payguard-poc/FULLSTACK_GUIDE.md

PayGuard Full Stack - Docker-Complete Setup

Complete guide for running the entire PayGuard fraud detection system in Docker.

🎉 COMPLETE DOCKER SETUP - Everything in Containers!

NEW: All services now run in Docker (infrastructure + API + frontend)!

# Start EVERYTHING with ONE command
docker compose up

# In another terminal, start just the consumer
cd consumer && python pipeline.py

# Open browser: http://localhost:3000

✨ Perfect setup! All services (Zookeeper, Kafka, Redis, API, Frontend) run in Docker. Only the Consumer (Python) runs on your machine for live development.

For more details: See DOCKER_QUICK_START.md

System Architecture

┌─────────────────────────────────────────────────────────────────┐
│                                                                 │
│                 React Frontend (Port 3000)                      │
│                  ✓ Token Generation                             │
│                  ✓ Transaction Submission                       │
│                  ✓ Results Display                              │
│                                                                 │
└────────────────┬─────────────────────────────────────────────────┘
                 │ HTTP/REST API
                 ▼
┌─────────────────────────────────────────────────────────────────┐
│                                                                 │
│              FastAPI Backend (Port 8000)                        │
│         • Authentication (JWT)                                  │
│         • Transaction Ingestion                                 │
│         • Kafka Producer                                        │
│                                                                 │
└────────────────┬─────────────────────────────────────────────────┘
                 │ Kafka Event Stream
                 ▼
┌─────────────────────────────────────────────────────────────────┐
│                                                                 │
│         Kafka Consumer Pipeline (6-Step Processing)             │
│         ✓ Feature Extraction (Redis profiling)                  │
│         ✓ Layer 1 - Fast Model Scoring                          │
│         ✓ Layer 2 - Deep Model Scoring (if needed)              │
│         ✓ Decision Engine (Circuit Breaker)                     │
│         ✓ Result Storage (Redis)                                │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘
        │              │              │
        ▼              ▼              ▼
    Kafka         Redis Cache     Zookeeper
   (Port 9092)    (Port 6379)    (Port 2181)

Prerequisites

• Docker Desktop - For Kafka, Zookeeper, Redis
• Python 3.9+ - For API and consumer
• Node.js 16+ - For React frontend
• npm or yarn - For package management

Quick Start (5 Steps)

Step 1: Start Infrastructure

# From payguard-poc root directory
docker compose up -d

# Verify services
docker compose ps

Expected output:

NAME          SERVICE        STATUS
payguard-poc-kafka-1         Up (healthy)
payguard-poc-redis-1         Up (healthy)
payguard-poc-zookeeper-1     Up (healthy)

Step 2: Start API Server

Terminal 1:

cd /Users/pat/Projects/PAT/payguard-poc
source venv/bin/activate  # On Windows: venv\Scripts\activate
uvicorn api.main:app --reload --host 0.0.0.0 --port 8000

Expected output:

INFO:     Uvicorn running on http://0.0.0.0:8000

Step 3: Start Fraud Detection Consumer

Terminal 2:

cd /Users/pat/Projects/PAT/payguard-poc
source venv/bin/activate
cd consumer
python pipeline.py

Expected output:

[INFO] PayGuard Fraud Detection Consumer - INITIALIZATION
[INFO] Waiting for transactions...

Step 4: Start React Frontend

Terminal 3:

cd /Users/pat/Projects/PAT/payguard-poc/frontend
npm run dev

Expected output:

VITE v5.4.21  ready in 543 ms

➜  Local:   http://localhost:3000/
➜  press h to show help

The frontend opens automatically in your browser.

Step 5: Test the System

1. In the React Frontend (http://localhost:3000):

   • User ID is auto-generated
   • Click "Generate Token"
   • Select a preset scenario or enter custom transaction details
   • Click "Submit Transaction"

2. Watch Terminal 2 (Consumer):

   • See the complete 6-step fraud detection pipeline
   • Observe feature extraction, model scoring, decision making

3. Check Results in the Frontend:

   • Transaction submitted status
   • Transaction ID
   • Submission timestamp

Detailed Component Startup

Docker Infrastructure

docker compose up -d

What starts:

• Kafka (9092) - Message broker
• Zookeeper (2181) - Kafka coordinator
• Redis (6379) - User profile cache

Verify:

docker compose logs -f kafka    # Watch Kafka logs
redis-cli ping                  # Should return PONG

Python API Server

cd /Users/pat/Projects/PAT/payguard-poc
python3 -m venv venv            # Create if not exists
source venv/bin/activate
pip install -r requirements.txt # Install if not exists
uvicorn api.main:app --reload --port 8000

Available endpoints:

• GET /health - Health check
• POST /token - Generate JWT token
• POST /transaction - Submit transaction

Test API:

curl http://localhost:8000/health

Python Consumer

cd /Users/pat/Projects/PAT/payguard-poc/consumer
python pipeline.py

What it does:

• Connects to Kafka topic "transactions"
• Runs 6-step fraud detection pipeline
• Stores results in Redis
• Logs detailed execution flow

React Frontend

cd /Users/pat/Projects/PAT/payguard-poc/frontend
npm install  # First time only
npm run dev

Features:

• Responsive web interface
• Real-time transaction submission
• Results history
• Pre-built test scenarios

Testing the Full Stack

Scenario 1: Normal Transaction

1. Frontend: Select "Normal Transaction" scenario
2. Click "Submit Transaction"
3. Watch consumer terminal for 6-step pipeline
4. Layer 1 score < 0.4 → Fast approval
5. See result in frontend

Scenario 2: Suspicious Transaction

1. Frontend: Select "Large Purchase" scenario
2. Click "Submit Transaction"
3. Watch consumer terminal:
   • Layer 1 detects suspicious amount
   • Escalates to Layer 2
   • Shows anomaly (85x normal amount)
   • Layer 2 blocks transaction
4. See result in frontend

Scenario 3: Custom Transaction

1. Frontend: Manually enter amount, merchant, description
2. Click "Submit Transaction"
3. Watch complete pipeline execution
4. Observe decision making

API Response Flow

Frontend → API → Backend

1. Frontend generates token
   POST http://localhost:8000/token?user_id=user_123
   Response: {"access_token": "eyJ...", "token_type": "bearer"}

2. Frontend submits transaction
   POST http://localhost:8000/transaction
   Headers: Authorization: Bearer eyJ...
   Body: {"amount": 100, "merchant": "Starbucks", "description": "..."}
   Response: {"status": "accepted", "transaction_id": "abc-123...", "message": "..."}

3. API publishes to Kafka
   Topic: "transactions"

4. Consumer processes (6 steps)
   Step 1: Feature Extraction
   Step 2: Layer 1 Scoring
   Step 3: Behavioral Analysis
   Step 4: Layer 2 Scoring (if needed)
   Step 5: Decision Engine
   Step 6: Result Storage (Redis)

5. Frontend polls/displays results

Monitoring & Debugging

View API Logs

Terminal 1 shows all API requests and responses.

View Consumer Logs

Terminal 2 shows detailed pipeline execution:

[PIPELINE START] Transaction: abc123...
[STEP 1/6] Feature Extraction & Behavioral Profiling
[REDIS LOOKUP] Found profile for user=user_001...
[FEATURE EXTRACTION] features=[amount, avg_amount, tx_count...]
[STEP 2/6] Layer 1 - Fast Model Scoring
[LAYER 1 SCORING] Fraud probability: 0.0500 → LOW RISK...
[DECISION] ✓ APPROVED
[PROFILE UPDATE] user=user_001 count=1→2...
[RESULT STORAGE] Stored decision in Redis...
[PIPELINE COMPLETE] Transaction abc123... → APPROVED

Check Redis Results

redis-cli
> KEYS result:*           # See all results
> GET result:<tx_id>      # View specific result
> HGETALL user:user_001:profile  # View user profile

View Frontend Network Requests

1. Open browser DevTools (F12)
2. Go to Network tab
3. Submit a transaction
4. See API calls and responses

Troubleshooting

API Connection Error

Error: "Failed to generate token. Is the API running?"

Solution:

# Check if API is running
curl http://localhost:8000/health
# Should return: {"status": "healthy", "service": "payguard-api"}

# If not running, start it:
uvicorn api.main:app --reload --port 8000

Consumer Not Processing

Error: Consumer shows "Waiting for transactions..." but no messages

Solution:

# Verify Kafka is running
docker compose logs kafka

# Check consumer is connected
# (should show Kafka CONNECT message in logs)

# Send a test transaction from frontend
# (consumer should start processing immediately)

Redis Connection Error

Error: "Connection refused" in consumer logs

Solution:

# Start Redis if not running
docker compose restart redis

# Test Redis
redis-cli ping
# Should return: PONG

Port Already in Use

Error: "Address already in use"

Solution:

# Kill process using the port
# For port 3000 (frontend):
kill -9 $(lsof -t -i :3000)

# For port 8000 (API):
kill -9 $(lsof -t -i :8000)

# Then restart the service

Frontend Shows CORS Error

Error: "Access to XMLHttpRequest has been blocked by CORS policy"

Solution: This is normal in development. The API needs to allow requests from localhost:3000.

In production, configure CORS properly in api/main.py:

from fastapi.middleware.cors import CORSMiddleware

app.add_middleware(
    CORSMiddleware,
    allow_origins=["http://localhost:3000", "your-production-domain.com"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

Performance Optimization

Frontend Optimization

# Production build
npm run build

# Analyze bundle size
npm install -g source-map-explorer
source-map-explorer 'dist/assets/*.js'

API Optimization

Use uvicorn workers in production:

uvicorn api.main:app --workers 4 --host 0.0.0.0 --port 8000

Consumer Optimization

Increase Kafka batch size in consumer/pipeline.py:

# Process multiple messages at once
# Improve throughput with multiple consumer threads

Production Deployment

Docker Compose Deployment

Full system in Docker:

docker compose -f docker-compose.prod.yml up -d

Cloud Deployment

Frontend → Vercel/Netlify

cd frontend
npm run build
# Deploy dist/ folder

API → AWS Lambda / Google Cloud Run / Heroku

# Containerize API
docker build -f Dockerfile.api -t payguard-api .

Consumer → Kubernetes / ECS

docker build -f Dockerfile.consumer -t payguard-consumer .

Stopping the System

Stop All Services

# Stop containers
docker compose down

# Stop frontend (Ctrl+C in terminal 3)
# Stop API (Ctrl+C in terminal 1)
# Stop consumer (Ctrl+C in terminal 2)

# Deactivate venv
deactivate

Clean Everything

# Remove all containers and volumes
docker compose down -v

# Remove frontend node_modules
cd frontend && rm -rf node_modules dist

Next Steps

1. ✅ System is running with full frontend
2. 📊 Test with various scenarios
3. 📝 Review logs to understand pipeline
4. 🔧 Customize scenarios in frontend
5. 🚀 Deploy to production

Documentation

• Frontend Details: frontend/README.md
• API Details: README_IMPLEMENTATION.md
• System Architecture: payguard_fraud_detection_flowchart.svg
• Quick Start: QUICKSTART.md

---

Quick Copy-Paste Commands:

# Terminal 1: Start infrastructure
docker compose up -d && docker compose ps

# Terminal 2: Start API
source venv/bin/activate && uvicorn api.main:app --reload

# Terminal 3: Start consumer
cd consumer && python pipeline.py

# Terminal 4: Start frontend
cd frontend && npm run dev

System is running! Open http://localhost:3000 in your browser.