CodeSentry

CodeSentry is an advanced, AI-powered security auditing and RAG (Retrieval-Augmented Generation) application designed for deep codebase analysis. It automatically ingests, chunks, embeds, and analyzes GitHub repositories to surface critical security vulnerabilities while allowing developers to semantically query and interact with their codebase.

CodeSentry was built with an experimental, cinematic, and minimal dark UI to emphasize data clarity, deep technical focus, and premium typography.

Source Code: https://github.com/dishasharma23-prog/CodeSentry

---

Architecture

CodeSentry operates on a microservice architecture built for production scale:

1. Frontend (Next.js 16)

   • React, Tailwind CSS, Framer Motion
   • Interactive Source Viewer, Citation Highlighting, and AST-aware Chat
   • Premium Dark/Cinematic UI

2. Backend (Node.js + Express + TypeScript)

   • Orchestrates GitHub cloning and repository lifecycle management
   • MongoDB Atlas integration for metadata, query history, and security findings
   • Provides RESTful bridging to the RAG Service

3. RAG Service (Python + FastAPI)

   • Parsing: Advanced Python AST parsing and generic regex parsing for multiple languages
   • Retrieval: Hybrid pipeline using ChromaDB (Dense/Embeddings) and BM25L (Sparse/Keyword)
   • Reranking: Cross-Encoder reranking (MS MARCO) for retrieval accuracy
   • LLM Engine: Gemini integration for synthesis, RAG answers, and security analysis
   • Strict Isolation: Repository-scoped retrieval and application-level isolation to prevent cross-repository context contamination

4. Database & Storage

   • MongoDB Atlas: Document store
   • ChromaDB: Vector storage (persistent volumes)
   • Local File System: BM25 .pkl indexes and cloned repository source code (persistent volumes)

---

Quick Start (Production via Docker)

CodeSentry includes a complete, production-ready docker-compose.yml for isolated deployment.

1. Prerequisites

• Docker & Docker Compose
• A MongoDB Atlas cluster (or any MongoDB instance)
• A Gemini API Key (or OpenAI API Key)

2. Environment Configuration

Create a .env file in the root directory:

# ==========================================
# Frontend Configuration
# ==========================================
# IMPORTANT: This must be the PUBLIC URL where your backend is accessible to the browser.
NEXT_PUBLIC_API_URL=http://localhost:3001/api

# ==========================================
# Backend Configuration
# ==========================================
# Production MongoDB Atlas String (DO NOT USE in-memory local instances for prod)
MONGODB_URI=mongodb+srv://<username>:<password>@<cluster>.mongodb.net/codesentry
PORT=3001

# ==========================================
# RAG Service Configuration
# ==========================================
LLM_PROVIDER=gemini
LLM_API_KEY=your_gemini_api_key_here
LLM_MODEL=gemini-2.0-flash

# (Optional) Map ports if overriding the defaults
FRONTEND_PORT=3000
BACKEND_PORT=3001
RAG_SERVICE_PORT=8000