Terraform AWS Bedrock Module

Overview

AWS Bedrock is a fully managed generative AI service that simplifies building and scaling AI applications by providing access to leading foundation models (FMs) through a unified API endpoint. This Terraform module facilitates the deployment and management of AWS Bedrock resources with enterprise-grade security and flexibility.

Key Features

• Budget Management for AI Resource Consumption
• Configurable Guardrails for Content Safety
• Multiple Inference Profiles for Different AI Models
• Flexible Tagging for Resource Management

Usage Example

module "bedrock" {
  source = "DNXLabs/bedrock/aws"

  # Budget Configuration
  budget_email = "contact@example.com"
  budget       = 500  # Monthly budget in USD

  # Guardrails Configuration
  guardrails = [
    {
      name            = "content-safety-guardrail"
      description     = "Prevent inappropriate content"
      filter_strength = "MEDIUM"
      filters         = ["SEXUAL", "VIOLENCE", "HATE", "INSULTS", "MISCONDUCT", "PROMPT_ATTACK"]
      tags = {
        "Unit" = "Cloud Operations Centre"
      }
    }
  ]

  # Inference Profiles for Different Models
  profiles = [
    {
      name                        = "claude-3-5-sonnet-profile"
      description                 = "Claude 3.5 Sonnet Inference Profile"
      source_inference_model_name = "us.anthropic.claude-3-5-sonnet-20241022-v2:0"
      tags = {
        "Unit" = "Cloud Operations Centre"
      }
    },
    {
      name                        = "claude-3-7-sonnet-profile"
      description                 = "Claude 3.7 Sonnet Inference Profile"
      source_inference_model_name = "us.anthropic.claude-3-7-sonnet-20250219-v1:0"
      tags = {
        "Unit" = "Cloud Operations Centre"
      }
    }
  ]
}

# Outputs
output "profile_arns" {
  description = "ARNs of created Bedrock inference profiles"
  value       = module.bedrock.profile_arns
}

output "guardrail_arns" {
  description = "ARNs of created Bedrock guardrails"
  value       = module.bedrock.guardrail_arns
}

output "guardrail_versions" {
  description = "Versions of created Bedrock guardrails"
  value       = module.bedrock.guardrail_versions
}

Requirements

Name	Version
Terraform	>= 0.13.0
AWS Provider	6.0.0-beta1

Inputs

Name	Description	Type	Default	Required
budget_email	Email address to send budget alerts	string	n/a	yes
budget	Monthly budget for AI resources in USD	number	n/a	yes
guardrails	List of guardrail configurations	list(object({...}))	[]	no
profiles	List of inference profile configurations	list(object({...}))	[]	no

Outputs

Name	Description	Type
profile_arns	List of created Bedrock inference profile ARNs	list(string)
guardrail_arns	List of created Bedrock guardrail ARNs	list(string)
guardrail_versions	Versions of created Bedrock guardrails	list(string)

Guardrails Configuration

Field	Description	Type
name	Name of the guardrail	string
description	Description of the guardrail	string
filter_strength	Strength of content filtering (e.g., LOW, MEDIUM, HIGH)	string
filters	Types of content to filter	list(string)
tags	Optional tags for the guardrail	map(string)

Profiles Configuration

Field	Description	Type
name	Name of the inference profile	string
description	Description of the profile	string
source_inference_model_name	ARN or name of the source model	string
tags	Optional tags for the profile	map(string)

Best Practices

1. Use comprehensive guardrails to ensure content safety
2. Configure multiple inference profiles for flexibility
3. Set appropriate budget alerts
4. Use tags for better resource management and tracking
5. Regularly review and update model configurations

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for details on how to contribute to this project.

Authors

Module managed by DNX Solutions.

License

Apache 2 Licensed. See LICENSE for full details.

Support

If you encounter any issues or have questions, please file an issue on the GitHub repository.