Documentation

AWS Bedrock

Integrate AWS Bedrock foundation models with ModelRed

🏔️

AWS Bedrock Foundation Models

Test managed foundation models from leading AI companies through AWS Bedrock's secure, scalable, and serverless platform.

Quick Setup

Get Started in 4 Steps

Connect AWS Bedrock foundation models for security testing.

1

Enable Bedrock Access

Enable access to foundation models in the AWS Bedrock Console.

Model Access → Request model access
2

Get AWS Credentials

Set up AWS IAM credentials with Bedrock permissions.

Access Key ID: AKIA...Secret Access Key: ...Region: us-east-1
3

Set Environment Variables

BASH
export AWS_ACCESS_KEY_ID="AKIA..."
export AWS_SECRET_ACCESS_KEY="your-secret-key"
export AWS_DEFAULT_REGION="us-east-1"
4

Register Your Model

PYTHON
from modelred import ModelRed

async with ModelRed() as client:
    await client.register_bedrock_model(
        model_id="my-bedrock-claude",
        bedrock_model_id="anthropic.claude-v2",
        aws_access_key_id="AKIA...",  # or use env vars
        aws_secret_access_key="your-secret-key",
        region="us-east-1"
    )

Foundation Models

🔮

Claude v2

Anthropic's Claude with Constitutional AI

Popular100K Context
🦙

Llama 2

Meta's open-source Llama 2 models

Open SourceMultiple Sizes
🏔️

Amazon Titan

Amazon's own foundation models

NativeText & Embeddings
⚡

AI21 Jurassic

AI21 Labs' Jurassic foundation models

Enterprise8K Context
🌟

Cohere Command

Cohere's command and generation models

Multilingual4K Context
🎨

Stable Diffusion

Stability AI's image generation models

Image GenXL Version

Configuration Options

Advanced Setup

Register different foundation models from Bedrock.

🎭

Multiple Models

PYTHON
# Register different Bedrock models
models = [
    ("claude-v2", "anthropic.claude-v2"),
    ("llama2-13b", "meta.llama2-13b-chat-v1"),
    ("titan-text", "amazon.titan-text-express-v1"),
    ("jurassic-ultra", "ai21.j2-ultra-v1")
]

for model_id, bedrock_id in models:
    await client.register_bedrock_model(
        model_id=model_id,
        bedrock_model_id=bedrock_id,
        region="us-east-1"
    )
🌍

Regional Deployment

PYTHON
# Different regions for model availability
await client.register_bedrock_model(
    model_id="claude-eu",
    bedrock_model_id="anthropic.claude-v2",
    region="eu-west-1",
    metadata={
        "region": "EU",
        "compliance": "GDPR",
        "use_case": "european_operations"
    }
)

Enterprise Features

🏢 Bedrock Advantages

Serverless & Managed

•

No infrastructure management

•

Pay-per-request pricing

•Automatic scaling
•High availability

Security & Governance

•

Data never leaves AWS

•

VPC endpoint support

•CloudTrail logging
•

IAM fine-grained access

Model IDs Reference

📋 Popular Model IDs

Text Generation

anthropic.claude-v2

Claude v2

meta.llama2-13b-chat-v1

Llama 2 13B

amazon.titan-text-express-v1

Titan Express

ai21.j2-ultra-v1

Jurassic-2 Ultra

More Models

cohere.command-text-v14

Command v14

meta.llama2-70b-chat-v1

Llama 2 70B

amazon.titan-embed-text-v1

Titan Embeddings

stability.stable-diffusion-xl-v1

Stable Diffusion XL

Common Issues

⚠️ Troubleshooting

Model Access Denied
AccessDeniedException: No access to model
Solutions:
  • • Request model access in Bedrock Console
  • • Wait for access approval (can take time)
  • • Check if model is available in your region
Invalid Model ID
ValidationException: Invalid model identifier
Solutions:
  • • Use exact model IDs from Bedrock docs
  • • Check model availability in your region
  • • Verify model ID spelling and format
Insufficient Permissions
UnauthorizedOperation: User not authorized
Solutions:
  • • Add bedrock:InvokeModel permission
  • • Verify AWS credentials are correct
  • • Check IAM policy allows Bedrock access

Quick Test

✅ Verify Your Setup

Run this test to confirm your Bedrock integration is working:

PYTHON
import asyncio
from modelred import ModelRed

async def test_bedrock():
    async with ModelRed() as client:
        # Register Bedrock model
        await client.register_bedrock_model(
            model_id="test-bedrock-claude",
            bedrock_model_id="anthropic.claude-v2",
            region="us-east-1"
        )
        print("✅ Bedrock model registered!")

        # Run security test
        result = await client.run_assessment(
            model_id="test-bedrock-claude",
            test_suites=["basic_security", "content_safety"]
        )
        print(f"🔍 Assessment started: {result.assessment_id}")

asyncio.run(test_bedrock())

Next Steps

🧪

Run Security Tests

Start testing your Bedrock models

🏗️

Try SageMaker

Deploy your own custom models