Strands Agent

An AI agent development project leveraging Amazon Bedrock and the Strands Agents SDK.

Project Overview

This project demonstrates how to build various AI agents using the Strands Agents SDK. You can create simple yet powerful agents centered around three core components: a model, tools, and a prompt.

Key Features

Included Utilities

• advanced_rag_utils.py: Advanced RAG (Retrieval-Augmented Generation) capabilities using AWS Bedrock

  • Knowledge Base retrieval and generation
  • Metadata filtering
  • Bedrock client configuration

• llm_cost_utils.py: LLM usage monitoring and cost calculation

  • Bedrock token count tracking
  • CloudWatch metrics integration
  • Cost analysis

Use Cases

The Strands.ipynb notebook includes 10 practical use cases:

1. Web Scraping: Extract article titles and links from Hacker News
2. Stock Price Analysis: Moving averages, volatility, return rate analysis with visualizations
3. Insurance Claims Inspection: Weather data validation and DynamoDB storage
4. Knowledge Base + Internet Search: 10K document retrieval and news aggregation
5. Custom Python Tools: LLM usage calculator
6. DataFrame Manipulation: Data processing with pandas
7. PySpark Data Processing: Large-scale data transformation and storage
8. Machine Learning Pipeline: Customer churn prediction model
9. Multi-Agent System: Financial advisory agent orchestration
10. MCP Integration: AWS documentation querying and architecture diagram generation

Installation

Prerequisites

pip install strands-agents strands-agents-tools strands-agents-builder nest_asyncio uv

Optional Libraries

# For financial analysis
pip install yfinance matplotlib

# For web scraping
pip install beautifulsoup4 pandas requests

# For machine learning
pip install scikit-learn joblib seaborn

Usage

Basic Agent Creation

from strands import Agent

agent = Agent()
agent("Explain Amazon Bedrock Agents?")

Extended Thinking (Reasoning Model)

Extended Thinking (also called "interleaved thinking" in Claude 4) enables Claude to think more deeply when solving complex problems, providing transparent reasoning processes alongside final answers.

For Claude 4 Sonnet (with Interleaved Thinking):

from strands import Agent
from strands.models.bedrock import BedrockModel

model = BedrockModel(
    model_id="us.anthropic.claude-sonnet-4-20250514-v1:0",
    additional_request_fields={
        "anthropic_beta": ["interleaved-thinking-2025-05-14"],
        "thinking": {"type": "enabled", "budget_tokens": 8000}
    }
)

agent = Agent(model=model)
response = agent("What are the benefits of reasoning models?")

The reasoning process is accessible via reasoningContent in the response blocks.

Agent with Tools

from strands_tools import python_repl, file_write
from strands import Agent
import os

os.environ["BYPASS_TOOL_CONSENT"] = "true"

agent = Agent(tools=[python_repl, file_write])
response = agent("Scrape data from a web page and save it as CSV")

Defining Custom Tools

from strands import tool, Agent

@tool
def my_custom_tool(query: str) -> str:
    """Custom tool description"""
    # Tool logic
    return result

agent = Agent(tools=[my_custom_tool])

Configuration

Create a variables.json file in the project root to configure AWS settings:

{
  "kbSemanticChunk": "your-knowledge-base-id",
  "region": "us-west-2",
  "account_number": "your-aws-account-number"
}

Supported Models

• Amazon Bedrock models (Nova, Claude, Titan, etc.)
• Anthropic Claude model family
• Ollama (for local development)
• Other providers via LiteLLM

File Structure

.
├── Strands.ipynb                      # Main demo notebook
├── research_agent_kb_internet.ipynb   # Research agent example
├── advanced_rag_utils.py              # RAG utility functions
├── llm_cost_utils.py                  # Cost calculation tools
├── kb_id.txt                          # Knowledge Base ID
├── strands.png                        # Project image
├── README.md                          # This file (English)
└── README-kr.md                       # Korean version

Key Highlights

• Model-Driven Approach: Easy integration with various LLM models
• Rich Tool Ecosystem: 20+ pre-built tools and MCP server support
• Simple API: Build powerful agents with just a few lines of code
• Scalable: From local development to production deployment
• AWS Integration: Native integration with Bedrock, DynamoDB, CloudWatch, and more

Resources

• Strands Agents Official Documentation
• Amazon Bedrock Documentation
• Strands Agents Announcement Blog
• Advanced RAG using Bedrock and SageMaker - AWS Samples

License

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

Contributing

Issues and pull requests are welcome!