AI Companion App - Infrastructure Documentation

Task 2: Infrastructure Setup

COMPLETED

This document describes the cloud infrastructure, Kubernetes orchestration, and CI/CD pipelines configured for the AI Companion App.

1. Cloud Infrastructure (AWS)

The infrastructure is defined using Terraform and includes:

Network Architecture

VPC: 10.0.0.0/16 with public and private subnets
Public Subnets: 10.0.1.0/24, 10.0.2.0/24 (AZ-a, AZ-b)
Private Subnets: 10.0.10.0/24, 10.0.11.0/24 (AZ-a, AZ-b)
ALB: Application Load Balancer with SSL termination

Compute & Containers

ECS Cluster: Fargate launch type for serverless containers
Services: API Server, WebSocket Server, Memory Service, Tool Agent
Auto Scaling: Based on CPU (70%) and memory (80%) utilization

Data Layer

Service	Type	Purpose
RDS PostgreSQL	db.r6g.large	User data, conversations
Aurora PostgreSQL	Serverless v2	Vector embeddings storage
ElastiCache Redis	cache.r6g.large	Session cache, real-time data
S3 Buckets	Standard	Assets, logs, model storage

Security

Secrets Manager: API keys, database credentials
Security Groups: Least-privilege access
IAM Roles: Task execution roles
Encryption: At-rest (RDS, S3) and in-transit (TLS)

2. Kubernetes Orchestration

Cluster Components

ALB

→

API Server

→

Memory Service

WebSocket

→

Tool Agent

Redis

|

PostgreSQL

|

Vector DB

Kubernetes Resources

Namespace: ai-companion
Deployments: 4 microservices with HPA
Services: ClusterIP for internal communication
Ingress: nginx controller with TLS
ConfigMaps: App config, LLM provider settings
Secrets: API keys, credentials
ResourceQuota: CPU, memory, pod limits
PodDisruptionBudget: High availability

Auto Scaling

API Server: 3-20 replicas, scale up 100%/15s, scale down 10%/60s
WebSocket: 3-15 replicas
Memory Service: 2-10 replicas
Tool Agent: 2-10 replicas

3. CI/CD Pipeline

Pipeline Stages

Stage	Job	Description
1	lint-and-test	Code linting (flake8, black, isort), type checking (mypy), unit tests
2	validate-prompts	YAML validation, variable extraction, injection pattern detection
3	validate-flows	Flow YAML structure, connectivity check
4	build-and-push	Build Docker images, push to ECR
5	terraform-deploy	Provision/update AWS infrastructure
6	deploy-kubernetes	Deploy to EKS, rolling update
7	deploy-prompts	Sync prompts to S3 config bucket
8	deploy-flows	Sync flows to S3 config bucket
9	smoke-tests	Health check verification

Triggers

Push to main: Full pipeline including infrastructure deploy
Push to develop: Build and deploy without infrastructure changes
PR to main: Lint, validate, and test only
Prompt/Flow changes: Auto-deploy prompts/flows to S3

Deployment Targets

ECR: Docker image registry
S3: Configuration storage (prompts, flows)
ECS/EKS: Container orchestration
CloudWatch: Logging and monitoring

4. Deployment Architecture

5. Files Delivered

infrastructure/terraform/main.tf - AWS infrastructure as code
infrastructure/kubernetes/manifests.yaml - K8s deployments, services, HPA
infrastructure/.github/workflows/ci-cd.yaml - GitHub Actions pipeline
infrastructure.html - This documentation

6. Next Steps

Task 3: Build Conversational Interface
Task 4: Implement Memory & Context System
Task 5: Build Tool/Action System

← Back to Main Index