Lexntra — AI-Powered Legal Document Generation Platform

Live: Frontend · API

Stack: React 19 · Express 5 · MongoDB · Groq AI · Puppeteer · Cloudinary

Lexntra is a full-stack platform that generates professional legal PDFs (Rent Agreements, NDAs, Freelance Contracts, Partnership Agreements, Service Agreements) from guided forms, augmented by an AI legal assistant and lawyer consultation booking.

---

Highlights

• Polymorphic schema design — 5 legal document types share a single MongoDB collection via Mongoose discriminators, with per-type validation and a unified query interface
• Server-side PDF pipeline — Form data → EJS template → Puppeteer headless Chrome → A4-formatted PDF download with embedded digital signatures, typography, and print CSS
• Domain-restricted AI chatbot — Real-time legal Q&A via Groq Cloud's Llama 3.1-8B, with server-enforced system prompt limiting to legal-only queries
• End-to-end auth security — bcrypt pre-save hooks, JWT Bearer tokens (1h expiry), protected routes, Mongoose injection protection, EJS output escaping
• Clean layered architecture — Controllers, routes, middlewares, utilities, and models separated with async error handling, custom error classes, and standardized API responses
• Production deployed — Backend on Render (Puppeteer Chrome via postinstall), frontend on Vercel, CORS whitelist, environment-driven config

---

Tech Stack

Layer	Technologies
Frontend	React 19, Vite 7, React Router DOM v7, Tailwind CSS 3, Axios (JWT interceptor), lucide-react
Backend	Node.js, Express 5, Mongoose (discriminators), JWT, bcryptjs, EJS, Puppeteer, Multer
AI	Groq Cloud API (Llama 3.1-8B instant), server-side domain enforcement
Media	Cloudinary (signature upload, PDF hosting)
Email	Nodemailer (Gmail SMTP)
Infrastructure	Render.com (backend), Vercel (frontend)

---

Architecture

React Form → FormData → Axios POST (JWT) → Multer → Auth Middleware
  → Controller → Cloudinary (signature) → Mongoose (discriminator save)
    → EJS render → Puppeteer (headless Chrome → A4 PDF)
      → Response stream → Temp file cleanup

Database: Mongoose Discriminators

All 5 document types inherit from a LegalDocument base schema keyed by DocumentType:

• Base: { DocumentType, user, DocumentCreationDate }
• RentAgreement — landlord, tenant, property, rent, lease, signatures
• NdaAgreement — parties, binding agreement, signatures
• FreelanceAgreement — client, freelancer, compensation, jurisdiction
• PartnershipAgreement — company, partner, product info, payment, legal
• ServiceAgreement — customer, services, compensation, contract term

Single collection (legaldocuments), queryable by type, with per-type Mongoose validation.

API Endpoints

Method	Endpoint	Auth	Description
POST	/api/v1/user/register	—	Create account
POST	/api/v1/user/login	—	Login, returns JWT
POST	/api/v1/documents	Bearer JWT	Submit form → generate PDF
GET	/api/v1/documents/history	Bearer JWT	User's document history
POST	/api/v1/consultation/book	—	Book lawyer consultation (email)
POST	/api/v1/chat	—	AI legal chatbot query

---

Getting Started

# Backend
cd backend
npm install
cp .env.example .env   # fill in secrets
npm run dev            # localhost:5000

# Frontend
cd frontend
npm install
npm run dev            # localhost:5173

Environment — MongoDB URI, JWT secret, Cloudinary credentials, Groq API key, Gmail app password.

---

Key Design Decisions

• Why discriminators? A single collection with polymorphic schemas avoids N collections with near-identical queries, enables unified history aggregation, and keeps the schema flexible for new document types
• Why server-side PDF? Puppeteer + EJS gives pixel-perfect A4 output with custom fonts, headers, and signature placement — browser print() cannot match this fidelity or automation
• Why server-enforced AI guardrails? The legal domain requires strict boundaries; the system prompt is enforced server-side, not client-side, preventing prompt injection
• Why no cookies? JWT Bearer tokens avoid CSRF entirely and align with the SPA + mobile API consumption pattern

---

Deployment

• Backend — Render Web Service: npm install (postinstall downloads Chrome) → npm start
• Frontend — Vercel: root frontend, build npm run build, output dist

---

Roadmap

• Redis caching for EJS templates and sessions
• Bull + Redis queue for async PDF generation
• Refresh token flow (httpOnly cookies)
• Zod validation schemas
• Pagination on document history
• Rate limiting (auth, chatbot)
• OAuth (Google, GitHub)
• Docker Compose for one-command setup
• CI pipeline with integration tests

---

License

Portfolio project for demonstration purposes.