GitLab x Vertex AI: Real-Time Carbon-Aware CI/CD

Overall Architecture
AI Agents Trace Example
Electricity Maps Visualization
Main Dashboard
Example Generated MR Comment

Inspiration

Every day, software teams run countless CI/CD jobs without knowing their true environmental impact. Inspired by the growing need for sustainability in tech, we built GitGreen to provide precise visibility into the carbon footprint of CI/CD pipelines and actionable insights for optimization.

What it does

GitGreen integrates seamlessly with GitLab CI/CD and Google Cloud Platform (GCP) to track real-time carbon emissions from pipeline jobs. It uses actual resource usage data from GCP Monitoring and regional carbon intensity from Electricity Maps. An AI agent, powered by Google Vertex AI Gemini, analyzes logs and configurations to suggest improvements directly within GitLab merge requests and Slack.

How we built it

Backend: NestJS deployed on Google Cloud Run
Data Queue: Redis with BullMQ for job management
Data Storage: PostgreSQL on Cloud SQL
Real-time Data: Google Cloud Monitoring API & Electricity Maps API
AI Analysis: Google Vertex AI Gemini with function calls for deep contextual analysis
Integration: GitLab OAuth for webhooks, Pub/Sub for event-driven architecture, Slack API for notifications
GitLab Component: Custom component published in GitLab CI/CD Catalog for easy pipeline integration

Challenges we ran into

Integrating real-time monitoring and AI analysis efficiently to keep reports timely and accurate.
Ensuring secure, scalable OAuth and webhook integration with GitLab.
Tuning Pub/Sub and Redis workflows for optimal performance.

Accomplishments that we're proud of

Successfully deployed a robust, scalable solution on Google Cloud.
Delivered precise carbon analytics using real monitoring data, significantly more accurate than industry averages.
Built comprehensive integration points for GitLab and Slack, enhancing developer experience and immediate actionability.

What we learned

Advanced utilization of Google Cloud services (Cloud Run, Vertex AI, Pub/Sub) for scalable, event-driven architectures.
Deep integration techniques with GitLab CI/CD and OAuth mechanisms.
Optimizing AI-driven data analysis for actionable sustainability insights.

What's next for GitGreen

Enhancing AI capabilities for broader sustainability suggestions.
Expanding integration with more cloud providers and additional collaboration tools.
Building advanced analytics dashboards for deeper organizational insights.