Deployment & Scaling · Advanced
Function Calling, Structured Output and Tools
Give Gemini declared tools it can ask you to run, force schema-valid JSON, and understand the control-flow rule that most teams get wrong: the model never executes anything itself.
- 1.3 hours lesson
- intermediate level
- Google Cloud AI Track
- Narrated presentation
What you’ll learn
- Function declarations: name, description, parameter schema
- The model requests calls; your code executes them — the core security property
- Descriptions and enums as the main lever on tool-selection accuracy
- Tool config modes: automatic, ANY (forced), NONE
- Parallel and compositional function calling
- Response schema for structured output when you want data not action
- Validating model-supplied arguments and capping tool-call turns
Inside this lesson
A narrated, slide-by-slide presentation with 1 prediction exercise, 1 knowledge check, 2 videos. You answer as you go, and the platform checks your work.