Overview
AI-powered GitHub repository reviewer that analyzes code quality, architecture, and security using NVIDIA's inference APIs.
The Problem
Reviewing large GitHub repositories manually is time-consuming. Developers need a quick way to assess the architecture, security flags, and code quality of public repos. GitLens AI automates this by fetching repository files and generating a structured, AI-driven review dashboard.
What I Did
- Built a full-stack web application that fetches repository trees and file contents via the GitHub REST API
- Integrated NVIDIA's inference API (Llama 3/DeepSeek) to provide structured JSON code reviews (Code Quality, Architecture, Security, Documentation)
- Engineered a smart file-filtering algorithm to select the most relevant source files while ignoring build artifacts and dependencies
- Reimagined the UI with a custom editorial brutalist design system inspired by nature photography (deep greens, rust accents)
- Implemented an animated review dashboard with score gauges and detailed category breakdowns
Tech Stack Details
Frontend
React, Vite, Tailwind CSS
AI Provider
NVIDIA API
APIs
GitHub REST API
UI/UX
Framer Motion, Recharts
Design System
Editorial Brutalism
Key Learnings
AI Orchestration
Prompt engineering for complex structured JSON responses from LLMs without relying on strict parser libraries
API Rate Limiting
Managing unauthenticated API limits and designing graceful error handling and smart fetching strategies
Design Token Systems
Translating a non-digital print aesthetic into a highly functional digital dashboard and data visualization interface