Architecture Overview

Infrastructure Components

StackFlow is deployed as a serverless application on AWS in the us-east-1 region under account 373544523367. The architecture is designed for high availability across two Availability Zones with automatic failover for all stateful components.

Networking & Security

All StackFlow Lambda functions and data services run within VPC vpc-0c4e3c18734dee8f7 across two private subnets: subnet-05eae5f255dec054f (us-east-1a) and subnet-03ab773ce82d704d1 (us-east-1b). The security group sg-0ada825cda6a75ed6 controls all inbound and outbound traffic.

All three CloudFront distributions are protected by WAF (stackflow-cloudfront-waf) with managed rule groups for OWASP Top 10, known bad inputs, and rate limiting. API Gateway has resource-level throttling configured at 10,000 RPS burst and 5,000 RPS steady state.

Data Layer

The Aurora PostgreSQL 16 cluster hosts the primary stackflow database containing all ITSM records, user data, configuration items, and workflow definitions. Connection pooling is handled by the Lambda runtime using pg connection pooling with a maximum of 10 connections per Lambda instance.

# Connect to Aurora main cluster (from within VPC)
psql -h stackflow-main-prod.cluster-c6pq0smgmlri.us-east-1.rds.amazonaws.com      -U stackflow_admin -d stackflow

# Check active connections
SELECT count(*), state FROM pg_stat_activity GROUP BY state;

AI & ML Layer

The AI layer is built on Amazon Bedrock with the StackFlow Knowledge Base (BXJGG7PIPS) providing RAG capabilities. Documents are embedded using Amazon Titan Embeddings v2 at 1024 dimensions and stored in OpenSearch Serverless. At query time, the model router selects the optimal Claude model based on task complexity, latency requirements, and cost budget.

import boto3

bedrock = boto3.client('bedrock-agent-runtime', region_name='us-east-1')

response = bedrock.retrieve_and_generate(
    input={'text': 'How do I reset a Cognito user password?'},
    retrieveAndGenerateConfiguration={
        'type': 'KNOWLEDGE_BASE',
        'knowledgeBaseConfiguration': {
            'knowledgeBaseId': 'BXJGG7PIPS',
            'modelArn': 'arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-3-sonnet-20240229-v1:0'
        }
    }
)

Observability

StackFlow emits structured logs to CloudWatch Logs and metrics to CloudWatch Metrics. All Lambda invocations include a correlation ID in the x-stackflow-request-id header, which is propagated through all downstream service calls for end-to-end tracing. X-Ray tracing is enabled on the API Gateway and Lambda functions.

The AI Observability module (see AI Observability) provides a unified view of Bedrock API costs, token usage, model performance, and semantic cache hit rates. Alerts are configured in CloudWatch Alarms with SNS notifications to the stackflow-security-findings topic for anomalous spending patterns.