Which safety threshold is MOST strict and blocks even low-probability harmful content?
本练习是课程的一部分
通过我们的互动练习之一，将理论转化为实践
This module transforms basic instructions into professional agent guidance systems. You’ll learn structured patterns for defining agent persona, role, boundaries, and communication style—turning simple prompts into sophisticated behavioral frameworks grounded in ADK best practices.
This module transforms unpredictable text responses into reliable, structured JSON output that systems can parse and use. You’ll learn how to use Pydantic schemas to enforce output format, making your agents production-ready for integration with applications, databases, and workflows.
This module empowers you to strategically select models and configure generation parameters for optimal performance. You’ll learn when to use Flash versus Pro models, how to set safety thresholds, and how to fine-tune temperature and token limits—transforming basic agents into production-optimized systems.
当前练习
This module adds multi-step reasoning to your agents, transforming them from reactive responders into thoughtful problem-solvers. You’ll learn how to enable Gemini’s built-in thinking capabilities using BuiltInPlanner, allowing agents to plan their approach before executing—essential for complex, multi-step tasks.
Reading List
