Complete AWS Bedrock Tutorial: Managed Generative AI on AWS

Amazon Bedrock is a fully managed service that provides access to foundation models (FMs) from leading AI companies through a unified API. It enables building generative AI applications without managing infrastructure.

Why AWS Bedrock?

• Multiple FMs: Access Claude, Llama, Titan, and more
• Fully managed: No infrastructure to manage
• Secure: Data privacy and VPC support
• Customizable: Fine-tune models with your data
• Integrated: Native AWS service integration

• Anthropic Claude (Claude 3, Claude 2)
• Meta Llama 2
• Amazon Titan
• AI21 Labs Jurassic
• Cohere Command
• Stability AI (images)

Prerequisites

pip install boto3

Configure AWS CLI
aws configure

Enable Bedrock model access in AWS Console

Quick Start

1. Basic Text Generation

import boto3
import json

Create Bedrock runtime client
bedrock = boto3.client(
    servicename="bedrock-runtime",

    regionname="us-east-1"
)

Invoke Claude model
def generatetext(prompt):

    body = json.dumps({
        "anthropicversion": "bedrock-2023-05-31",
        "maxtokens": 1024,

        "messages": [
            {"role": "user", "content": prompt}
        ]
    })

    response = bedrock.invokemodel(
        modelId="anthropic.claude-3-sonnet-20240229-v1:0",
        body=body
    )

    result = json.loads(response["body"].read())
    return result["content"][0]["text"]

Generate text
response = generatetext("Explain machine learning in simple terms.")

print(response)

def generatetextstreaming(prompt):
    body = json.dumps({
        "anthropicversion": "bedrock-2023-05-31",
        "maxtokens": 1024,

        "messages": [
            {"role": "user", "content": prompt}
        ]
    })

    response = bedrock.invokemodelwithresponsestream(

        modelId="anthropic.claude-3-sonnet-20240229-v1:0",
        body=body
    )

    for event in response["body"]:
        chunk = json.loads(event["chunk"]["bytes"])
        if chunk["type"] == "contentblockdelta":

            print(chunk["delta"]["text"], end="", flush=True)

generatetextstreaming("Write a short poem about AI.")

def invoketitan(prompt):
    body = json.dumps({
        "inputText": prompt,
        "textGenerationConfig": {
            "maxTokenCount": 1024,
            "temperature": 0.7,
            "topP": 0.9
        }
    })

    response = bedrock.invokemodel(

        modelId="amazon.titan-text-express-v1",
        body=body
    )

    result = json.loads(response["body"].read())
    return result["results"][0]["outputText"]

response = invoketitan("What is cloud computing?")
print(response)

2. Meta Llama 2

def invokellama(prompt):
    body = json.dumps({
        "prompt": f"[INST] {prompt} [/INST]",

        "maxgenlen": 512,

        "temperature": 0.7,
        "topp": 0.9
    })

    response = bedrock.invokemodel(

        modelId="meta.llama2-70b-chat-v1",
        body=body
    )

    result = json.loads(response["body"].read())
    return result["generation"]

response = invokellama("Explain neural networks.")
print(response)

def invokecohere(prompt):
    body = json.dumps({
        "prompt": prompt,
        "maxtokens": 500,
        "temperature": 0.7
    })

    response = bedrock.invokemodel(

        modelId="cohere.command-text-v14",
        body=body
    )

    result = json.loads(response["body"].read())
    return result["generations"][0]["text"]