Amazon Nova AI Hackathon 2025 · Agentic AI Category · Team NovaMind
CodeSense is a multi-agent AI code intelligence system powered by Amazon Nova 2 Lite that reads your entire codebase, reasons across multiple files, and delivers a prioritized bug report with root cause explanations and ready-to-use code fixes — all in plain English.
"CodeSense — Where Amazon Nova thinks like a senior developer, so you don't have to debug alone."
- 🤖 4 Specialist Nova Agents — Bug Hunter, Security, Performance, Validator running in parallel
- 🔁 Self-Correction Loop — Validator Agent eliminates false positives automatically
- 📊 ML Anomaly Detection — IsolationForest scores every file for risk before analysis begins
- 🔒 OWASP Top 10 — Security vulnerability detection with attack scenarios
- 💬 Nova Chat Interface — Ask follow-up questions about any finding
- 🗺️ Risk Heatmap — Visual ML anomaly scores across all modules
Developer → React UI → FastAPI Backend → Orchestrator Agent (Nova 2 Lite)
↓ dispatches to:
┌────────────────────────────────┐
│ Bug Hunter │ Security Agent │
│ Perf Agent │ Validator Agent │
└────────────────────────────────┘
↓ uses tools:
GitHub API · AST Parser · Code Executor
↓
ML Anomaly Layer (IsolationForest)
↓
Prioritized Report + Fix Plans
codesense/
├── backend/
│ ├── agents/
│ │ ├── orchestrator.py # Master controller
│ │ └── specialist_agents.py # 4 Nova-powered agents
│ ├── ml/
│ │ └── anomaly_detector.py # IsolationForest ML layer
│ ├── tools/
│ │ └── tool_definitions.py # Nova tool definitions
│ ├── main.py # FastAPI server
│ ├── nova_client.py # Amazon Nova 2 Lite client
│ └── requirements.txt
├── frontend/
│ ├── src/
│ │ ├── pages/
│ │ │ ├── HomePage.jsx # Landing page
│ │ │ └── ResultsPage.jsx # Results dashboard
│ │ ├── App.jsx
│ │ └── index.css
│ ├── index.html
│ └── package.json
└── README.md
Composite Bug Risk Score Formula:
Risk Score = (40% × IsolationForest Anomaly Score)
+ (30% × Complexity Metrics)
+ (30% × Agent Confidence Score)
This composite scoring guides Nova agents to investigate the highest-risk files first — making analysis faster and more accurate.
- Python 3.10+
- Node.js 18+
- AWS Account with Bedrock access
cd backend
python -m venv venv
venv\Scripts\activate # Windows
pip install -r requirements.txtcopy env.example .env
# Edit .env with your AWS credentialsuvicorn main:app --reload --port 8000cd frontend
npm install
npm run dev| Layer | Technology |
|---|---|
| AI Model | Amazon Nova 2 Lite (Amazon Bedrock) |
| Agent Framework | Custom orchestration + Python |
| Backend | FastAPI + AWS Lambda |
| Data Science | scikit-learn IsolationForest |
| Frontend | React + Vite |
| Code Analysis | Python AST + GitHub API |
| Storage | AWS S3 |
| Member | Role |
|---|---|
| Dua Bano | AI & Backend Lead |
| Ushba Jamshid | Data Science Lead |
| Maryam Noreen | Frontend & Demo Lead |
- Category: Agentic AI
- Core Model: Amazon Nova 2 Lite (
amazon.nova-lite-v1:0) - Hashtag: #AmazonNova
MIT License — feel free to use and build on this project!