Google AI

Agent Creation with Google ADK

3 - 6 hours 4 Modules

Overview

Main Topic: Google AI

Duration: 3 - 6 Hours

Level: Advanced

Target Audience: Developers, Software Architects, and AI Engineers

ABOUT THE TRAINING

The Google ADK is an open-source, modular, and flexible framework designed for AI agent development to follow software engineering standards. The focus is on building robust architectures that can be orchestrated and scaled in the Vertex AI Agent Engine runtime environment. The ADK allows the creation of agents that not only converse but also leverage tools and manage persistent memories professionally.

Objective

Empower developers to build, test, and deploy production-ready agents using the ADK, integrating capabilities for code execution, external tools, and memory banks to solve complex business problems.

TRAINING MODULES

Modular Architecture and ADK Setup

Introduction to the "Agent as Software" philosophy. Setting up the development environment and using the SDK to create model-independent agents, focusing on compatibility with the Gemini ecosystem and other orchestration frameworks.

Tools and Model Context Protocol (MCP)

Empowering the agent through tools: connecting to Cloud API logs and using integrated tools. Implementing the Model Context Protocol (MCP) to enable the agent to interact securely with databases and external systems.

Memory Bank and Session Management

Advanced context and persistence configuration. The student will learn to set up the memory bank so that the agent can generate, retrieve, and review memories, as well as manage conversation sessions via API or console.

Deployment and Agent Engine (A2A)

Publication workflow in the Vertex AI Agent Engine, Google Cloud's managed environment. Introduction to the Agent2Agent (A2A) protocol for communication between multiple agents and configuration of governance (IAM), monitoring, and execution logs.

Additional materials

Structuring your application with Google ADK brings significant advantages: modularity, specialization, reuse, better maintainability, and the ability to define structured control flows using dedicated workflow agents. ADK lets you build sophisticated applications by composing several distinct agent instances into a multi-agent system.

Agent initialization example

from google.adk.agents import Agent
from vertexai.agent_engines import AdkApp

agent = Agent(
    model=model,                  # Required.
    name='currency_exchange_agent', # Required.
    generate_content_config=generate_content_config, # Optional.
)
app = AdkApp(agent=agent)