David Deal, Director of Platforms
Coding languages used: Golang, Python, Angular, TypeScript
Challenge: Developers spend excessive time on error-prone manual code reviews.
Key Result: Freeing up 25% of developer time, letting them build more features.

The Linux Foundation is a global non-profit that supports various open-source projects that help customers build modern web applications by unlocking the power of shared technology. It hosts and supports over 200 open-source projects, including CNCF, PyTorch, OpenJS, and Automotive Grade Linux. With an engineering team of 40-60 engineers, spread globally, the organization provides a neutral space for developers and companies to collaborate. The team develops and maintains tools and services for project memberships, telemetry data management, and IT infrastructure.

Business Challenges - Manual Code Reviews are a Bottleneck

The Linux Foundation faced the challenge of a distributed software development team, where code reviews were slow, inefficient, and error-prone. Code reviews were done manually by technical leads and peers, with often two cycles to complete the review and the Pull Request (PR) being ready to merge. This manual process introduced delays especially when engineers in different time zones were waiting for feedback. They were looking for tools to identify inconsistencies and gaps in code quality without burdening team leads.

Key Challenges:

• Manual code reviews have high variance and are error-prone due to critical bugs being missed

• Manual code reviews are limited by individual reviewer knowledge and lack the consistency needed for top-quality engineering teams

• Distributed teams with engineers in different time zones lead to delayed code reviews

• Gaps in code quality end up heavily burdening the team leads when they have to take time out from developing new features

Improving Dev Productivity with AI Code Reviews

The Linux Foundation greatly improved its software development workflow after adopting AI Code Reviews with CodeRabbit. The onboarding was quick (just a few clicks), and it took less than a day to set up, integrate with the Foundation’s GitHub repositories, and receive helpful AI code review recommendations. With CodeRabbit’s SaaS deployment model and support from the customer success team, the setup was completed very quickly.

CodeRabbit’s AI code reviews freed up the engineering team lead’s time and provided a high-quality and automated first pass of the review cycle. Some examples of issues identified by AI code reviews include bug fixes, missed documentation, unit test insertion, and code refactoring suggestions. These are the key features of CodeRabbit that the Linux Foundation used:

Key Features from AI Code Reviews

Always-on code reviews:

• CodeRabbit’s AI delivers senior engineer-level code reviews that are always available, just one click away. Unlike human reviewers, it eliminates the need for engineers to wait for peers or technical leads in different time zones, ensuring an efficient workflow.

PR Summaries and Suggestions:

• Concise and easy-to-understand summaries of complex changes in a PR with a walkthrough of changes in each file associated with the PR.

• Line-by-line recommendations for optimizing code quality with refactoring suggestions.

1-click for bug fixes:

• Immediately identify bugs and errors before they make it to production

• 1-click to accept AI suggestions on fixing bugs and automatically commit the fix to the PR.

Extensive Integrations:

• A wide array of languages and frameworks are supported including:

- Golang, Python, Angular, TypeScript, Terraform and SQL.

• Out-of-the-box integrations with static analyzers and linters ensure CodeRabbit is a one-stop shop for code reviews.

Docstrings & Unit Tests:

• CodeRabbit’s AI reviews identified documentation gaps and recommended inserting unit tests, especially in SQL and DBT testing frameworks.

• Recommendations for Terraform files improved infrastructure management.

Automated Learnings:

• CodeRabbit’s AI learns over time and identifies certain best practices specific to a repository.

• Engineers can see and tweak the automated learnings by providing contextual feedback to the chatbot, further improving the quality of the AI code reviews.

“CodeRabbit has proven invaluable in uncovering discrepancies between our documentation and actual test coverage. Highlighting inconsistencies like missing null checks or mismatched value ranges significantly improved the quality of our codebase and prevented numerous potential issues.” — David Deal, Senior Director of Engineering, The Linux Foundation.

Conclusion

The Linux Foundation is excited to roll out CodeRabbit to additional teams in their organization. They are also looking forward to using CodeRabbit’s upcoming analytics features like tracking PR merge time and human comments reduction to more easily measure the impact of AI in speeding up their dev workflows. They are also excited to see how the AI learns over time as more users engage with the chatbot to reinforce learnings specific to their coding standards in their review cycle.

The Linux Foundation’s adoption of CodeRabbit shows the power of AI code reviews, driving team efficiency and delivering a 25% reduction in the manual time spent on code reviews. CodeRabbit automates the boring stuff and provides actionable insights so the engineers can focus on innovation and maintaining quality in open-source projects.

Get Started with CodeRabbit

You can utilize the power of AI Code Reviews with CodeRabbit as well. It takes less than 5 minutes to get started, and no credit card is required to integrate CodeRabbit with your git platform. Start Your Free Trial, create a new pull request, and watch AI simplify your code reviews in just a few minutes. Questions? Reach out to our team for support.