Inspiration

Every developer has experienced the frustration of a failed CI/CD pipeline. The hardest part is often not fixing the issue, but identifying what actually caused the failure. In modern software systems, services, repositories, dependencies, and deployment pipelines are deeply interconnected. Developers often spend hours investigating logs, commits, and dependencies across multiple repositories.

We wanted to build a solution that uses GitLab Orbit's knowledge graph to provide immediate context and actionable insights whenever a pipeline fails.

What it Does

Orbit Rover is an AI-powered root cause analysis agent for GitLab.

When a pipeline fails, Orbit Rover automatically:

Retrieves pipeline logs and recent commits
Queries GitLab Orbit for dependency relationships and affected services
Identifies the most likely root cause
Calculates blast radius and downstream impact
Suggests remediation steps
Posts a structured diagnostic report directly into the Merge Request

Developers receive immediate feedback without manually investigating multiple repositories or services.

How We Built It

Orbit Rover is built using:

GitLab Webhooks
GitLab APIs
GitLab Orbit Knowledge Graph APIs
FastAPI
Next.js
OpenAI / Claude LLMs
Tailwind CSS

The backend listens for pipeline failure events and gathers context from GitLab and Orbit. The collected information is passed to an LLM which generates a structured root cause analysis report. The report is then posted back into GitLab and visualized through an optional dashboard.

Challenges

The biggest challenge was combining pipeline logs with graph-based context from Orbit. Traditional log analysis only shows symptoms, while Orbit provides relationships between services, dependencies, repositories, pipelines, and ownership. Combining both allowed us to produce significantly more useful diagnostics.

Another challenge was ensuring that the generated explanations remained concise and actionable for developers.