Skip to content

end-to-end-weather-agent

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History
 
 

end-to-end-weather-agent

README.md

End-to-End Weather Agent with Tools and Memory - CDK

This CDK stack deploys a complete Amazon Bedrock AgentCore Runtime with a sophisticated weather-based activity planning agent. This demonstrates the full power of AgentCore by integrating Browser tool, Code Interpreter, Memory, and S3 storage in a single deployment.

Table of Contents

Overview

This CDK stack creates a comprehensive AgentCore deployment that showcases:

Core Components

  • AgentCore Runtime: Hosts a Strands agent with multiple tools
  • Browser Tool: Web automation for scraping weather data from weather.gov
  • Code Interpreter: Python code execution for weather analysis
  • Memory: Stores user activity preferences
  • S3 Bucket: Stores generated activity recommendations
  • ECR Repository: Container image storage
  • IAM Roles: Comprehensive permissions for all components

Agent Capabilities

The Weather Activity Planner agent can:

  1. Scrape Weather Data: Uses browser automation to fetch 8-day forecasts from weather.gov
  2. Analyze Weather: Generates and executes Python code to classify days as GOOD/OK/POOR
  3. Retrieve Preferences: Accesses user activity preferences from memory
  4. Generate Recommendations: Creates personalized activity suggestions based on weather and preferences
  5. Store Results: Saves recommendations as Markdown files in S3

Use Cases

  • Weather-based activity planning
  • Automated web scraping and data analysis
  • Multi-tool agent orchestration
  • Memory-driven personalization
  • Asynchronous task processing

Architecture

The architecture demonstrates a complete AgentCore deployment with multiple integrated tools:

Core Components:

  • User: Sends weather-based activity planning queries
  • AWS CodeBuild: Builds the ARM64 Docker container image with the agent code
  • Amazon ECR Repository: Stores the container image
  • AgentCore Runtime: Hosts the Weather Activity Planner Agent
    • Weather Agent: Strands agent that orchestrates multiple tools
    • Invokes Amazon Bedrock LLMs for reasoning and code generation
  • Browser Tool: Web automation for scraping weather data from weather.gov
  • Code Interpreter Tool: Executes Python code for weather analysis
  • Memory: Stores user activity preferences (30-day retention)
  • S3 Bucket: Stores generated activity recommendations
  • IAM Roles: Comprehensive permissions for all components

Workflow:

  1. User sends query: "What should I do this weekend in Richmond VA?"
  2. Agent extracts city and uses Browser Tool to scrape 8-day forecast
  3. Agent generates Python code and uses Code Interpreter to classify weather
  4. Agent retrieves user preferences from Memory
  5. Agent generates personalized recommendations
  6. Agent stores results in S3 bucket using use_aws tool

Prerequisites

AWS Account Setup

  1. AWS Account: You need an active AWS account with appropriate permissions

  2. AWS CLI: Install and configure AWS CLI with your credentials

    aws configure

  3. Python 3.10+ and AWS CDK v2 installed

    # Install CDK
npm install -g aws-cdk

# Verify installation
cdk --version

  4. CDK version 2.218.0 or later (for BedrockAgentCore support)

  5. Bedrock Model Access: Enable access to Amazon Bedrock models in your AWS region

  6. Required Permissions: Your AWS user/role needs permissions for:

    • CloudFormation stack operations
    • ECR repository management
    • IAM role creation
    • Lambda function creation
    • CodeBuild project creation
    • BedrockAgentCore resource creation (Runtime, Browser, CodeInterpreter, Memory)
    • S3 bucket operations (for CDK assets and results storage)

Deployment

CDK vs CloudFormation

This is the CDK version of the end-to-end weather agent. If you prefer CloudFormation, see the CloudFormation version.

Option 1: Quick Deploy (Recommended)

# Install dependencies
pip install -r requirements.txt

# Bootstrap CDK (first time only)
cdk bootstrap

# Deploy
cdk deploy

Option 2: Step by Step

# 1. Create and activate Python virtual environment
python3 -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# 2. Install Python dependencies
pip install -r requirements.txt

# 3. Bootstrap CDK in your account/region (first time only)
cdk bootstrap

# 4. Synthesize the CloudFormation template (optional)
cdk synth

# 5. Deploy the stack
cdk deploy --require-approval never

# 6. Get outputs
cdk list

Deployment Time

  • Expected Duration: 15-20 minutes
  • Main Steps:
    • Stack creation: ~2 minutes
    • Docker image build (CodeBuild): ~10-12 minutes
    • Runtime and tools provisioning: ~3-5 minutes
    • Memory initialization: ~1 minute

Testing

Using AWS CLI

# Get the Runtime ARN from CDK outputs
RUNTIME_ARN=$(aws cloudformation describe-stacks \
  --stack-name WeatherAgentDemo \
  --region us-east-1 \
  --query 'Stacks[0].Outputs[?OutputKey==`AgentRuntimeArn`].OutputValue' \
  --output text)

# Get the S3 bucket name
BUCKET_NAME=$(aws cloudformation describe-stacks \
  --stack-name WeatherAgentDemo \
  --region us-east-1 \
  --query 'Stacks[0].Outputs[?OutputKey==`ResultsBucketName`].OutputValue' \
  --output text)

# Invoke the agent
aws bedrock-agentcore invoke-agent-runtime \
  --agent-runtime-arn $RUNTIME_ARN \
  --qualifier DEFAULT \
  --payload $(echo '{"prompt": "What should I do this weekend in Richmond VA?"}' | base64) \
  response.json

# View the immediate response
cat response.json

# Wait a few minutes for processing, then check S3 for results
aws s3 ls s3://$BUCKET_NAME/

# Download the results
aws s3 cp s3://$BUCKET_NAME/results.md ./results.md
cat results.md

Using AWS Console

  1. Navigate to Bedrock AgentCore Console
  2. Go to "Runtimes" in the left navigation
  3. Find your runtime (name starts with WeatherAgentDemo_)
  4. Click on the runtime name
  5. Click "Test" button
  6. Enter test payload: 
    {
  "prompt": "What should I do this weekend in Richmond VA?"
}
  7. Click "Invoke"
  8. View the immediate response
  9. Wait 2-3 minutes for background processing
  10. Navigate to S3 Console to download results.md from the results bucket

Sample Queries

Try these queries to test the weather agent:

  1. Weekend Planning:

    {"prompt": "What should I do this weekend in Richmond VA?"}

  2. Specific City:

    {"prompt": "Plan activities for next week in San Francisco"}

  3. Different Location:

    {"prompt": "What outdoor activities can I do in Seattle this week?"}

  4. Vacation Planning:

    {"prompt": "I'm visiting Austin next week. What should I plan based on the weather?"}

How It Works

Step-by-Step Workflow

  1. User Query: "What should I do this weekend in Richmond VA?"

  2. City Extraction: Agent extracts "Richmond VA" from the query

  3. Weather Scraping (Browser Tool):

    • Navigates to weather.gov
    • Searches for Richmond VA
    • Clicks "Printable Forecast"
    • Extracts 8-day forecast data (date, high, low, conditions, wind, precipitation)
    • Returns JSON array of weather data

  4. Code Generation (LLM):

    • Agent generates Python code to classify weather days
    • Classification rules:
      • GOOD: 65-80°F, clear, no rain
      • OK: 55-85°F, partly cloudy, slight rain
      • POOR: <55°F or >85°F, cloudy/rainy

  5. Code Execution (Code Interpreter):

    • Executes the generated Python code
    • Returns list of tuples: [('2025-09-16', 'GOOD'), ('2025-09-17', 'OK'), ...]

  6. Preference Retrieval (Memory):

    • Fetches user activity preferences from memory
    • Preferences stored by weather type: 
      {
  "good_weather": ["hiking", "beach volleyball", "outdoor picnic"],
  "ok_weather": ["walking tours", "outdoor dining", "park visits"],
  "poor_weather": ["indoor museums", "shopping", "restaurants"]
}

  7. Recommendation Generation (LLM):

    • Combines weather analysis with user preferences
    • Creates day-by-day activity recommendations
    • Formats as Markdown document

  8. Storage (S3 via use_aws tool):

    • Saves recommendations to S3 bucket as results.md
    • User can download and review recommendations

Asynchronous Processing

The agent runs asynchronously to handle long-running tasks:

  • Immediate response: "Processing started..."
  • Background processing: Completes all steps
  • Results available in S3 after ~2-3 minutes

Cleanup

Using CDK (Recommended)

cdk destroy

Using AWS CLI

# Step 1: Empty the S3 bucket (required before deletion)
BUCKET_NAME=$(aws cloudformation describe-stacks \
  --stack-name WeatherAgentDemo \
  --region us-east-1 \
  --query 'Stacks[0].Outputs[?OutputKey==`ResultsBucketName`].OutputValue' \
  --output text)

aws s3 rm s3://$BUCKET_NAME --recursive

# Step 2: Terminate any active browser sessions
# Get the Browser ID
BROWSER_ID=$(aws cloudformation describe-stacks \
  --stack-name WeatherAgentDemo \
  --region us-east-1 \
  --query 'Stacks[0].Outputs[?OutputKey==`BrowserId`].OutputValue' \
  --output text)

# List active sessions
aws bedrock-agentcore list-browser-sessions \
  --browser-id $BROWSER_ID \
  --region us-east-1

# Terminate each active session (replace SESSION_ID with actual session ID from list command)
# Repeat this command for each active session
aws bedrock-agentcore terminate-browser-session \
  --browser-id $BROWSER_ID \
  --session-id SESSION_ID \
  --region us-east-1

# Step 3: Delete the stack
aws cloudformation delete-stack \
  --stack-name WeatherAgentDemo \
  --region us-east-1

# Wait for deletion to complete
aws cloudformation wait stack-delete-complete \
  --stack-name WeatherAgentDemo \
  --region us-east-1

Important: Browser sessions are automatically created when the agent uses the browser tool. Always terminate active sessions before deleting the stack to avoid deletion failures.

Using AWS Console

  1. Navigate to S3 Console
  2. Find the bucket (name format: <stack-name>-results-<account-id>)
  3. Empty the bucket
  4. Navigate to Bedrock AgentCore Console
  5. Go to "Browsers" in the left navigation
  6. Find your browser (name starts with WeatherAgentDemo_browser)
  7. Click on the browser name
  8. In the "Sessions" tab, terminate any active sessions
  9. Navigate to CloudFormation Console
  10. Select the WeatherAgentDemo stack
  11. Click "Delete"
  12. Confirm deletion

Troubleshooting

CDK Bootstrap Required

If you see bootstrap errors:

cdk bootstrap aws://ACCOUNT-NUMBER/REGION

Permission Issues

Ensure your IAM user/role has:

  • CDKToolkit permissions or equivalent
  • Permissions to create all resources in the stack
  • iam:PassRole for service roles

Python Dependencies

Install dependencies in the project directory:

pip install -r requirements.txt

Build Failures

Check CodeBuild logs in the AWS Console:

  1. Go to CodeBuild console
  2. Find the build project (name contains "weather-agent-build")
  3. Check build history and logs

Browser Session Issues

If deployment fails due to browser sessions:

  1. List active sessions using AWS CLI
  2. Terminate all active sessions
  3. Retry deployment or cleanup

Memory Initialization Issues

If memory initialization fails:

  1. Check Lambda function logs in CloudWatch
  2. Verify IAM permissions for memory access
  3. Retry deployment