GitlabAgents

Ship better code on GitLab: AI agents run static analysis for you.

Comment

Inspiration

Writing software is complex and time consuming. Most software inevitably becomes bigger and more complex as time goes on. Ensuring code quality is one of the best ways to reduce long-term complexity and technical debt.

GitlabAgents was built to help teams automatically detect and address quality issues in their codebases, ensuring high quality software. What's more, for every problem identified in your code, a solution is suggested.

What it does

GitlabAgents scans repos or commits during CI/CD pipelines. Issues are logged and prioritized for teams to take action.

The suggested solutions can help developers improve their coding skills.

How we built it

  • The web app is built with Next.js, TypeScript and PostgreSQL.
  • Integrates with Gitlab: scans repos on-demand and as part of CI.
  • Powered by Gemini 2.0 Flash for fast and effective code analysis.
  • Hosted on a Google Cloud Platform (GCP) VM.
  • Uses AI Construx, my no-code AI agent platform, to manage analysis agents.

Challenges we ran into

  • Time constraints: initially started with a different project idea (a web-service code generator) but pivoted to better align with GitLab track requirements.
  • AI Construx in beta: I had to resolve some issues during integration.
  • Gitlab CI: I had some trouble importing the static analysis component, I kept getting errors about an invalid path, but this was eventually resolved.
  • Potentially high LLM use: I've put a 15 file max limit on each analysis run to avoid rate-limits with large repos.

Accomplishments that we're proud of

  • Seamless GitLab integration with minimal user setup.
  • Hosted by Google's GCP and uses Gemini 2.0 Flash for fast but accurate static analysis of source code.
  • Integrates with AI Construx for AI agent services (I built this platform).

What we learned

  • Static analysis by LLMs ranges from critical bugs to general style advice and should be used more widely.
  • I was new to GitLab and learned about its approach and API, which are well structured and developer friendly.

What's next for GitlabAgents

  • I'm definitely going to use it with my own projects to increase quality.
  • I would like to get feedback from the development community regarding GitlabAgents.
  • If there is interest it would be great to gain users and add features.

Built With

Share this project:

Updates