Agent design patterns

Why it matters

"Let's build an agent" is often the wrong first move. Most successful LLM features are a single well-prompted call, or a few calls wired together in fixed ways. A handful of named patterns cover nearly every design, and knowing them lets you pick the least complicated one that works: cheaper, faster, easier to test, and easier to explain to the people who depend on it.

What you'll be able to do

• Tell workflows (code decides the steps) from agents (the model decides)
• Choose between chaining, routing, parallelization, orchestrator-workers and evaluator-optimizer
• Implement routing, voting and an evaluator-optimizer loop
• Argue for the simplest design that meets the goal

Before this lesson

• The agent loop
• Prompt chaining

Next in this phase: LLMs in automations