Agentic AI with Google ADK — Learning Repository

A structured, hands-on repository to learn Google Agent Development Kit (ADK) from zero to multi-agent production systems. Each folder is a self-contained project that builds on the previous one.

🚀 Learning Roadmap

Folder	Topic	Concepts Covered
01_basic_agent/	Simple Agent + Tool	LlmAgent, Tool, Runner, Session, Events
02_multi_agent/	Multi-Agent (Delegation)	Intent-based delegation, transfer_to_agent
03_sequential_pipeline/	Sequential Pipelines	SequentialAgent, fixed linear workflows
04_parallel_agents/	Parallel Agents	ParallelAgent, high-speed concurrency
05_stateful_agent/	Stateful & Persistent	SessionService, JSON file persistence
06_rag_agent/	RAG (Retrieval)	Retrieval-Augmented Generation, Search tools

---

🛠️ Quick Start

1. Clone the repo

git clone https://github.com/ayshajaved/Agentic-AI.git
cd Agentic-AI

2. Create a virtual environment

python -m venv venv
venv\Scripts\activate

3. Install dependencies

pip install google-adk python-dotenv

4. Set your Gemini API Key

Create a .env in the root folder:

GOOGLE_API_KEY=your_api_key_here

---

📂 Modules Overview

01. Basic Agent + Tool

Build a single agent that can fetch real-time weather using function calling (tools).

• Run: cd 01_basic_agent; python main.py

02. Multi-Agent Delegation

A coordinator agent that intelligently transfers tasks to specialized sub-agents.

• Run: cd 02_multi_agent; python main.py

03. Sequential Pipelines

Create a "Production Line" where agents hand off data in a fixed order (Researcher → Writer).

• Run: cd 03_sequential_pipeline; python main.py

04. Parallel Agents

Gather diverse data from multiple sources simultaneously for high-speed execution.

• Run: cd 04_parallel_agents; python main.py

05. Stateful & Persistent Agents 🧠

Give your agent long-term memory that persists even after you restart the script.

• Run: cd 05_stateful_agent; python main.py

06. RAG Agent (Retrieval)

A factual assistant that searches local documents to answer questions without hallucinating.

• Run: cd 06_rag_agent; python main.py

---

🧩 Core ADK Concepts

• LlmAgent: Your AI "Brain" with instructions and tools.
• Tool: Python functions the LLM can "act" through.
• Runner: The engine that drives the agentic loop.
• SessionService: Manages the agent's memory (RAM or Disk).
• Sequential/Parallel Agent: specialized containers for complex multi-agent flows.

---

🚀 Tech Stack

• Google ADK (google-adk)
• Gemini 2.0 Flash — Recommended for high-speed agentic loops.
• Python 3.11+
• dotenv for API security.