Code review
AI code review for code your agents write
Code Review on Longwave is an opinionated interface to help you review pull requests faster and with higher quality. It is built for a world where agents write all the code. Longwave's Code Review surface feels almost like agents are going above and beyond to get their code reviewed and merged, just like a good software engineer would in the past.
Intuitive way to view changes
Longwave groups changed files in the most intuitive way instead of displaying everything as a flattened list of changes. The interface uses a layered approach to display changes where you see the most critical changes along with textual and visual summaries where relevant first. Changes and change summaries are grouped logically so it is possible to review them faster. However, you also have the ability to dig deep when there is a need.
Outcome first interface
The code review interface on Longwave doesn't just show code, it also shows outcomes that were achieved during the effort in a clear and concise way. These include end-to-end functionality tests and performance related outcomes.
Built-in video demos
When the changes involve UI updates, the code review interface can show screenshots and video demos recorded by agents. This makes code reviews faster and more efficient as you don't have to manually run the app and test each updated ui flow or design changes anymore.
Questions people ask
What is the best way to review AI generated code?
The best way to review AI generated code is to review outcomes such as end to end tests and performance changes and understanding the changes on a system level as opposed to reading each line of code. Screenshots and video demos created by agents and submitted with the PR can also significantly speed up the review process.
How does Longwave organize a code review?
Longwave groups files based on the systems they touch and the type of changes. It shows changes intuitively and in an opinionated way with the goal of making the entire change easy to understand as a whole. The reviewer, still has the ability to dig deeper into the actual code, but ideally most of their time will be spent one layer above the raw code verifying tests that were run by the agent, looking at demos and understanding regression risks along wtih their criticality.
Can I review code written by the AI agent on Longwave before creating a pull request?
Yes. Open Change Review work area. Review the changes by picking the workstream where the changes exist.
Can an AI code review tool summarize a change?
Yes. Ask Longwave for a summary of all the changes before creating a pull request. It shows the changes intuitively similar to our PR interface. From here, you can directly create a PR.
Does an AI summary replace a code review?
No. The summary is meant for you to verify the agent's changes before you create the PR. The summary interface is very similar to the PR interface so it is easy for you to understand the changes faster and better before submitting a PR. Code review happens after the PR is created.
Can I commit and push changes from Longwave?
Yes. For uncommitted changes, stage the files, enter a commit message or ask your agent to generate one, and commit. Longwave can push the commit if the repository has a remote. Longwave is deeply integrated with git.
Does Longwave automatically review GitHub pull requests?
It is possible to setup Longwave such that you have an agent review PRs and automatically add comments to it. This is, however, not the default setup currently. You can, however, use Longwave Automations to do it.
Can I review more than one repository on Longwave?
Yes. Change Review lists the repositories that have been added to Longwave. Choose a repository to inspect its pull requests or uncommitted changes on your local machine.
Merge agent work with confidence.
Grouped changes, passing flows, and a recorded demo whenever the UI changes.