Agent Development Kit (ADK) Crash Course

This repository contains examples for learning Google's Agent Development Kit (ADK), a powerful framework for building LLM-powered agents.

Getting Started

Setup Environment

You only need to create one virtual environment for all examples in this course. Follow these steps to set it up:

# Create virtual environment in the root directory
python -m venv .venv

# Activate (each new terminal)
# macOS/Linux:
source .venv/bin/activate
# Windows CMD:
.venv\Scripts\activate.bat
# Windows PowerShell:
.venv\Scripts\Activate.ps1

# Install dependencies
pip install -r requirements.txt

Once set up, this single environment will work for all examples in the repository.

Setting Up API Keys

Create an account in Google Cloud https://cloud.google.com/?hl=en
Create a new project
Go to https://aistudio.google.com/apikey
Create an API key
Assign key to the project
Connect to a billing account

Each example folder contains a .env.example file. For each project you want to run:

Navigate to the example folder
Rename .env.example to .env
Open the .env file and replace the placeholder with your API key:
```
GOOGLE_API_KEY=your_api_key_here
```

You'll need to repeat this for each example project you want to run.

Examples Overview

Here's what you can learn from each example folder:

1. Basic Agent

Introduction to the simplest form of ADK agents. Learn how to create a basic agent that can respond to user queries.

2. Tool Agent

Learn how to enhance agents with tools that allow them to perform actions beyond just generating text.

3. LiteLLM Agent

Example of using LiteLLM to abstract away LLM provider details and easily switch between different models.

4. Structured Outputs

Learn how to use Pydantic models with output_schema to ensure consistent, structured responses from your agents.

5. Sessions and State

Understand how to maintain state and memory across multiple interactions using sessions.

6. Persistent Storage

Learn techniques for storing agent data persistently across sessions and application restarts.

7. Multi-Agent

See how to orchestrate multiple specialized agents working together to solve complex tasks.

8. Stateful Multi-Agent

Build agents that maintain and update state throughout complex multi-turn conversations.

9. Callbacks

Implement event callbacks to monitor and respond to agent behaviors in real-time.

10. Sequential Agent

Create pipeline workflows where agents operate in a defined sequence to process information.

11. Parallel Agent

Leverage concurrent operations with parallel agents for improved efficiency and performance.

12. Loop Agent

Build sophisticated agents that can iteratively refine their outputs through feedback loops.

Official Documentation

For more detailed information, check out the official ADK documentation:

https://google.github.io/adk-docs/get-started/quickstart

Support

Need help or run into issues? Join our free AI Developer Accelerator community on Skool:

AI Developer Accelerator Community

In the community you'll find:

Weekly coaching and support calls
Early access to code from YouTube projects
A network of AI developers of all skill levels ready to help
Behind-the-scenes looks at how these apps are built

agent-development-kit-crash-course