LexAI

LexAI is an AI-assisted legal drafting and review platform built for handling real-world case documents. It helps ingest PDFs and text files, extract key facts, search across evidence, detect inconsistencies, and generate structured legal outputs.

What it does

• Ingests case documents such as FIRs, witness statements, affidavits, and related attachments
• Runs OCR on scanned documents when needed
• Classifies user intent and routes requests through the right workflow
• Retrieves relevant evidence from a vector store for grounded responses
• Detects contradictions and supports entailment-style verification
• Drafts and assembles legal content from templates and structured context
• Exports generated work into document formats for downstream use

Project layout

• Backend: Django + Django REST Framework
• Async processing: Celery + Redis
• Retrieval and embeddings: ChromaDB + Voyage AI
• LLM orchestration: Anthropic Claude
• OCR: PyMuPDF and Tesseract-based support
• Frontend: React + Vite + TypeScript + Tailwind CSS

The backend is organized into domain apps for documents, orchestration, templates, compilation, retrieval, verification, and agent workflows.

Requirements

• Python 3.10+
• Node.js 18+
• Redis
• API access for the configured AI services

For local development, SQLite is enabled by default. PostgreSQL can be used in production.

Environment setup

Create the backend environment file:

1. Copy the example file in the project root:
   • .env.example to .env
2. Fill in the required values, especially:
   • SECRET_KEY
   • ANTHROPIC_API_KEY
   • VOYAGE_API_KEY
   • REDIS_URL
3. If you are using the frontend locally, copy the example file in the frontend folder:
   • frontend/.env.local.example to frontend/.env.local
4. Set VITE_API_BASE_URL to your backend URL, for example:
   • http://localhost:8000/api/v1

Installation

From the project root:

1. Install Python dependencies:
   • pip install -r requirements.txt
2. Apply database migrations:
   • python manage.py migrate
3. Install frontend dependencies:
   • cd frontend
   • npm install

Run locally

Run the backend, worker, and frontend in separate terminals.

1) Django backend

From the project root:

• python manage.py runserver

2) Celery worker

From the project root:

• celery -A config worker -l info

3) Frontend

From the frontend folder:

• npm run dev

Open the frontend URL shown by Vite, usually http://localhost:5173.

API documentation

If API docs are enabled in the backend settings, the schema and Swagger UI are available at:

• /api/v1/schema/
• /api/v1/schema/ui/

Frontend scripts

Available scripts in frontend/package.json:

• npm run dev — start the Vite development server
• npm run build — type-check and build for production
• npm run preview — preview the production build locally
• npm run generate:types — generate TypeScript API types from the backend schema

Testing

Run the test suite from the project root:

• pytest

Deployment notes

• Use PostgreSQL in production
• Use a persistent Redis instance for Celery and cached workflows
• Set CORS_ALLOWED_ORIGINS to your frontend domain
• Update VITE_API_BASE_URL to the deployed backend API URL

Repository structure

• apps/ — core backend domains and agents
• config/ — Django project configuration
• frontend/ — React user interface
• infrastructure/ — OCR and storage integrations
• tests/ — backend test suite