Ellipsis (YC W24) - AI code reviews & bug fixes

Hi ProductHunt, Hunter and Nick here! Allow me to introduce Ellipsis, a GitHub bot that automatically performs deep, LLM-powered code reviews, fills in PR descriptions, and can even write working, tested code to fix bugs. Simply open a pull request to get an AI Code Review and summary. As you keep pushing code, Ellipsis will update the summary and review the new changes. If a bug pops up, you can squash it with 1-click fixes. 🚀 Here's an example from jxnl/instructor (10K stars) 🚀 👨‍💻 What does it catch? Ellipsis will catch logical errors, anti-patterns, security issues, spelling & grammar mistakes, documentation drift, and more. Reviews take less than 3 minutes. You can even have Ellipsis enforce your team style guide rules, like this project. For example, you might add a rule like "Make sure new code has corresponding unit tests." Ellipsis will subsequently leave reminders on any pull requests that try to submit untested code. 🐛 How does it fix bugs? When Ellipsis finds a problem, it'll leave a comment on the offending line, just like a human would. Then, you can tag Ellipsis to have it implement the fix. This allows you to write working, tested code right from the GitHub UI! Internally, Ellipsis actually tests the code it writes by building your project, running the unit tests, linter, etc. 🆒 🚀 Why did we build this? We built Ellipsis because we are TIRED of the status quo code review process: create a PR, ask for review, move on to new task, receive feedback, reacquaint with old code, fix bug, go back to new task... there's too much context switching. With Ellipsis, the reviews are nearly instant and addressing feedback doesn't require checking out an old branch. ❓ Does it work? Yes. State-of-the-art LLM's aren't good enough (yet) to implement entire features, but they are good enough to catch and fix small mistakes in code. Our data shows that teams that use Ellipsis merge pull requests 13% faster than those that don't. Typically, a developer will get a "Looks good to me!" from Ellipsis before asking for a peer review, resulting in a higher quality discussions with fewer [nit] comments. Start a free trial to get 7 days of AI Code Reviews at ellipsis.dev 🤖

Several of my teams have been using Ellipsis for 4-5 months now and the code reviews and ticket summaries are on point. It constantly catches things that human reviewers would have missed. It’s been a great tool!

Nice job! @hunter_brooks Congrats on the launch 🚀

Very cool application, and super excited for documentation drift!

This is really interesting, It will really Open Source maintainers like me. Great work Team Ellipsis.👏🏼 All the best for the launch 🔥

Congrats on the launch! Developer workflows will definitely be improved!

Congrats on the launch @hunter_brooks! How does it ensure the generated code snippets are optimized for performance and best practices?

Ha LGTM-as-a-service! Love the little humor there. This honestly is a great addition especially if the team is stretched super thin and tech leads don't have time to do good code reviews. I've seen junior engineers suffer because tech leads don't give good reviews and they leave as a result. I assume the AI can leverage style guide you created so it will stay consistent and personalize the reviews? Congrats on the launch @hunter_brooks and team!

This is an impressive tool, Hunter! The ability to catch logical bugs and enforce style guide rules with minimal hassle will definitely streamline our code review process. It's great to see LLMs being used effectively in this way—speeding up PRs by 13% is a significant improvement. I can’t wait to give the free trial a go and see how it enhances our workflow. Thanks for sharing this with the community! 🚀

Ellipsis sounds like a game-changer for code reviews, Hunter! The ability to catch logical errors and enforce style guide rules automatically is exactly what teams need to reduce context switching and improve productivity. I love the 1-click bug fix feature - it's so efficient! Plus, merging PRs 13% faster is impressive; this could really enhance the build process for many devs. Can't wait to see how it progresses! Definitely upvoting this!