Lexi - Legal Document Templating System

A full-stack application that transforms legal documents into reusable templates with smart variable extraction and automated drafting capabilities.

🌟 Features

• Document Ingestion: Upload and process PDF/DOCX legal documents
• Smart Template Generation: Automatically extract variables and create reusable templates
• AI-Powered Drafting: Generate documents using natural language queries
• Variable Management: Define and manage template variables with validation
• Web Search Integration: Find relevant templates online when no local match exists

🛠️ Tech Stack

Backend

• Framework: FastAPI (Python)
• LLM: Google Gemini 1.5/Flash
• Database: PostgreSQL/SQLite
• Document Processing: Python-Docx, PyPDF2
• Vector Search: FAISS (for template similarity)

Frontend

• Framework: Next.js 14 (App Router)
• UI: Shadcn UI + Tailwind CSS
• State Management: React Query
• Form Handling: React Hook Form + Zod

🚀 Getting Started

Prerequisites

• Node.js 18+
• Python 3.10+
• PostgreSQL (or SQLite for development)
• Google AI API Key

Installation

1. Clone the repository

   git clone https://github.com/yourusername/lexi.git
   cd lexi

2. Backend Setup

   cd backend
   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -r requirements.txt

3. Frontend Setup

   cd frontend
   npm install

4. Environment Variables Create .env files in both backend/ and frontend/ directories:

   Backend (.env)

   DATABASE_URL=postgresql://user:password@localhost:5432/lexi
   GOOGLE_AI_API_KEY=your-google-ai-key
   EXA_API_KEY=your-exa-ai-key  # For bonus feature
   

   Frontend (.env.local)

   NEXT_PUBLIC_API_URL=http://localhost:8000
   

Running the Application

1. Start Backend

   cd backend
   uvicorn main:app --reload

2. Start Frontend

   cd frontend
   npm run dev

3. Access the Application

   • Frontend: http://localhost:3000
   • API Docs: http://localhost:8000/docs

🏗️ Architecture

Core Components

1. Document Ingestor

   • Handles file uploads (PDF/DOCX)
   • Extracts text and metadata
   • Processes documents into clean text

2. Template Engine

   • Identifies and extracts variables using Gemini
   • Creates Markdown templates with YAML front-matter
   • Manages template storage and retrieval

3. Drafting System

   • Processes natural language queries
   • Matches queries to templates
   • Handles variable collection and validation
   • Generates final documents

4. Web Integration (Bonus)

   • Searches for relevant templates online
   • Processes and ingests new templates on-the-fly
   • Expands template library automatically

🧠 Smart Prompting

Variable Extraction

SYSTEM_PROMPT = """
You are a legal document templating assistant. Extract reusable fields from the provided legal document.
Return a JSON object with variables and their metadata.
"""

Template Matching

TEMPLATE_MATCH_PROMPT = """
Given the user's request and available templates, select the most appropriate template.
Consider the document type, jurisdiction, and content.
Return the template_id and a confidence score.
"""

📚 Data Model

Templates

CREATE TABLE templates (
    id SERIAL PRIMARY KEY,
    title VARCHAR(255) NOT NULL,
    doc_type VARCHAR(100),
    jurisdiction VARCHAR(100),
    similarity_tags TEXT[],
    body_md TEXT,
    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

Template Variables

CREATE TABLE template_variables (
    id SERIAL PRIMARY KEY,
    template_id INTEGER REFERENCES templates(id),
    key VARCHAR(100) NOT NULL,
    label VARCHAR(255) NOT NULL,
    description TEXT,
    example TEXT,
    required BOOLEAN DEFAULT true,
    dtype VARCHAR(50),
    regex TEXT,
    enum_values TEXT[]
);

📄 Example Usage

1. Upload a Document

   • Go to /templates/upload
   • Select a legal document (PDF/DOCX)
   • Review extracted variables
   • Save as a new template

2. Generate a Draft

   • Go to /draft
   • Enter your request (e.g., "Draft a notice to insurer in India")
   • Answer the variable prompts
   • Download or copy the generated document

📝 License

MIT

---

Created with ❤️ by UOIONHHC