Day 2: AWS Bedrock — Call Claude and Other Models via AWS

What You'll Build Today

A Bedrock inference script that calls Claude via boto3, handles responses, and demonstrates streaming. You'll also build a simple comparison wrapper that shows the syntax difference between Bedrock and the direct Anthropic API.

Bedrock vs Direct API

When to use Bedrock vs calling Claude directly

Factor	AWS Bedrock	Direct Anthropic API
Data residency	Stays in your AWS region	Anthropic's infrastructure
AWS integrations	Native (IAM, VPC, CloudWatch)	Manual integration needed
Multiple model providers	Yes (Titan, Llama, Cohere, etc.)	Anthropic models only
Setup complexity	More initial setup	Simpler to start
Best for	Enterprise, compliance-sensitive, AWS-native apps	Prototypes, direct use

Rule of thumb: If you're already on AWS and need compliance (healthcare, government, finance), use Bedrock. If you're prototyping or the app doesn't need AWS services, use the Anthropic API directly.

Install boto3

Set up boto3 for Bedrock

bash

pip install boto3

boto3 picks up your AWS credentials automatically from the aws configure setup in Day 1. No additional setup needed.

First Bedrock Call

Call Claude via Bedrock

pythonbedrock_call.py

import boto3
import json

client = boto3.client("bedrock-runtime", region_name="us-east-1")

def call_claude(prompt, max_tokens=512):
    body = json.dumps({
        "anthropic_version": "bedrock-2023-05-31",
        "max_tokens": max_tokens,
        "messages": [{"role": "user", "content": prompt}]
    })
    response = client.invoke_model(
        modelId="anthropic.claude-opus-4-5-20251101-v1:0",
        body=body,
        contentType="application/json",
        accept="application/json"
    )
    result = json.loads(response["body"].read())
    return result["content"][0]["text"]

# Test it
response = call_claude("Summarize what AWS Bedrock is in 2 sentences.")
print(response)

# Check usage metadata
print(f"\nInput tokens: {json.loads(boto3.client('bedrock-runtime', region_name='us-east-1').invoke_model(modelId='anthropic.claude-opus-4-5-20251101-v1:0', body=json.dumps({'anthropic_version':'bedrock-2023-05-31','max_tokens':10,'messages':[{'role':'user','content':'hi'}]}), contentType='application/json', accept='application/json')['body'].read())['usage']['input_tokens']}")

Streaming

Stream responses for real-time output

pythonbedrock_stream.py

import boto3
import json

client = boto3.client("bedrock-runtime", region_name="us-east-1")

response = client.invoke_model_with_response_stream(
    modelId="anthropic.claude-opus-4-5-20251101-v1:0",
    body=json.dumps({
        "anthropic_version": "bedrock-2023-05-31",
        "max_tokens": 1024,
        "messages": [{"role": "user", "content": "Write a short poem about cloud computing."}]
    }),
    contentType="application/json",
    accept="application/json"
)

for event in response["body"]:
    chunk = json.loads(event["chunk"]["bytes"])
    if chunk["type"] == "content_block_delta":
        print(chunk["delta"]["text"], end="", flush=True)
print()

Day 2 Complete

Understood when to use Bedrock vs the direct Anthropic API
Made a synchronous Bedrock call with boto3
Implemented streaming response handling
Wrapped calls in a reusable function

Day 2 Done

Next: S3 + Lambda — Serverless AI Pipelines

Day 3 wires Bedrock into a Lambda function triggered by S3 file uploads. A fully serverless, event-driven AI pipeline.

Go to Day 3

← Day 1 Course Overview Join Live Bootcamp