Architecture Diagram

🛠️ About AutoDevOps

💡 Inspiration

Modern DevOps workflows are often disjointed, manual, and hard to scale. We were inspired to streamline and automate this entire lifecycle using intelligent agents — each responsible for a critical task in the software delivery pipeline. Our goal was to make DevOps smarter, faster, and more collaborative using AI and cloud-native services.

🚀 What It Does

AutoDevOps is a fully automated, multi-agent DevOps system that:

Reviews code using Vertex AI Gemini
Generates and executes test cases autonomously
Orchestrates CI/CD pipeline events
Scans code for secrets and vulnerabilities
Logs all results and actions in BigQuery for full traceability

All of this is coordinated via Pub/Sub and integrated with GitHub Actions for seamless CI/CD workflows.

🛠️ How We Built It

We developed a system of specialized agents:

Code Reviewer Agent: Uses Gemini to analyze code and publish review summaries
Test Generator Agent: Listens for reviews, generates tests, executes them, and logs output
CI/CD Agent: Triggers pipelines or halts builds based on test outcomes
Security Agent: Performs vulnerability and secret scanning

Communication between agents is handled through Google Cloud Pub/Sub, and all logs are pushed to BigQuery. We used GitHub Actions to tie everything together in the CI/CD process.

⚠️ Challenges We Ran Into

Coordinating asynchronous Pub/Sub messages between multiple agents
Managing IAM roles and secrets securely in GitHub Actions
Handling test generation across different programming languages
Ensuring BigQuery schema integrity and consistent logging across agents
Debugging failure cases where one agent would lag or crash

🏆 Accomplishments That We're Proud Of

Built a fully functional and scalable multi-agent system in under a week
Seamless integration of Vertex AI with real-world DevOps workflows
Achieved reliable and reusable automation with clear modularity
Created a system that can adapt to real development pipelines and scale with minimal changes