anvesh35/bedrock-chatbot repository overview

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   User Browser  │───►│  Application    │───►│   ECS Fargate   │
│                 │    │  Load Balancer  │    │   Container     │
└─────────────────┘    └─────────────────┘    └─────────────────┘
                                                        │
                                                        ▼
                                               ┌─────────────────┐
                                               │  Amazon Bedrock │
                                               │  (Claude/Nova)  │
                                               └─────────────────┘

# AWS CLI v2
aws --version

# AWS CDK
npm install -g aws-cdk
cdk --version

# Python 3.11+
python --version

# Docker
docker --version

05-chatbot/
├── source/                     # Application source code
│   ├── chatbot_backend.py      # Backend logic with LangChain
│   └── chatbot_frontend.py     # Streamlit UI
├── chatbot/                    # CDK infrastructure
│   ├── app.py                  # CDK app entry point
│   └── chatbot/
│       ├── chatbot_ecr_stack.py    # ECR repository
│       └── chatbot_ecs_stack.py    # ECS service & ALB
├── Dockerfile                  # Container configuration
├── docker-compose.yml          # Local development
├── requirements.txt            # Python dependencies
└── README.md                   # This guide

# Singleton pattern for efficient resource usage
_bedrock_llm = None

def get_bedrock_client():
    global _bedrock_llm
    if _bedrock_llm is None:
        _bedrock_llm = ChatBedrockConverse(
            region_name='us-east-1',
            model='amazon.nova-pro-v1:0',
            temperature=0.1,      # Low randomness for consistent responses
            max_tokens=1000       # Reasonable response length
        )
    return _bedrock_llm

def create_chat_memory():
    return ConversationSummaryBufferMemory(
        llm=get_bedrock_client(), 
        max_token_limit=2000      # Prevents context overflow
    )

def get_ai_response(user_input, chat_memory):
    conversation_chain = ConversationChain(
        llm=get_bedrock_client(), 
        memory=chat_memory, 
        verbose=True              # Debug logging
    )
    return conversation_chain.invoke(user_input)['response']

# Initialize persistent memory across page reloads
if 'chat_memory' not in st.session_state: 
    st.session_state.chat_memory = backend.create_chat_memory()

# Initialize chat history for UI display
if 'chat_history' not in st.session_state:
    st.session_state.chat_history = []

# Clone repository
git clone https://github.com/anveshmuppeda/amazon-bedrock.git
cd amazon-bedrock/05-chatbot

# Create virtual environment
python -m venv chatbot-env
source chatbot-env/bin/activate  # Windows: chatbot-env\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Configure AWS CLI
aws configure
# Enter: Access Key, Secret Key, Region (us-east-1), Output format (json)

# Verify configuration
aws sts get-caller-identity

# List available models
aws bedrock list-foundation-models --region us-east-1

# Test model access
aws bedrock invoke-model \
    --model-id "amazon.nova-pro-v1:0" \
    --body '{"messages":[{"role":"user","content":[{"text":"Hello"}]}],"inferenceConfig":{"maxTokens":100}}' \
    --cli-binary-format raw-in-base64-out \
    --region us-east-1 \
    output.json

# Start the application
docker-compose up --build

# Access application
open http://localhost:8501

# Navigate to source directory
cd source

# Set environment variables
export AWS_DEFAULT_REGION=us-east-1
export BEDROCK_MODEL_ID=amazon.nova-pro-v1:0

# Run Streamlit
streamlit run chatbot_frontend.py

cd chatbot
cdk bootstrap aws://$(aws sts get-caller-identity --query Account --output text)/us-east-1

# Deploy ECR stack
cdk deploy ChatbotEcrStack

# Note the ECR repository URI from output

# Get ECR login token
aws ecr get-login-password --region us-east-1 | \
docker login --username AWS --password-stdin \
$(aws sts get-caller-identity --query Account --output text).dkr.ecr.us-east-1.amazonaws.com

# Build for Linux/AMD64 (required for ECS)
docker build --platform linux/amd64 -t chatbot-ecr-repository:latest .

# Tag for ECR
docker tag chatbot-ecr-repository:latest \
$(aws sts get-caller-identity --query Account --output text).dkr.ecr.us-east-1.amazonaws.com/chatbot-ecr-repository:latest

# Push to ECR
docker push \
$(aws sts get-caller-identity --query Account --output text).dkr.ecr.us-east-1.amazonaws.com/chatbot-ecr-repository:latest

# Deploy ECS stack
cdk deploy ChatbotEcsStack

# Note the LoadBalancer DNS from output

# Get ALB DNS name
aws elbv2 describe-load-balancers \
    --names chatbot-alb \
    --query 'LoadBalancers[0].DNSName' \
    --output text

# Access application
open http://[ALB-DNS-NAME]

User: "Remember that I work as a software engineer at AWS"
AI: "I'll remember that you work as a software engineer at AWS."

User: "What's my job?"
AI: "You work as a software engineer at AWS."

User: "I have a meeting at 3 PM about the new project"
AI: "Got it, you have a meeting at 3 PM about the new project."

User: "What time is my meeting?"
AI: "Your meeting is at 3 PM."

User: "What's it about?"
AI: "It's about the new project."

# Test API response time
curl -w "@curl-format.txt" -s -o /dev/null http://[ALB-DNS]/

# Simulate multiple users
for i in {1..5}; do
    curl -X POST http://[ALB-DNS]/ &
done

# Check container status
docker ps

# Check health endpoint
curl http://localhost:8501/_stcore/health

# Check ECS service status
aws ecs describe-services \
    --cluster chatbot-ecs-cluster \
    --services chatbot-ecs-service

Error: AccessDeniedException
Solution: Enable model access in Bedrock console

Error: Token limit exceeded
Solution: Reduce max_token_limit in ConversationSummaryBufferMemory

# Check logs
docker logs [container-id]

# Common issues:
# - Missing AWS credentials
# - Wrong model ID
# - Network connectivity

# Check ECS logs
aws logs get-log-events \
    --log-group-name "/chatbot/logs" \
    --log-stream-name [stream-name]

# Test Bedrock connectivity
aws bedrock list-foundation-models --region us-east-1

# Check ECR repository
aws ecr describe-repositories --repository-names chatbot-ecr-repository

# Verify ECS service
aws ecs list-services --cluster chatbot-ecs-cluster

# Check ALB health
aws elbv2 describe-target-health --target-group-arn [target-group-arn]

# Destroy ECS stack
cdk destroy ChatbotEcsStack

# Destroy ECR stack (after confirming no images needed)
cdk destroy ChatbotEcrStack

# Verify cleanup
aws ecs list-clusters
aws ecr describe-repositories