Question 4
Domain 1: Agent Architecture, Design, and DevelopmentYour team is selecting a framework for building agents (LangChain, LlamaIndex, CrewAI, or a custom Python stack). Select TWO criteria that should drive this decision in production agent engineering.
Correct answer: AD
Explanation
Framework choice should be driven by whether its abstractions match the dominant workload: "LlamaIndex for retrieval-heavy pipelines" and "CrewAI for multi-agent role-based orchestration." A custom Python stack is appropriate for "highly bespoke flows where the framework's primitives don't fit," because production engineering should favor the tool that fits the required control and orchestration model.
Why each option is right or wrong
A. Match the framework's abstractions to your dominant workload (e.g., LlamaIndex for retrieval-heavy pipelines, CrewAI for multi-agent role-based orchestration, custom for highly bespoke flows where the framework's primitives don't fit).
Production framework selection turns on fit-to-workload, not brand preference: LlamaIndex is optimized around retrieval-centric pipelines, CrewAI around multi-agent role/task orchestration, and a custom Python stack is justified when the required control flow cannot be expressed cleanly by the framework’s primitives. In production, the decisive criterion is whether the framework’s abstractions reduce glue code and orchestration overhead for the dominant path; if they do not, the engineering cost and brittleness rise immediately.
B. Always pick a custom-built framework, since open-source frameworks introduce dependencies you don't control and are slower than hand-rolled code.
C. The framework choice doesn't matter long-term because the team will inevitably switch frameworks every six months as new ones emerge.
D. Evaluate observability/telemetry integrations (does the framework expose OpenTelemetry tracing? Is it compatible with NeMo Agent Toolkit, Phoenix, Weave, Langfuse?) since debuggability is the dominant operational concern in production.
E. Pick whichever framework has the highest GitHub star count, since popularity correlates strictly with quality and longevity.