Environment Setup

Why do prepared materials exist for an 'unscripted' event?

Agentic π Day was designed as a fully unscripted live-coding session. But with 110+ sign-ups, prepared these docs and a presentation as a safety net, not a script. The live session stays messy and real. These materials ensure every attendee walks away with complete code, concepts, and references regardless of where the live conversation goes.

Workshop Modules

This workshop is structured as a progressive, hands-on journey. You’ll start by building a simple conversational agent and incrementally add capabilities (tools, memory, multi-agent orchestration, safety guardrails, and production deployment) until you have a fully functional SupportBot running on AWS. Each module builds directly on the previous one, so by the end you’ll understand how every piece fits together.

Module	Topic	Duration
0	Environment Setup & Intro	20 min
1	Your First Agent with Strands	25 min
2	Custom Tools & MCP Servers	35 min
3	Memory & Context Management	30 min
4	Multi-Agent Patterns	30 min
5	Evals, Safety & Observability	30 min
6	Deploy to AWS AgentCore	25 min

Prerequisites

Before the workshop, make sure you have the following installed:

• Python 3.12+: Download Python
• Git: For cloning the workshop repository
• A code editor (VS Code recommended)
• One of the following model providers:
  • AWS credentials + Bedrock access (for Claude on Bedrock), or
  • OpenAI API key (OPENAI_API_KEY), or
  • OpenRouter API key (OPENROUTER_API_KEY), or
  • Any other provider supported by Strands Agents SDK

AWS is optional for Modules 1–5

AWS credentials and Bedrock access are only required for Module 6 (deployment to AgentCore). For Modules 1–5, you can use any supported model provider by configuring shared/model.py.

Setup Steps

1. Clone the workshop repository

   git clone <workshop-repo-url>
   cd agentic-ai-workshops

2. Create a Python virtual environment

   cd workshop
   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

4. Configure your model

   Open shared/model.py and uncomment one model option. This is the single place that controls which LLM is used across all modules.

   Option	Provider	What you need
   1	AWS Bedrock (Claude Sonnet — default)	AWS credentials configured
   2	AWS Bedrock (Claude Haiku — faster/cheaper)	AWS credentials configured
   3	OpenAI (GPT-4o, GPT-4o-mini)	export OPENAI_API_KEY="..."
   4	LiteLLM (Anthropic, Ollama, Cohere, etc.)	Depends on provider
   5	OpenRouter via LiteLLM	export OPENROUTER_API_KEY="..."

   For example, to use AWS Bedrock (default), uncomment Option 1:

   from strands.models.bedrock import BedrockModel
   
   model = BedrockModel(
       max_tokens=1000,
       temperature=0.7,
   )

   Or to use OpenAI, uncomment Option 3:

   from strands.models.openai import OpenAIModel
   
   model = OpenAIModel(
       model_id="gpt-4o",
       params={"max_tokens": 1000, "temperature": 0.7},
   )

   Change it here once, and every module picks it up automatically.

5. AWS setup (only if using Bedrock or for Module 6 deployment)

   Using AWS Bedrock

   Configure credentials and enable Bedrock model access:

   aws configure
   # Enter your AWS Access Key ID, Secret Access Key, and region (us-east-1)

   Go to the Amazon Bedrock console in us-east-1 and enable access to Anthropic models. Anthropic models require submitting brief use case details in the console; approval is automatic and typically takes a few seconds. Bedrock uses serverless cross-region inference profiles, so once your use case is approved, inference is available across regions.

   Using another provider

   No AWS setup needed for Modules 1–5. Just set the appropriate API key environment variable for your chosen provider and configure shared/model.py.

6. Verify your setup

   python module_00_setup/verify_setup.py

   You should see checks passing. AWS-specific checks (credentials, Bedrock) will show as [SKIP] if you’re using a non-AWS provider — that’s fine for Modules 1–5.

Tip

If you’re using AWS Bedrock and having trouble with credentials, ask the workshop facilitator for temporary credentials.

What’s Installed

Package	Purpose
strands-agents	Core agent framework
strands-agents-tools	Pre-built tool collection
litellm	Unified LLM gateway (OpenRouter, Anthropic, Ollama, etc.)
mcp	Model Context Protocol client
fastmcp	Fast MCP server builder
boto3	AWS SDK for Python (needed for Bedrock & Module 6)
opentelemetry-*	Observability and tracing
rich	Pretty terminal output

Project Structure

workshop/
├── module_00_setup/       # This setup module
├── module_01_first_agent/ # Basic agent
├── module_02_tools_mcp/   # Tools + MCP server
├── module_03_memory/      # Memory patterns
├── module_04_multi_agent/ # Multi-agent system
├── module_05_evals/       # Evals + safety
├── module_06_deploy/      # AgentCore deployment
├── shared/                # Shared config & data
│   ├── model.py           # ⬅ Configure your LLM here (one place for all modules)
│   └── data.py            # Mock product/order data
├── requirements.txt
└── pyproject.toml

Note

Each module builds on the previous one. We start with a simple agent and progressively add tools, memory, multi-agent patterns, evals, and deployment.