About the Project

Inspiration

Open source projects thrive on collaboration, but code reviews are often slow, inconsistent, and a bottleneck for innovation. We were inspired to build an AI-powered GitHub PR reviewer bot using Amazon Nova Act and Nova Premier to automate high-quality code reviews, helping maintainers and contributors focus on what matters most: building great software.

What We Learned

Integrating advanced AI models (Nova Act, Nova Premier) into real-world developer workflows
Securely managing credentials and environment variables for cloud-based AI services
Handling GitHub API permissions and webhooks for seamless automation
The importance of clear documentation and demo readiness for hackathons

How We Built It

Server: Uvicorn running the app from src.server:app to handle GitHub webhook events
AI Analysis: Amazon Nova Premier (with fallback to Pro) for code review suggestions
PR Metadata: Nova Act API (with fallback to GitHub API) for PR context
Review Posting: Automated review/comments via GitHub API
DevOps: ngrok for local webhook testing, .env for secure config, and robust error handling
Documentation: Markdown with Mermaid diagrams and LaTeX for clarity

Challenges Faced

Authentication: Managing multiple API keys and permissions for GitHub and AWS Bedrock
Model Fallbacks: Ensuring strict use of Nova Act and Premier for hackathon compliance, with clear error handling if unavailable
Webhook Delivery: Debugging webhook failures and ensuring reliable event processing
Demo Readiness: Preparing a publish-ready repo, excluding secrets, and providing clear setup instructions
Cross-Platform Issues: Handling file locks and environment differences, especially on Windows