Creating a Stack Configuration

When I need to define my infrastructure, I want to create a stack, so I can specify model, framework, and hardware options.

---

Introduction

A stack defines your complete AI infrastructure configuration. From the model provider to the cloud region, stacks capture the technical decisions that power your application.

Stack Components

Model Configuration

• Provider — Anthropic, OpenAI, Google, NVIDIA, etc.
• Model ID — Specific model version
• Temperature — Response variability
• Max Tokens — Output length limit
• Context Length — Input window size

Framework Selection

• Orchestration — LangGraph, LangChain, custom
• Observability — Logging and monitoring
• Caching — Response caching strategy

Hardware Options

• Cloud Provider — AWS, GCP, Azure
• Region — Deployment location
• Instance Type — CPU/GPU configuration
• Scaling — Auto-scaling settings

Using Templates

Start from pre-built templates:

• Speed Stack — Optimized for low latency
• Cost Stack — Optimized for budget
• Quality Stack — Maximum capability

Linking to Scenarios

Stacks can be linked to scenarios to enable:

• Automatic validation against requirements
• Cost projections based on workload
• Compatibility checking

Next steps: Link your stack to a scenario for validation.