InsightX - AI-powered Multilingual Document Assistant

InsightX is an AI-powered document processing and analysis system designed for Kochi Metro Rail Limited (KMRL) to handle document overload through automated processing, summarization, and intelligent search capabilities.

🎯 Project Overview

Problem: Document overload at KMRL requiring automated solution to process and summarize documents Goal: Build an MVP that ingests documents (English & Malayalam), extracts insights, summarizes content, and provides a searchable dashboard with alerts

🏗️ Architecture

InsightX/
├── backend/           # RestAPI Python backend
├── frontend/          # React.js TypeScript frontend  
├── infrastructure/    #  GCP configs
└── docs/             # Documentation

🚀 Tech Stack

Backend

• API: FastAPI (Python)
• Database: PostgreSQL (structured data)
• Vector DB: Pinecone (embeddings)
• Storage: Google Cloud Storage
• Queue: Redis (background jobs)

AI/ML

• OCR: Google Cloud Document AI
• LLM: Vertex AI (Gemini models)
• Embeddings: BERT via transformers
• Voice: Google Speech-to-Text

Frontend

• Framework: React.js with TypeScript
• Styling: Tailwind CSS
• Auth: JWT-based authentication

Infrastructure

• Cloud: Google Cloud Platform (GCP)
• Containers: Docker + Kubernetes (GKE)
• CI/CD: GitHub Actions

👥 User Roles

1. Front-line Managers - Operational summaries
2. Compliance Officers - Regulatory alerts & deadlines
3. Cross-Department Teams - Shared updates
4. Engineers/Retiring Employees - Knowledge preservation

🎯 Core Features (MVP)

Document Ingestion

• Upload PDFs, Word, scanned images
• OCR support (English + Malayalam)
• Bulk processing via Redis workers

AI Summarization

• Role-based summaries (compliance, operations, technical)
• Extract deadlines and key actions
• Cross-link related documents

Dashboard & Search

• Web dashboard with document viewer
• Text and voice search capabilities
• Role-based access control

Alerts & Notifications

• Automated compliance alerts
• Email notifications for regulatory updates

🔒 Security

• JWT-based authentication
• Role-based access control
• AES-256 encryption for sensitive data

📊 Success Metrics (MVP)

• OCR accuracy ≥ 80% (English), ≥ 70% (Malayalam)
• Summary relevance rating ≥ 0.8
• Handle 500 documents/week
• Support 3 user roles with separate views

🚀 Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• Docker & Docker Compose
• Google Cloud Platform account

Backend Setup

cd backend
pip install -r requirements.txt
uvicorn app.main:app --reload

Frontend Setup

cd frontend
npm install
npm start

Docker Setup

docker-compose up -d

📝 API Documentation

Once running, visit:

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

🔮 Future Phases

• SharePoint/Maximo integration
• Mobile application
• Advanced analytics & reporting
• Multi-language support expansion

📄 License

MIT License - see LICENSE file for details

---

InsightX - Transforming document chaos into actionable insights