ADK

详解 ADK Go 的 Rewind 机制——如何将 Agent 回溯到历史状态重新执行，适用于 Agent 走错路、需要重新来的场景。

详解 ADK Go 的 Context Compression 机制——如何压缩历史消息，保持上下文精简，控制 token 消耗。

详解 ADK Go 的 Context Caching 机制——如何缓存上下文减少 token 消耗，以及适用场景和限制。

详解 ADK Go 中 Event 的类型、EventHandler 的使用，以及如何监听和处理 Agent 运行过程中的各种事件。

详解 ADK Go 中 State 的概念、读写 API，以及如何在多轮对话中保持和共享自定义状态数据。

Detailed explanation of Session concept, creation, persistence and recovery mechanism in ADK Go, understanding how to maintain Agent context in multi-turn conversations.

Detailed Tool Confirmation mechanism—when Agent needs to execute risky operations, how to let user confirm before executing; plus API Key, OAuth and other auth mechanism configuration.

Detailed guide on using OpenAPI Tool to dynamically generate Tools from OpenAPI specs, suitable for services with API documentation but no MCP interface.

Detailed MCP (Model Context Protocol) working principles, and how to connect external services (like databases, APIs) to ADK Go Agents through MCP.

Detailed Function Tool performance optimization strategies: timeout control, concurrent calls, connection pool management, and how to prevent slow Tools from affecting entire Agent’s response.