RepoSense

It is an AI-powered code review platform that automatically reviews your GitHub Pull Requests using RAG (Retrieval Augmented Generation) and LLM analysis. When a PR is opened or updated, RepoSense fetches the diff, retrieves relevant codebase context from a knowledge graph, and posts inline review comments directly on the PR in GitHub.

Screenshots

Landing Page

Dashboard & Repositories

Code Quality Trends & Analytics

Detailed Pull Request Reviews

---

Features

• Automatic PR Reviews — triggers on every PR open or update via GitHub webhooks
• RAG-Enhanced Analysis — indexes your entire codebase into a knowledge graph for context-aware reviews
• Inline GitHub Comments — posts review comments directly on the exact line of code in GitHub
• Quality Scoring — assigns a quality score (0–100) to each PR based on issues found
• Knowledge Graph — extracts functions, classes, and relationships (imports, calls, inheritance) from Python and JS/TS files
• Per-chunk Diff Embedding — each changed hunk is embedded separately for targeted context retrieval
• Dashboard — view all repositories, PR history, and quality scores
• Analytics — track code quality trends, issue categories, and averages over time
• PR Detail View — see all review comments with severity, category, and file location
• Multi-user — each user connects their own GitHub account independently

---

Tech Stack

Frontend

• React 18, Vite, Tailwind CSS
• React Router, Recharts
• React Context API (Global Authentication Context)
• Lucide React icons

Backend

• FastAPI (Python)
• SQLAlchemy + PostgreSQL
• PyGithub, HTTPX

AI / ML

• NVIDIA NIM API (LLM inference — Qwen3 Coder 480B)
• NVIDIA NV-EmbedQA (embeddings)
• Pinecone (vector database)

Auth

• GitHub OAuth 2.0
• JWT cookies

---

How It Works

1. User connects GitHub account via OAuth
2. User enables a repository — optionally grants crawl permission
3. If crawl permission granted, RepoSense indexes the entire codebase:
   • Extracts file, function, and class nodes
   • Detects import, call, and inheritance edges
   • Stores embeddings in Pinecone
4. When a PR is opened/updated, GitHub sends a webhook to the backend
5. Backend fetches the PR diff, retrieves relevant codebase context via RAG
6. LLM reviews the diff with full context and returns structured comments
7. Comments are posted inline on the PR in GitHub
8. PR appears in dashboard with quality score and full review detail
9. Frontend manages global session status (AuthContext) and guards routes (/dashboard, /analytics, /pr/:id) to prevent unauthorized access while redirecting authenticated users to the dashboard.

Quality Score Formula

Starts at 100 and deducts:

• −15 per critical issue
• −5 per warning
• −1 per suggestion

Minimum score is 0.

---

Local Setup

Prerequisites

• Python 3.11+
• Node.js 18+
• PostgreSQL database
• GitHub OAuth App
• NVIDIA NIM API key — build.nvidia.com
• Pinecone account — pinecone.io

1. Clone the repository

git clone https://github.com/FriToS-Ban/RepoSense.git
cd RepoSense

2. Create a GitHub OAuth App

Go to GitHub → Settings → Developer settings → OAuth Apps → New OAuth App

Field	Value
Homepage URL	http://localhost:5173
Authorization callback URL	http://localhost:8000/api/auth/callback

Save the Client ID and Client Secret.

3. Backend setup

cd backend
pip install -r requirements.txt

Create a .env file in the backend folder:

GITHUB_CLIENT_ID=your_github_client_id
GITHUB_CLIENT_SECRET=your_github_client_secret
GITHUB_WEBHOOK_SECRET=any_random_string
NVIDIA_API_KEY=your_nvidia_api_key
NVIDIA_MODEL=qwen/qwen3-coder-480b-a35b-instruct
PINECONE_API_KEY=your_pinecone_api_key
PINECONE_INDEX_NAME=reposense
DATABASE_URL=postgresql://user:password@localhost/reposense
SECRET_KEY=any_random_secret_string
FRONTEND_URL=http://localhost:5173
BACKEND_URL=http://localhost:8000
ENVIRONMENT=development

Start the backend:

cd ..
uvicorn backend.main:app --reload --host 0.0.0.0 --port 8000

4. Frontend setup

cd frontend
npm install

Create a .env file in the frontend folder:

VITE_API_URL=http://localhost:8000

Start the frontend:

npm run dev

5. Expose backend for webhooks (local development)

GitHub needs a public URL to send webhook events. Use ngrok:

ngrok http 8000

Copy the ngrok URL (e.g. https://abc123.ngrok.io) and update:

• BACKEND_URL in your backend .env
• The callback URL in your GitHub OAuth App

6. Open the app

Go to http://localhost:5173, connect your GitHub account, and enable a repository.

---

Project Structure

RepoSense/
├── backend/
│   ├── api/
│   │   ├── deps.py
│   │   └── routes/
│   │       ├── analytics.py
│   │       ├── auth.py
│   │       ├── prs.py
│   │       ├── repos.py
│   │       └── webhooks.py
│   ├── core/
│   │   ├── config.py
│   │   ├── database.py
│   │   └── security.py
│   ├── models/
│   │   └── models.py
│   ├── services/
│   │   ├── github.py
│   │   ├── indexer.py
│   │   ├── llm.py
│   │   ├── rag.py
│   │   └── review.py
│   ├── main.py
│   └── requirements.txt
└── frontend/
    ├── src/
    │   ├── components/
    │   │   ├── GithubIcon.jsx
    │   │   └── Navbar.jsx
    │   ├── context/
    │   │   └── AuthContext.jsx
    │   ├── pages/
    │   │   ├── Analytics.jsx
    │   │   ├── Dashboard.jsx
    │   │   ├── Landing.jsx
    │   │   └── PRDetail.jsx
    │   ├── App.jsx
    │   └── main.jsx
    └── package.json

---

Environment Variables Reference

Variable	Description
GITHUB_CLIENT_ID	GitHub OAuth App client ID
GITHUB_CLIENT_SECRET	GitHub OAuth App client secret
GITHUB_WEBHOOK_SECRET	Secret for verifying webhook payloads
NVIDIA_API_KEY	NVIDIA NIM API key for LLM and embeddings
NVIDIA_MODEL	NVIDIA model string for reviews
PINECONE_API_KEY	Pinecone vector database API key
PINECONE_INDEX_NAME	Pinecone index name (default: reposense)
DATABASE_URL	PostgreSQL connection string
SECRET_KEY	Secret for signing JWT tokens
FRONTEND_URL	Frontend base URL (for CORS and redirects)
BACKEND_URL	Backend base URL (for OAuth callback and webhooks)
ENVIRONMENT	development or production

---

License

MIT