10000 GitHub - theailifestyle/google-adk-demos: This repo contains the demos for the google agent development kit tutorial on The AI Lifestyle channel.
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

theailifestyle/google-adk-demos

BranchesTags

Repository files navigation

Google Agent Development Kit (ADK) Tutorial Examples

This repository contains the example code demonstrated in the YouTube tutorial "Building Powerful AI Agents with Google ADK: A Complete Step-by-Step Guide".

Overview

These examples illustrate key concepts of the Google Agent Development Kit (ADK), a Python framework for building, evaluating, and deploying AI agents. They cover:

  • Basic agent creation and tool definition
  • Workflow orchestration using Sequential, Parallel, and Loop agents
  • Advanced concepts like Agents-as-Tools and built-in tools (Search, Code Execution)
  • Deployment to Google Cloud (Vertex AI Agent Engine)
  • Building simple UIs for agents using Streamlit

Prerequisites

  • Python (3.9+ recommended)
  • Virtual Environment (venv)
  • Google Cloud Project (for deployment examples)
  • API Keys (Google AI Gemini API Key) stored securely in .env files or environment variables

Setup

  1. Clone the repository:

    git clone https://github.com/yourusername/googleADK1.git
cd googleADK1

  2. Create and activate a virtual environment:

    python -m venv venv
source venv/bin/activate  # Linux/macOS
# venv\Scripts\activate  # Windows

  3. Install dependencies:

    pip install -r requirements.txt

  4. Configure API Keys:

    • Create a .env file in the project root and add your API keys: 
      GOOGLE_API_KEY=your_google_api_key_here
    • Some scripts may expect environment variables instead of .env files.

Examples

1. Multi-Tool Agent Quickstart (multi_tool_agent_quickstart/)

  • File: agent.py
  • Concept: Demonstrates building your first simple agent (weather_time_agent) using basic Python functions (get_weather, get_current_time) as tools. Illustrates agent definition, instructions, and tool docstrings.
  • Corresponds to Tutorial: Setup, Building Your First Agent, Defining Tools & Agent Logic.
  • To Run: Typically run via adk dev <path_to_agent_py> or integrated into other scripts.

2. Core ADK Samples (samples from adk documention/)

These scripts showcase core ADK features, often run directly.

  • sequentialagent.py: Demonstrates SequentialAgent for a code generation pipeline (Write -> Review -> Refactor). Uses LlmAgent and session state.

    • Concept: Sequential Workflow Orchestration.
    • Corresponds to Tutorial: SequentialAgent for Pipelines.
    • To Run: python "samples from adk documention/sequentialagent.py"

  • parallelagent.py: Demonstrates ParallelAgent for concurrent web research on multiple topics using LlmAgent and the built-in google_search tool.

    • Concept: Parallel Workflow Orchestration, Built-in Tools.
    • Corresponds to Tutorial: ParallelAgent for Concurrent Tasks, Built-in Tools.
    • To Run: python "samples from adk documention/parallelagent.py"

  • loopagent.py: Demonstrates LoopAgent for iterative document writing and refinement. Uses LlmAgent and session state.

    • Concept: Iterative Workflow Orchestration.
    • Corresponds to Tutorial: LoopAgent for Iterative Processes.
    • To Run: python "samples from adk documention/loopagent.py"

  • deploy_agent.py: Demonstrates defining agents-as-tools (specialized translators) and deploying the root agent to Vertex AI Agent Engine. Requires Google Cloud setup.

    • Concept: Agents-as-Tools, Deployment.
    • Corresponds to Tutorial: Agents-as-Tools, Deployment Option 1: Vertex AI Agent Engine.
    • To Run: python "samples from adk documention/deploy_agent.py" (Requires GCP configuration)

3. Streamlit UI Examples (strealit use cases/)

These examples wrap ADK agents in Streamlit web interfaces. Run them using streamlit run <path_to_script.py>.

  • streamlit_code_gen_and_enhancer_sequential_agent.py: UI for the sequential code pipeline, including code execution using built_in_code_execution.

  • streamlit_research_parallel_agent.py: UI for parallel research, including dynamic subtopic generation and report synthesis.

  • streamlit_pitch_optimizer_loop_agent.py: UI for the iterative pitch refinement loop (Critic -> Writer).

  • streamlit_multi_agents_agents_as_tools.py: UI demonstrating a multi-agent "Tutor Bot" where specialists (Math, Spanish, Search) are used as tools (AgentTool). Includes configurable instructions and guardrails.

  • streamlit_multi_agents_subagents.py: UI demonstrating a hybrid multi-agent "Tutor Bot" using sub_agents for Math/Spanish and AgentTool for Search.

  • streamlit_agent_engine_remote_agent.py: UI to connect to and interact with an agent already deployed on Vertex AI Agent Engine (like the one from deploy_agent.py).

4. Agent Management CLI (agent_management_cli.py)

  • Concept: A command-line tool to manage agents deployed on Vertex AI (List, Update, Delete).
  • Corresponds to Tutorial: Deployment (Management aspect).
  • To Run: python agent_management_cli.py (Requires GCP configuration)

Resources

License

This project is licensed under the MIT License - see the LICENSE file for details.

About

This repo contains the demos for the google agent development kit tutorial on The AI Lifestyle channel.

Resources

Readme

License

MIT license
Activity

Stars

28 stars

Watchers

1 watching

Forks

17 forks
Report repository

Releases

No releases published

Packages

No packages published

Languages
0