Inspiration
Modern software teams face a common problem: when a production issue occurs, developers often spend significant time manually searching through merge requests, commits, deployments, pipelines, and related files to understand what went wrong.
We wanted to create a simple investigation assistant that helps developers quickly identify likely causes of incidents and receive actionable recommendations. The idea was inspired by the growing role of AI agents in developer workflows and the vision behind GitLab Orbit's software lifecycle intelligence.
What it does
Orbit Trailblazer Agent is an intelligent incident investigation assistant.
A developer can describe an incident such as:
- Login failures
- Pipeline failures
- API outages
- Database migration issues
- Production performance problems
The agent analyzes the incident description and generates:
- Likely root cause analysis
- Recommended investigation steps
- Confidence assessment
- Actionable next steps
The goal is to reduce investigation time and provide a structured starting point for debugging production issues.
How I built it
The project was implemented as a custom GitLab Duo Agent. The agent was configured with Orbit investigation capabilities including Graph Schema inspection, Graph Status analysis, Orbit command execution, command discovery, and repository file access. These capabilities were combined with an incident investigation workflow focused on helping developers understand failures, identify likely root causes, and determine next actions. I built Orbit Trailblazer Agent using:
- Python
- Flask
- HTML
- CSS
- GitLab
The backend processes incident descriptions and categorizes common production failures. Based on keywords and incident patterns, the application generates investigation guidance and recommended actions.
A clean web interface was created to allow users to submit incidents and immediately receive investigation results.
We also created multiple documented investigation scenarios to demonstrate how the agent responds to different classes of incidents.
Challenges I ran into
One of the biggest challenges was understanding how to integrate Orbit concepts into a practical investigation workflow.
We spent considerable time exploring GitLab Orbit documentation, project provisioning, and Orbit-related capabilities while designing a solution that could demonstrate incident investigation workflows.
Another challenge was designing useful investigation outputs that remain simple, understandable, and actionable for developers.
We also had to balance functionality, documentation, user experience, and hackathon timelines while building the project.
Accomplishments that we're proud of
- Built a working incident investigation web application
- Created multiple production incident scenarios
- Developed a clean user interface for incident reporting
- Implemented structured investigation outputs
- Successfully documented the project and investigation workflows
- Demonstrated how AI-assisted investigation can improve developer productivity
What I learned
Through this project we learned:
- How incident response workflows operate in modern software teams
- How AI agents can assist developers during debugging and investigation
- Flask application development and deployment workflows
- GitLab-based project collaboration and documentation practices
- The importance of structured troubleshooting and root cause analysis
What's next for Orbit Trailblazer Agent
Future improvements include:
- Real integration with GitLab repositories
- Merge request and commit analysis
- Pipeline investigation capabilities
- Deployment history correlation
- Automated impact analysis
- Incident severity scoring
- AI-powered root cause detection
- Integration with GitLab Orbit graph capabilities when available
Our long-term vision is to evolve Orbit Trailblazer into a developer investigation companion capable of connecting software lifecycle data and helping teams resolve incidents faster.
Log in or sign up for Devpost to join the conversation.