Accessing Amazon Bedrock

Introduction to Amazon Bedrock

Nikhil Rangarajan

Data Scientist

Meet your instructor!

Machine Learning Engineer
Cloud Technology (AWS, Azure)
Data, Cloud, and Computer Engineering

The profile photo of the instructor

What is Amazon Bedrock?

Amazon's service

The AWS and Amazon Bedrock logos

What is Amazon Bedrock?

Amazon's service
AI models called via API
- API: a messenger between software systems
Bedrock provides access to models from companies such as Anthropic, Meta, etc.
No model setup or training needed - pay as you go

The AWS and Amazon Bedrock logos, and a diagram of an API.

Foundation models in Generative AI

Pre-trained models
Serve as a base for multiple AI tasks
Generate text, code, images based on prompts
- Text/Chat: Claude, Jurassic
- Image: Stable Diffusion
- Embedding: Titan

A diagram showing a Foundation Model at the center with four outward arrows connecting to different capabilities: Text/Chat, Embeddings, Code, and Images.

Understanding access to Amazon Bedrock

Note: Throughout this course, Access Key and Secret Key have been pre-loaded.

Enable Bedrock Access in AWS:
- Create an IAM User in AWS Console
- Add Amazon Bedrock Full Access
- Get your access credentials:
  - Access Key
  - Secret Key
  - Region (e.g., 'us-east-1')

A flowchart showing three steps to enable Amazon Bedrock access in AWS. Step 1: Create IAM User, Step 2: Add Bedrock Access, Step 3: Get Credentials.

Setting up Bedrock runtime access

Configure AWS credentials in your environment

# Not secure for production, only for teaching purposes
AWS_REGION = 'us-east-1'

AWS_ACCESS_KEY = 'your-access-key'

AWS_SECRET_KEY = 'your-secret-key'

Initialize Bedrock client

Use 'bedrock-runtime' for model inference

import boto3 
bedrock = boto3.client('bedrock-runtime', # use -runtime
  region_name=AWS_REGION,
  aws_access_key_id=AWS_ACCESS_KEY,
  aws_secret_access_key=AWS_SECRET_KEY)

Setting up Bedrock access

Check client initialization status
Use 'bedrock' for operational requests

# Verify Bedrock client configuration
import boto3
bedrock_client = boto3.client('bedrock', region_name='us-east-1')

models = bedrock_client.list_foundation_models()
print(f"Connected successfully! 
      Found {len(models['modelSummaries'])} available models.")

Connected successfully!  Found 13 available models.

Information on foundation models

Use get_foundation_model() to retrieve model details

Need a ModelID, unique identifier
- Example: amazon.nova-micro-v1:0, anthropic.claude-3-5-sonnet-v2:0

response = bedrock_client.get_foundation_model
    (modelIdentifier='anthropic.claude-3-5-sonnet-v2:0')

print(response)

Information on foundation models

{
 'modelDetails': {

   'modelName': 'anthropic.claude-3-5-sonnet-v2:0',
   'modelId': 'anthropic.claude-3-5-sonnet-v2:0',
   'providerName': 'Anthropic',
   'modelArn': 'arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-3-5-sonnet',

   'inputModalities': ['TEXT'],
   'outputModalities': ['TEXT'],
   'responseStreamingSupported': True,
   'inferenceTypesSupported': ['ON_DEMAND'],
   'modelLifecycle': {'status': 'ACTIVE'}

   }
}

Let's practice!

Introduction to Amazon Bedrock