AWS Bedrock Integration

Sypha supports seamless integration with Amazon Bedrock, providing access to a broad spectrum of high-performing foundation models (FMs) via a single, unified API. This provider communicates directly with AWS infrastructure, authenticating via your localized AWS credentials.

Official Site: aws.amazon.com/bedrock

Pre-Configuration Requirements

• AWS Infrastructure: An active AWS account is required.
• Service Initialization: You must request and receive authorization for Amazon Bedrock. Refer to the AWS Bedrock Kickstart Guide for detailed instructions.
• Model Provisioning: Within the AWS Console, you must explicitly request access to specific model families (e.g., Anthropic Claude, Meta Llama).
• Environment Setup: Ensure your development machine is configured via the AWS CLI:

  aws configure

Authentication Framework

Sypha accommodates three primary authentication methodologies:

1. Bedrock-Specific API Keys:
   • Initialize a service-specific key within the AWS Console. This is an efficient, streamlined method for dedicated Bedrock access.
2. Standard AWS IAM Access Keys (Development Standard):
   • Create an IAM identity with the requisite permissions (specifically bedrock:InvokeModel).
   • Capture the Access Key ID and Secret Access Key.
3. AWS Profile Mapping:
   • Utilise an existing profile configured via the ~/.aws/credentials file. This is ideal for developers managing multiple AWS environments.

Optimized Model Registry

Sypha is optimized for several key model families on Bedrock:

• Amazon Nova & Titan: High-efficiency proprietary models.
• Anthropic Claude: Including Claude 3.7 Sonnet, 3.5 Sonnet (v2), and Haiku.
• Meta Llama: Ranging from the ultra-fast Llama 3.1 8B to the high-capacity 3.1 405B.
• DeepSeek: Including specialized reasoning models like R1.

Note: For a comprehensive list of current model IDs, consult the AWS Bedrock Supported Models Registry.

Configuring Sypha

1. Access Settings: Activate the gear icon in the Sypha interface.
2. Set Provider: Choose "Bedrock" from the provider catalog.
3. Select Auth Strategy:
   • API Key: Enter your Bedrock-specific key.
   • IAM Credentials: Input your Access Key, Secret, and (optionally) Session Token.
   • Profile: Specify your local profile identifier (e.g., default).
4. Choose Region: Designate the AWS region hosting your Bedrock resources (e.g., us-east-1).
5. Cross-Region Inference: Toggle this option if you need to access models beyond your primary configured region.
6. Assign Primary Model: Choose your intended engine from the selector.

Operational Strategy

• Granular Permissions: Ensure your IAM identity possesses the mandatory bedrock:InvokeModel policy.
• Economic Planning: Model expenditures are billed directly through your AWS account. View the Amazon Bedrock Rate Table for details.
• Latency Considerations: Cross-region inference may introduce marginal interaction delays.