Agentic Design Patterns

Agentic Design Patterns are architectural and behavioral patterns for building autonomous AI agents and agentic systems.

These patterns provide proven solutions for designing agents that can:

• Perceive their environment
• Make decisions
• Take actions
• Iterate toward achieving defined goals

They encompass various approaches including:

• Reasoning frameworks
• Planning strategies
• Tool integration
• Memory management
• Multi-agent orchestration

This enables developers to create intelligent systems that operate with minimal human intervention while maintaining control and predictability.

Getting Started

Prerequisites

• Python 3.10 or higher is required.
• Provider API keys:
  • Anthropic (for Claude models)
  • Google (for Gemini models)
  • OpenAI (for OpenAI models)

Library dependencies are listed in the requirements.txt file.

Setup environment

1. Create a virtual environment:

   python3 -m venv .venv

2. Activate the virtual environment:

   source .venv/bin/activate

3. Install the required dependencies:

   pip install -r requirements.txt

Configuration API Keys

Provider API Keys are required to run the examples. You can put them in your shell configuration file (e.g. ~/.zshrc or ~/.bashrc):

export ANTHROPIC_API_KEY="your-anthropic-api-key"
export GEMINI_API_KEY="your-gemini-api-key"
export OPENAI_API_KEY="your-openai-api-key"

Available Patterns

Chaining Pattern

Sequential task decomposition

• Anthropic implementation: chaining/chaining_anthropic.py
• OpenAI implementation: chaining/chaining_openai.py

Routing Pattern

Intelligent request routing

• Google ADK implementation: routing/routing_adk.py