AI Chat Application 💬

What is this? A real-time chat platform where users converse with AI characters powered by Google Gemini, built with microservices architecture.

Why does it exist? Demonstrates production-grade distributed systems patterns: event-driven architecture, message queues, caching, real-time WebSockets, and containerized deployments.

Problem solved: Handles scalable, real-time AI conversations with persistent state, async processing, and fault-tolerant message handling across independent services.

Quick Start (2 minutes)

Prerequisites

• Docker Desktop, Minikube, kubectl, Python 3.12+, Google API Key

Run Locally

cd /Users/devesh1011/code_with_devesh/chat-app
minikube start
eval $(minikube docker-env)

# Build & deploy all services
./deploy.sh  # (create this script with all kubectl apply commands)

# Or manually:
cd src && for dir in auth chat-ws ai_character ai-consumer message-dispatcher gateway; do
  cd $dir && docker build -t deveshk237572/$dir:latest . && cd ..
done
cd ../.. && kubectl apply -f postgres/k8s && kubectl apply -f rabbit/k8s && kubectl apply -f redis/k8s
kubectl apply -f src/*/k8s

Test the API

# Login
curl -X POST http://localhost:8000/login \
  -H "Content-Type: application/x-www-form-urlencoded" \
  -d "username=testuser&password=password123"

# Get characters
curl -X GET http://localhost:8000/characters \
  -H "Authorization: Bearer {access_token}"

# Connect WebSocket
ws://localhost:8004/ws/{character_id}?token={access_token}
# Send: {"content": "Hello!"}

Architecture

Frontend: React + TypeScript (WebSocket client) Backend: 6 microservices (FastAPI)

• Gateway: API routing & auth
• Auth: JWT token generation
• Chat-WS: WebSocket server, session management
• AI Character: Character data & prompts
• AI Consumer: LLM processing (RabbitMQ consumer)
• Message Dispatcher: Persistence & caching

Data Flow: User Message → WebSocket → RabbitMQ → AI Consumer → Gemini API → Message Dispatcher → PostgreSQL/Redis → Response to Client

Tech Stack

Layer	Tech
Frontend	React, TypeScript, WebSocket
Backend	FastAPI, Python, Uvicorn
Database	PostgreSQL 16
Queue	RabbitMQ
Cache	Redis
AI/LLM	Google Generative AI (Gemini 3-Flash), LangChain
Infrastructure	Kubernetes (Minikube), Docker
Auth	JWT

Services Detail

Service	Port	Purpose
Gateway	8000	API routing, auth, character listing
Auth	3001	JWT token generation & validation
Chat-WS	8004	WebSocket connections, session mgmt
AI Character	3000	Character data & system prompts
AI Consumer	-	Gemini API calls, response generation
Message Dispatcher	-	Persistence, caching, notification

Key Design Patterns

• Event-Driven: RabbitMQ for decoupled async processing
• Caching: Redis for fast conversation history retrieval
• Containerized: All services in Docker, orchestrated by Kubernetes
• Stateless: Services can scale horizontally
• Async-First: FastAPI with asyncio for high concurrency

Environment Setup

Set these in k8s/configmap.yaml files:

• GOOGLE_API_KEY - For Gemini AI
• DATABASE_URL - PostgreSQL connection
• RABBITMQ_HOST, REDIS_HOST - Infrastructure addresses

Monitoring

# Check all pods
kubectl get pods

# View logs
kubectl logs -f deployment/gateway
kubectl logs -f deployment/chat-ws
kubectl logs -f deployment/ai-consumer

# RabbitMQ UI
kubectl port-forward svc/rabbitmq 15672:15672
# Visit http://localhost:15672 (guest/guest)

Development

# Run individual service locally
cd src/gateway
pip install -r requirements.txt
uvicorn main:app --reload

# Frontend
cd frontend
bun install
bun run dev

License

MIT