Skip to main content

16 docs tagged with "scope:design"

View all tags

AgentCore Hybrid Strategy

Decision framework and pattern catalog for combining Bedrock AgentCore managed service with EKS-based self-hosted agents in hybrid deployment

Architecture Decision Guide

Covers the six design decisions to finalize when adopting EKS Hybrid Nodes — hybrid connectivity, cluster topology, Pod CIDR exposure, CNI and routing, node authentication, and workload exposure — with decision criteria and inter-decision dependencies.

CIDR Design and Range Minimization

IP range design best practices for EKS Hybrid Nodes — routing requirement assessment procedure, VPC minimum sizing rationale, dedicated control plane ENI subnet strategy, proactive Pod CIDR allocation, and multi-environment address planning.

EKS Hybrid Nodes Concepts and How They Work

Covers Amazon EKS Hybrid Nodes from definition, use cases, and pricing model to the nodeadm registration flow, RemoteNodeNetwork/RemotePodNetwork networking structure, traffic flows, and key technical characteristics.

Knowledge Feature Store Expansion

3-plane design integrating ontology and Knowledge Graph into traditional Feature Store to reduce hallucinations, enable provenance tracking, and enhance domain entity utilization

MCP Tool Token Optimization Patterns

Token optimization patterns for MCP-based agents. Explains the upfront loading cost that follows from the host's tool-injection strategy and four techniques — Progressive Discovery, a tool compression proxy, Code Execution and prompt cache alignment — with published savings figures and the conditions they were measured under.

Node Authentication Methods — SSM vs IAM Roles Anywhere

IAM credential provider selection guide for EKS Hybrid Nodes — SSM hybrid activation vs IAM Roles Anywhere comparison, the Vault PKI integration pattern, Hybrid Nodes IAM role minimum permissions, and credential lifecycle management.

Self-Improving Agent Loop (Autoresearch)

5-stage loop design and safety mechanisms for self-hosted SLMs to autonomously learn and improve from production traces based on Karpathy's autoresearch concept