Skip to lesson

NVIDIANVIDIA-Certified Professional: Agentic AI2026-101.6 Apply logic trees, prompt chains, and stateful orchestration for multi-step reasoning.

Hybrid Control Flow Example

Completed by opening the official lab

Why it matters

Most production workflows mix an LLM’s choice with fixed sequences. This example puts both in one configuration, so you can see exactly where each decision is made.

What to practise

Read a workflow whose top level is a router_agent with three branches: a sequential_executor pipeline, a second pipeline with a router nested inside it, and another router. The steps are mock functions, so the control flow is all there is to watch. Run each path with nat run and the README’s inputs, and check which branch every input reaches. Note that both pipelines set raise_type_incompatibility: false, and decide what you would check instead before data moves from one step to the next.

How it fits

The pipelines are 1.6.1’s prompt chains — compare their compatibility setting with runtime validation of each intermediate result. The routers are 1.6.2’s LLM-based routing composed into a bounded decision tree, and the whole configuration is a fixed graph in the sense of 1.6.3.