🤖 Agent Development Kit (ADK) — Learning Repository

A hands-on, progressive learning repository for building AI agents using Google's Agent Development Kit (ADK). Each module builds on the previous one, taking you from a simple greeting agent to a fully stateful, multi-agent customer service system.

Built with: Google ADK • LiteLLM • OpenRouter • Python 3.13

---

📚 Modules

1️⃣ Basic Agent — 1-basic-agent/

Your first AI agent — a simple greeting agent.

• Introduction to the ADK framework
• Creating your first LlmAgent
• Running agents with adk web

1-basic-agent/
└── greeting_agent/
    ├── __init__.py
    └── agent.py

---

2️⃣ Tool Agent — 2-tool-agent/

Giving your agent superpowers with custom tools.

• Defining Python functions as tools
• Connecting tools to an agent
• Understanding how ADK calls tools automatically

2-tool-agent/
└── tool-agent/
    ├── __init__.py
    └── agent.py

---

4️⃣ Structured Output — 4-structured-output/

Generating structured, predictable responses from your agent.

• Enforcing output schemas
• Working with structured email-style responses

4-structured-output/
└── emal_agent/
    ├── __init__.py
    └── agent.py

---

5️⃣ Sessions & State — 5-sessions-and-state/

Making agents remember things within a conversation.

• Introduction to InMemorySessionService
• Managing session state (read/write)
• Building a stateful Q&A agent

5-sessions-and-state/
├── basic_stateful_session.py
└── question_answering_agent/
    ├── __init__.py
    └── agent.py

---

6️⃣ Persistent Storage — 6-persistatt-storage/

Persisting agent memory across sessions using a database.

• Switching from InMemorySessionService → DatabaseSessionService
• SQLite-backed persistence
• Building a reminder/memory agent with CRUD operations
• Using ToolContext for state management inside tools

6-persistatt-storage/
├── main.py
├── utils.py
└── memory_agent/
    ├── __init__.py
    └── agent.py

---

7️⃣ Multi-Agent System — 7-multi-agent/

Orchestrating multiple specialized agents under a manager.

• Creating a Manager Agent that delegates tasks
• Sub-agents: stock_analyst, funny_nerd, news_analyst
• Using AgentTool to call agents as tools
• Using transfer_to_agent for agent-to-agent handoff
• Custom tools: stock price lookup (yfinance), news search (duckduckgo-search)

7-multi-agent/
└── manager/
    ├── __init__.py
    ├── agent.py              # Root manager agent
    └── sub_agents/
        ├── funny_nerd/
        │   └── agent.py      # Joke-telling agent
        ├── news_analyst/
        │   └── agent.py      # News fetching agent (DuckDuckGo)
        └── stock_analyst/
            └── agent.py      # Stock price agent (yfinance)

---

8️⃣ Stateful Multi-Agent System — 8-stateful-multi-agent/

A production-style customer service system with state management.

• Full customer service workflow with 4 specialized sub-agents
• Shared state across agents (purchased_courses, interaction_history)
• XML-style prompt tagging for structured instructions
• Tools that modify session state (purchase, refund)
• Interaction history tracking

8-stateful-multi-agent/
├── main.py                           # Entry point with chat loop
├── utils.py                          # State display & agent response processing
└── customer_service_agent/
    ├── __init__.py
    ├── agent.py                      # Root customer service agent
    └── sub_agents/
        ├── course_support_agent/     # Course content help
        ├── order_agent/              # Purchase history & refunds
        ├── policy_agent/             # Community guidelines & policies
        └── sales_agent/              # Course purchases

---

🚀 Getting Started

Prerequisites

• Python 3.13+
• An OpenRouter API key (free tier available)

Installation

# Clone the repository
git clone https://github.com/itsharshitb/agent-development-kit.git
cd agent-development-kit

# Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate  # macOS/Linux
# .venv\Scripts\activate   # Windows

# Install dependencies
pip install -r requirements.txt

Environment Setup

Create a .env file in the relevant module directory (e.g., 7-multi-agent/manager/.env):

OPENROUTER_API_KEY=your_openrouter_api_key_here

Running an Agent

Option 1: ADK Web UI (for modules 1–7)

cd 7-multi-agent
adk web
# Open http://localhost:8000 in your browser

Option 2: Terminal (for modules 6 & 8)

cd 8-stateful-multi-agent
python main.py

---

🛠️ Tech Stack

Technology	Purpose
Google ADK	Agent framework (agents, tools, sessions, runners)
LiteLLM	Unified LLM API interface
OpenRouter	LLM provider (Arcee, Llama, etc.)
yfinance	Stock market data
DuckDuckGo Search	Web/news search (no API key needed)
python-dotenv	Environment variable management

---

📖 Key Concepts Covered

• Agent Creation — Building agents with LlmAgent and custom instructions
• Tool Integration — Wrapping Python functions as ADK tools
• Session Management — In-memory and persistent (SQLite) session storage
• State Management — Reading/writing state via ToolContext
• Multi-Agent Orchestration — Manager → sub-agent delegation patterns
• Agent-as-Tool — Using AgentTool to call agents like functions
• Agent Transfer — Using transfer_to_agent for conversation handoff
• Prompt Engineering — XML-style tagging, structured instructions, guard clauses

---

📝 License

This project is for educational purposes. Feel free to use and modify the code for your own learning.

---

👤 Author

Harshit Bhatt — @itsharshitb