RetainAI 🧠

AI-Powered Customer Retention Platform

RetainAI is a full-stack SaaS platform that uses machine learning and multi-agent AI to predict customer churn, explain risk factors, and automatically generate retention strategies — across multiple industries including Telecom, Banking, and E-commerce.

---

🌍 Why RetainAI?

• Global churn costs businesses $1.6 trillion/year (Accenture)
• Acquiring a new customer costs 5–7x more than retaining one
• A 5% improvement in retention increases profits by 25–95% (Harvard Business Review)
• RetainAI helps organizations identify at-risk customers 30–60 days early and take automated, data-driven action

---

⚙️ Tech Stack

Layer	Technology
Frontend	React.js, Axios, Recharts, React Dropzone
Backend	Spring Boot (Java), Spring Data JPA, Apache Commons CSV
ML / Agent	Python, Flask, LangChain, LangGraph, Gemini, SHAP
Database	MySQL 8.0
Integration	REST APIs
DevOps	Docker, Docker Compose

---

🏗️ Architecture

┌─────────────────┐     REST      ┌──────────────────┐     REST     ┌─────────────────────┐
│   React Frontend│ ────────────► │  Spring Boot API  │ ──────────► │  Python Agent (Flask)│
│   (Port 3000)   │               │   (Port 8080)     │             │   (Port 5000)        │
└─────────────────┘               └──────────────────┘             └─────────────────────┘
                                           │                                   │
                                           ▼                                   ▼
                                   ┌──────────────┐                 ┌──────────────────────┐
                                   │  MySQL 8.0   │                 │  Multi-Agent System  │
                                   │  (Port 3307) │                 │  - Prediction Agent  │
                                   └──────────────┘                 │  - Explanation Agent │
                                                                     │  - Recommendation    │
                                                                     │  - Monitoring Agent  │
                                                                     │  - Copilot Agent     │
                                                                     └──────────────────────┘

---

🔌 API Endpoints (Backend)

Method	Endpoint	Description
POST	/customers/ml-predict	Predict churn for a single customer
GET	/customers	Fetch all customers
GET	/customers/stats	Churn analytics and retention stats
POST	/upload-csv	Bulk upload customers via CSV

🤖 Agent API Endpoints (Python)

Method	Endpoint	Description
POST	/analyze	Full analysis — prediction, explanation, recommendation
POST	/copilot	AI Copilot for natural language retention queries

---

📊 Current Features

• ✅ Multi-agent AI system (prediction, explanation, recommendation, monitoring, copilot)
• ✅ Spring Boot REST API with MySQL integration
• ✅ CSV bulk upload with automatic customer parsing
• ✅ React dashboard with analytics charts (Recharts)
• ✅ AI Copilot chat interface
• ✅ AI Explanation cards with SHAP-based reasoning
• ✅ Customer form for manual entry + predictions
• ✅ File upload with drag-and-drop (React Dropzone)
• ✅ Stats cards — total customers, high risk, churn rate, retention rate
• ✅ Customer table with risk badges
• ✅ Fully Dockerized — one command to run everything
• ✅ Public/protected route split with dual navbar layout
• 🚧 Industry selector — Telecom, Banking, E-commerce (in progress)
• 🚧 Automated retention email campaigns (in progress)
• 🚧 Customer segment view — Critical / At Risk / Stable (planned)
• 🚧 Authentication & role-based access (planned)

---

🤖 Multi-Agent System

The Python agent layer uses LangChain + LangGraph and consists of 5 specialized agents:

Agent	Role
Prediction Agent	ML model inference — churn probability + binary prediction
Explanation Agent	SHAP-based feature importance — explains why a customer churns
Recommendation Agent	Generates personalized retention strategies per customer
Monitoring Agent	Tracks model performance and detects data drift
Copilot Agent	Natural language interface for retention queries

---

🚀 Running with Docker (Recommended)

Prerequisites

• Docker Desktop installed and running

One command to start everything:

docker-compose up --build

This starts 4 services:

Service	URL
Frontend	http://localhost:3000
Backend	http://localhost:8080
Agent	http://localhost:5000
MySQL	localhost:3307

To stop:

docker-compose down

To rebuild a specific service:

docker-compose build --no-cache frontend
docker-compose up

---

🛠️ Manual Setup (Without Docker)

1. MySQL

mysql -u root -p
CREATE DATABASE churn_db;

2. Backend

cd backend/backend
./mvnw spring-boot:run

3. Agent

cd agent
pip install -r requirements.txt
python agent_api.py

4. Frontend

cd Frontend/my-app
npm install
npm start

---

📁 Project Structure

RetainAI/
├── Frontend/
│   └── my-app/
│       ├── src/
│       │   ├── components/       # AICopilot, AnalyticsCharts, CustomerForm,
│       │   │                     # CustomerTable, FileUpload, AIInsights,
│       │   │                     # AIExplanationCard, Navbar, Navbar2, Layout
│       │   ├── pages/            # LandingPage, Login, Signup, Dashboard,
│       │   │                     # Home, Report, Features, Pricing,
│       │   │                     # Resources, Working
│       │   └── styles/           # CSS per page/component
│       └── public/
├── backend/
│   └── backend/
│       └── src/main/java/com/churn/backend/
│           ├── controller/       # CustomerController, FileUploadController
│           ├── model/            # Customer.java
│           ├── repository/       # CustomerRepository.java
│           └── service/          # CustomerService.java
├── agent/
│   ├── agents/                   # 5 specialized agents
│   ├── agent_api.py              # Flask entry point
│   └── requirements.txt
├── Dockerfile.agent
├── docker-compose.yaml
└── README.md

---

🌐 Industry Verticals (Roadmap)

Industry	Key Churn Signals	Business Impact
Telecom	Contract type, charges, tenure	20–30% annual churn → ₹60L+/yr losses
Banking	Credit score, products, account activity	₹15–40Cr acquisition cost to replace
E-commerce	Login frequency, cart abandonment, plan	7% monthly churn → ₹70L/month lost

---

👥 Contributors

Name	Role
randomlyclueless	Multi-agent system, Dashboard, Full-stack integration
Rutuja	Backend APIs, Docker setup, Frontend pages

---

📄 License

MIT License