← Back to Projects

CodeSage

AI-driven code review backend with modular pipelines, automating reviews while preserving developer privacy through local LLM inference.

  • FastAPI
  • Python
  • CrewAI
  • Ollama
Source Code

Overview

An AI-powered code review system that automates the review process while keeping your code private. Built with a hybrid architecture that combines cloud and local LLM inference.

Why I Built This

Code review is essential but time-consuming. Most AI code review tools send your code to external servers. CodeSage gives you the benefits of AI-assisted review while keeping sensitive code local.

Technical Approach

  • FastAPI backend for high-performance async processing
  • CrewAI for orchestrating multi-agent review pipelines
  • Ollama for local LLM inference (privacy-preserving)
  • Hybrid architecture: Gemini API for general tasks, local models for sensitive code

Recognition

Recognized at a hackathon for advancing secure, developer-centric tooling. The project demonstrates how AI tools can be both powerful and privacy-respecting.