RepoAI - RAG Based Repository Analysis

Problem

Understanding large codebases is time-consuming. Developers spend hours searching files, tracing dependencies, understanding architecture, and onboarding into unfamiliar repositories.

Solution

RepoAI converts any Git repository into an AI-powered knowledge base. It analyzes source code, generates embeddings, stores them in a vector database, and enables developers to ask natural language questions about the codebase.

Features

• GitHub URL & ZIP repository ingestion
• Tree-sitter based code analysis
• Function, class, method, route, and component extraction
• Dependency and code flow analysis
• Semantic code chunk generation
• RAG (Retrieval-Augmented Generation)
• BAAI/bge-base-en-v1.5 embedding generation
• PostgreSQL + pgvector vector database
• Semantic code search using cosine similarity
• AI-powered repository chat
• Architecture discovery and tracing
• JWT Authentication
• Google OAuth Integration
• Razorpay Payment Integration
• Multi-user repository access
• Context-window optimization for accurate AI responses

Tech Stack

Frontend

• React
• TypeScript
• Vite
• Tailwind CSS

Backend

• Node.js
• Express.js
• PostgreSQL
• Gemini API

Worker

• Python
• FastAPI
• Tree-sitter
• Sentence Transformers

Workflow

Repository Upload
        ↓
Repository Analysis
        ↓
Symbol & Dependency Extraction
        ↓
Code Chunk Generation
        ↓
Embedding Generation
        ↓
PostgreSQL (pgvector)
        ↓
User Query
        ↓
Semantic Retrieval
        ↓
Context Construction
        ↓
LLM Response

Architecture

React Frontend
       ↓
Express Backend
       ↓
PostgreSQL + pgvector
       ↓
FastAPI Analysis Worker
       ↓
LLM + RAG Pipeline

Key Highlights

• AI-powered repository understanding
• Semantic code search
• Vector-based retrieval
• Repository-specific RAG pipeline
• Automated architecture analysis
• Scalable worker-based processing
• Production-ready authentication and payments

images :

Screenshots