AWS Bedrock

WatchThatSite integrates with AWS Bedrock to access a variety of foundational models from providers like Anthropic, Meta, and more.

Prerequisites

• AWS account with Bedrock access
• User or role with Bedrock permissions
• Models must be enabled in your Bedrock setup

Get Your Credentials

1. Create IAM User (Recommended for Security)

1. Go to AWS IAM Console
2. Click Users > Create user
3. Name it (e.g., "watchthatsite-bedrock")
4. Proceed through the wizard

2. Create Access Keys

1. In IAM, select your user
2. Go to Security credentials
3. Click Create access key
4. Choose Application running outside AWS
5. Click Next
6. Copy:
   • Access Key ID
   • Secret Access Key (won't be shown again)

3. Attach Bedrock Policy

1. In IAM, select your user
2. Click Add permissions > Attach policies directly
3. Search for and select AmazonBedrockFullAccess
4. Click Add permissions

For production, use a more restrictive policy limited to specific models and regions.

4. Find Your Region

Bedrock is available in specific regions. Common ones:

• us-east-1 — US East (N. Virginia)
• us-west-2 — US West (Oregon)
• eu-west-1 — Europe (Ireland)
• ap-northeast-1 — Asia Pacific (Tokyo)

Check AWS documentation for current region availability.

Add to WatchThatSite

1. Go to Settings > AI Providers
2. Click Add Provider Key
3. Select AWS Bedrock from the provider dropdown
4. Enter:
   • Access Key ID: Your IAM access key
   • Secret Access Key: Your IAM secret key
   • Region: AWS region (e.g., us-east-1)
   • Session Token (optional): For temporary credentials
5. Optionally select specific models
6. Click Save

Enable Models

Before using models, enable them in Bedrock:

1. Go to AWS Bedrock Console
2. In left sidebar, go to Model access
3. Click Enable for each model you want to use
4. Some models may require review and approval

Recommended Models

Latest Bedrock models with tool calling:

• Claude 3.5 Sonnet (Anthropic) — Latest, fast + capable
• Claude 3 Opus (Anthropic) — Most powerful reasoning
• Llama 3.1 70B (Meta) — Large, capable open model
• Mistral Large (Mistral) — Fast, capable

Model availability varies by region. Check Bedrock Models.

All Available Models

Your account has access to all enabled models supporting tool calling:

1. Go to Settings > AI Providers
2. Click Add Provider Key or Edit existing
3. Models dropdown shows all enabled models
4. Select specific models to restrict, or leave empty for all

Pricing

AWS Bedrock uses on-demand pricing:

• Per-token pricing: Charged by input and output tokens
• Varies by model: Different models have different costs
• No minimum commitment: Pay only for what you use

Check AWS Bedrock Pricing for current rates.

Security Best Practices

• Use IAM policies: Restrict access to specific models and regions
• Rotate keys: Periodically refresh access keys
• Use temporary credentials: Consider using STS for time-limited credentials
• CloudTrail logging: Enable CloudTrail for audit logging
• VPC endpoints: Use PrivateLink for private connectivity

Troubleshooting

"Invalid Access Key" or "AccessDenied"

• Verify Access Key ID and Secret Access Key are correct
• Check IAM user has AmazonBedrockFullAccess or equivalent permission
• Verify credentials haven't been revoked

"Invalid Region"

• Verify region format is correct (e.g., us-east-1)
• Check Bedrock is available in your region
• Some models may not be available in all regions

"Model Not Found" or "Model Access Denied"

• Enable the model in Model access in Bedrock console
• Some models require approval — wait for AWS review
• Verify your region supports the model
• Check IAM policy allows access to the specific model

"Session Token Expired"

If using temporary credentials:

• Refresh your session token
• Generate a new STS token
• Ensure token hasn't expired

Throttling or Rate Limits

For high-volume usage:

• Contact AWS to increase Bedrock quotas
• Implement exponential backoff for retries
• Consider batching requests