Module 10 · Frameworks

Handoffs, guardrails, and typed output

The runtime style frameworks, rebuilt small. A handoff is a tool call that reassigns one variable, a guardrail is a function that can raise before the model is ever called, and structured output is a validator wired back into the prompt.

4 lessons · 3 graded exercises·4 hours·intermediate·Prerequisite: State graphs and durable execution

Two runtimes, one loop

30 min

Architecture choices between the OpenAI Agents SDK and Pydantic AI, dependency stability risks, and strategies to minimize vendor lock-in.

Handoffs are tool calls

55 min · exercise

Multi-agent handoffs operate by exposing synthetic tool calls for target agents and reassigning the active agent reference in the execution loop.

Guardrails that can stop a run

40 min · exercise

Input guardrails evaluate user prompts before model execution to prevent unnecessary token costs, while output guardrails validate generated responses.

Typed output and the retry loop

55 min · exercise

Structured output uses tool schemas to validate responses, feeding validation errors directly back into the model context for automated correction.