⚖️ LegalAi: The Intelligent Justice Engine

Democratizing Legal Justice with NVIDIA-Powered AI. Winner/Participant at Rubix TSEC Hackathon

LegalAi is a high-performance "Judge-Safe" legal research platform. Designed for both citizens and legal professionals, it bridges the gap between the complex Indian Penal Code (IPC) and the new Bharatiya Nyaya Sanhita (BNS) using state-of-the-art Retrieval-Augmented Generation (RAG).

🏗️ The Architecture (Techny & Fancy)

LegalAi is built on a distributed microservices architecture designed for reliability and speed.

🧠 Core Intelligence: NVIDIA NIM Integration

We utilize NVIDIA NIM (NVIDIA Inference Microservices) for ultra-low latency legal reasoning.

• Primary Model: meta/llama-3.1-70b-instruct for complex legal analysis and drafting.
• Secondary Model: meta/llama-3.1-8b-instruct for fast greetings and general intent classification.

🔍 Search Engine: Cloud-Native RAG

LegalAi uses a high-performance Cloud-Native RAG pipeline to eliminate local memory overhead and maximize speed:

• NVIDIA Embeddings: nv-embedqa-e5-v5 for ultra-precise semantic retrieval via NVIDIA NIM.
• Vector DB: ChromaDB (Self-managing, high-speed vector storage).
• Processing: A 12-stage text cleaning pipeline with OCR support for complex PDF legal documents.

🛡️ Security & Scalability: API Gateway

A specialized Node.js Gateway ensures the system remains stable and secure:

• Rate Limiting: Tiered protection (20 requests/15m for AI, 100 requests/15m for general API) to prevent infrastructure overuse.
• Dynamic Routing: Intelligent intent routing between the RAG engine and lightweight classification models.

---

🚀 Key Features

🏆 "Judge-Safe" Innovations

• ⚖️ Neutral Legal Analysis: Instead of giving advice, it breaks down "Key Factors" and "Possible Interpretations", acting like a neutral legal clerk.
• 🔥🧊 Balanced Arguments: Instantly generates "Arguments For" and "Arguments Against" a case to assist in strategic brainstorming.
• 📄 Professional Memo Export: Convert AI research into a formatted, citation-heavy PDF with one click.

🇮🇳 Specialized for India

• IPC ↔️ BNS Mapping: Real-time cross-referencing between old and new Indian laws.
• 🗣️ Vernacular Intelligence: Native support for Hindi (Input/Output) with voice-to-text integration.
• 📚 Citation-First: Every answer is backed by direct links to IndiaCode statutes.

---

🛠️ Tech Stack

• Frontend: React + Vite + Tailwind CSS + ShadCN UI
• Intelligence: NVIDIA NIM API + Llama 3.1 70B/8B
• Data Engine: Python (FastAPI) + ChromaDB + Sentence Transformers
• Orchestration: Node.js (Express) + http-proxy-middleware + Docker

---

⚡ Getting Started

📦 Installation

# Clone the repository
git clone https://github.com/WillyEverGreen/TSEC_LEGAL_AI.git
cd TSEC_LEGAL_AI

# Install all dependencies (Unified Setup)
npm install

🔑 Environment Setup

Create a .env file in the root based on .env.example:

NVIDIA_API_KEY=your_nvidia_api_key
VITE_API_URL=http://localhost:8000

🏃 Running the Engine

Launch the entire stack (Frontend + Gateway + RAG) with a single command:

npm run dev:all

The app will be available at http://localhost:5173.

---

📦 Deployment

Docker Compose (Recommended)

docker-compose -f deployment/docker-compose.yml up --build

Manual Production Start

npm run build
npm start

---

📂 System Organization

• rag_service/: The heart of the AI—Python FastAPI service managing ChromaDB and NVIDIA NIM.
• server/: Node.js Gateway providing rate limiting, proxying, and static file serving.
• scripts/maintenance/: Maintenance suite for data ingestion and system setup.
• deployment/: Production-grade container configurations.

📜 Example Queries

1. "What changed in IPC 302 under the new BNS system?"
2. "Draft a balanced legal analysis for a workplace dispute."
3. "Summarize this 50-page judgment and extract key IPC sections."

---

Built for the Rubix TSEC Hackathon.