Terraform Bedrock Module

Amazon Bedrock is a fully managed service that offers a choice of foundation models (FMs) along with a broad set of capabilities for building generative AI applications.

This module includes resources to deploy Bedrock features.

Knowledge Bases

With Knowledge Bases for Amazon Bedrock, you can give FMs and agents contextual information from your company’s private data sources for Retrieval Augmented Generation (RAG) to deliver more relevant, accurate, and customized responses.

Create a Knowledge Base

A vector index on a vector store is required to create a Knowledge Base. This construct currently supports Amazon OpenSearch Serverless, Amazon RDS Aurora PostgreSQL, Pinecone, and MongoDB. By default, this resource will create an OpenSearch Serverless vector collection and index for each Knowledge Base you create, but you can provide an existing collection to have more control. For other resources you need to have the vector stores already created and credentials stored in AWS Secrets Manager.

The resource accepts an instruction prop that is provided to any Bedrock Agent it is associated with so the agent can decide when to query the Knowledge Base.

To create a knowledge base, make sure you pass in the appropriate variables and set the create_kb variable to true.

Example default Opensearch Serverless Agent with Knowledgebase

provider "opensearch" {
  url         = module.bedrock.default_collection[0].collection_endpoint
  healthcheck = false
}

module "bedrock" {
  source  = "aws-ia/bedrock/aws"
  version = "0.0.5"
  create_kb = true
  create_default_kb = true
  foundation_model = "anthropic.claude-v2"
  instruction = "You are an automotive assisant who can provide detailed information about cars to a customer."
}

Knowledge Base - Data Sources

Data sources are the various repositories or systems from which information is extracted and ingested into the knowledge base. These sources provide the raw content that will be processed, indexed, and made available for querying within the knowledge base system. Data sources can include various types of systems such as document management systems, databases, file storage systems, and content management platforms. Suuported Data Sources include Amazon S3 buckets, Web Crawlers, SharePoint sites, Salesforce instances, and Confluence spaces.

• Amazon S3. You can either create a new data source by passing in the existing data source arn to the input variable kb_s3_data_source or create a new one by setting create_s3_data_source to true.

• Web Crawler. You can create a new web crawler data source by setting the create_web_crawler input variable to true and passing in the necessary variables for urls, scope, etc.

• SharePoint. You can create a new SharePoint data source by setting the create_sharepoint input variable to true and passing in the necessary variables for site urls, filter patterns, etc.

• Salesforce. You can create a new Salesforce data source by setting the create_salesforce input variable to true and passing in the necessary variables for site urls, filter patterns, etc.

• Confluence. You can create a new Confluence data source by setting the create_confluence input variable to true and passing in the necessary variables for site urls, filter patterns, etc.

Agents

Enable generative AI applications to execute multistep tasks across company systems and data sources.

Create an Agent

The following example creates an Agent with a simple instruction and without any action groups or knowedlge bases.

module "bedrock" {
  source  = "aws-ia/bedrock/aws"
  version = "0.0.5"
  foundation_model = "anthropic.claude-v2"
  instruction = "You are an automotive assisant who can provide detailed information about cars to a customer."
}

To create an Agent with a default Knowledge Base you simply set create_kb and create_default_kb to true:

module "bedrock" {
  source  = "aws-ia/bedrock/aws"
  version = "0.0.5"
  create_kb = true
  create_default_kb = true
  foundation_model = "anthropic.claude-v2"
  instruction = "You are an automotive assisant who can provide detailed information about cars to a customer."
}

Action Groups

An action group defines functions your agent can call. The functions are Lambda functions. The action group uses an OpenAPI schema to tell the agent what your functions do and how to call them. You can configure an action group by passing in the appropriate input variables.

Prepare the Agent

The Agent constructs take an optional parameter shouldPrepareAgent to indicate that the Agent should be prepared after any updates to an agent, Knowledge Base association, or action group. This may increase the time to create and update those resources. By default, this value is true.

Prompt Overrides

Bedrock Agents allows you to customize the prompts and LLM configuration for its different steps. You can disable steps or create a new prompt template. Prompt templates can be inserted from plain text files.

Bedrock Guardrails

Amazon Bedrock's Guardrails feature enables you to implement robust governance and control mechanisms for your generative AI applications, ensuring alignment with your specific use cases and responsible AI policies. Guardrails empowers you to create multiple tailored policy configurations, each designed to address the unique requirements and constraints of different use cases. These policy configurations can then be seamlessly applied across multiple foundation models (FMs) and Agents, ensuring a consistent user experience and standardizing safety, security, and privacy controls throughout your generative AI ecosystem.

With Guardrails, you can define and enforce granular, customizable policies to precisely govern the behavior of your generative AI applications. You can configure the following policies in a guardrail to avoid undesirable and harmful content and remove sensitive information for privacy protection.

Content filters – Adjust filter strengths to block input prompts or model responses containing harmful content.

Denied topics – Define a set of topics that are undesirable in the context of your application. These topics will be blocked if detected in user queries or model responses.

Word filters – Configure filters to block undesirable words, phrases, and profanity. Such words can include offensive terms, competitor names etc.

Sensitive information filters – Block or mask sensitive information such as personally identifiable information (PII) or custom regex in user inputs and model responses.

You can create a Guardrail by setting create_guardrail to true and passing in the appropriate input variables:

module "bedrock" {
  source  = "aws-ia/bedrock/aws"
  version = "0.0.5"
  create_kb = false
  create_default_kb = false
  create_guardrail = true
  blocked_input = "I can provide general info about services, but can't fully address your request here. For personalized help or detailed questions, please contact our customer service team directly. For security reasons, avoid sharing sensitive information through this channel. If you have a general product question, feel free to ask without including personal details."
  blocked_output = "I can provide general info about services, but can't fully address your request here. For personalized help or detailed questions, please contact our customer service team directly. For security reasons, avoid sharing sensitive information through this channel. If you have a general product question, feel free to ask without including personal details."
  filters_config = [
      {
        input_strength  = "MEDIUM"
        output_strength = "MEDIUM"
        type            = "HATE"
      },
      {
        input_strength  = "HIGH"
        output_strength = "HIGH"
        type            = "VIOLENCE"
      }
  ]
  pii_entities_config = [
      {
        action = "BLOCK"
        type   = "NAME"
      },
      {
        action = "BLOCK"
        type   = "DRIVER_ID"
      },
      {
        action = "ANONYMIZE"
        type   = "USERNAME"
      },
  ]
  regexes_config = [{
      action      = "BLOCK"
      description = "example regex"
      name        = "regex_example"
      pattern     = "^\\d{3}-\\d{2}-\\d{4}$"
  }]
  managed_word_lists_config = [{
      type = "PROFANITY"
  }]
  words_config = [{
    text = "HATE"
  }]
  topics_config = [{
      name       = "investment_topic"
      examples   = ["Where should I invest my money ?"]
      type       = "DENY"
      definition = "Investment advice refers to inquiries, guidance, or recommendations regarding the management or allocation of funds or assets with the goal of generating returns ."
  }]
  foundation_model = "anthropic.claude-v2"
  instruction = "You are an automotive assisant who can provide detailed information about cars to a customer."
}

Prompt Management

Amazon Bedrock provides the ability to create and save prompts using Prompt management so that you can save time by applying the same prompt to different workflows. You can include variables in the prompt so that you can adjust the prompt for different use case.

Prompt Variants

Prompt variants in the context of Amazon Bedrock refer to alternative configurations of a prompt, including its message or the model and inference configurations used. Prompt variants allow you to create different versions of a prompt, test them, and save the variant that works best for your use case. You can add prompt variants to a prompt by passing in the values for the ```variants_list

```hcl
  variants_list = [
    {
      name          = "variant-example"
      template_type = "TEXT"
      model_id      = "amazon.titan-text-express-v1"
      inference_configuration = {
        text = {
          temperature    = 1
          top_p          = 0.9900000095367432
          max_tokens     = 300
          stop_sequences = ["User:"]
          top_k          = 250
        }
      }
      template_configuration = {
        text = {
          input_variables = [
            {
              name = "topic"
            }
          ]
          text = "Make me a {{genre}} playlist consisting of the following number of songs: {{number}}."
        }
      }
    }
  ]

Prompt Version

A prompt version is a snapshot of a prompt at a specific point in time that you create when you are satisfied with a set of configurations. Versions allow you to deploy your prompt and easily switch between different configurations for your prompt and update your application with the most appropriate version for your use-case.

You can create a Prompt version by setting create_prompt_version to true and adding an optional prompt_version_description and optional prompt_version_tags.

Creating a prompt with a prompt version would look like:

module "bedrock" {
  source = "../.." # local example
  create_kb = false
  create_default_kb = false
  create_s3_data_source = false
  create_agent = false

  # Prompt Management
  prompt_name = "prompt"
  default_variant = "variant-example"
  create_prompt = true
  create_prompt_version = true
  prompt_version_description = "Example prompt version"
  variants_list = [
    {
      name          = "variant-example"
      template_type = "TEXT"
      model_id      = "amazon.titan-text-express-v1"
      inference_configuration = {
        text = {
          temperature    = 1
          top_p          = 0.9900000095367432
          max_tokens     = 300
          stop_sequences = ["User:"]
          top_k          = 250
        }
      }
      template_configuration = {
        text = {
          input_variables = [
            {
              name = "topic"
            }
          ]
          text = "Make me a {{genre}} playlist consisting of the following number of songs: {{number}}."
        }
      }
    }

  ]

}

Requirements

Name	Version
terraform	>= 1.0.7
aws	~>5.0
awscc	>= 1.0.0
opensearch	= 2.2.0
random	>= 3.6.0
time	~> 0.6

Providers

Name	Version
aws	~>5.0
awscc	>= 1.0.0
opensearch	= 2.2.0
random	>= 3.6.0
time	~> 0.6

Modules

No modules.

Resources

Name	Type
aws_bedrockagent_data_source.knowledge_base_ds	resource
aws_cloudwatch_log_group.knowledge_base_cwl	resource
aws_iam_policy.bedrock_kb_s3_decryption_policy	resource
aws_iam_policy.bedrock_knowledge_base_policy	resource
aws_iam_policy.bedrock_knowledge_base_policy_s3	resource
aws_iam_role.agent_role	resource
aws_iam_role.bedrock_knowledge_base_role	resource
aws_iam_role_policy.action_group_policy	resource
aws_iam_role_policy.agent_policy	resource
aws_iam_role_policy.bedrock_kb_oss	resource
aws_iam_role_policy.guardrail_policy	resource
aws_iam_role_policy.kb_policy	resource
aws_iam_role_policy_attachment.bedrock_kb_s3_decryption_policy_attachment	resource
aws_iam_role_policy_attachment.bedrock_knowledge_base_policy_attachment	resource
aws_iam_role_policy_attachment.bedrock_knowledge_base_policy_s3_attachment	resource
aws_lambda_permission.allow_bedrock_agent	resource
aws_opensearchserverless_access_policy.data_policy	resource
aws_opensearchserverless_security_policy.nw_policy	resource
aws_opensearchserverless_security_policy.security_policy	resource
awscc_bedrock_agent.bedrock_agent	resource
awscc_bedrock_agent_alias.bedrock_agent_alias	resource
awscc_bedrock_data_source.knowledge_base_confluence	resource
awscc_bedrock_data_source.knowledge_base_salesforce	resource
awscc_bedrock_data_source.knowledge_base_sharepoint	resource
awscc_bedrock_data_source.knowledge_base_web_crawler	resource
awscc_bedrock_guardrail.guardrail	resource
awscc_bedrock_guardrail_version.guardrail	resource
awscc_bedrock_knowledge_base.knowledge_base_default	resource
awscc_bedrock_knowledge_base.knowledge_base_mongo	resource
awscc_bedrock_knowledge_base.knowledge_base_opensearch	resource
awscc_bedrock_knowledge_base.knowledge_base_pinecone	resource
awscc_bedrock_knowledge_base.knowledge_base_rds	resource
awscc_bedrock_prompt.prompt	resource
awscc_bedrock_prompt_version.prompt_version	resource
awscc_logs_delivery.knowledge_base_log_delivery	resource
awscc_logs_delivery_destination.knowledge_base_log_destination	resource
awscc_logs_delivery_source.knowledge_base_log_source	resource
awscc_opensearchserverless_collection.default_collection	resource
awscc_s3_bucket.s3_data_source	resource
opensearch_index.default_oss_index	resource
random_string.solution_prefix	resource
time_sleep.wait_after_index_creation	resource
time_sleep.wait_before_index_creation	resource
aws_caller_identity.current	data source
aws_iam_policy_document.agent_permissions	data source
aws_iam_policy_document.agent_trust	data source
aws_iam_policy_document.knowledge_base_permissions	data source
aws_partition.current	data source
aws_region.current	data source

Inputs

Name	Description	Type	Default	Required
action_group_description	Description of the action group.	string	null	no
action_group_name	Name of the action group.	string	null	no
action_group_state	State of the action group.	string	null	no
agent_alias_description	Description of the agent alias.	string	null	no
agent_alias_name	The name of the guardrail.	string	"TerraformBedrockAgentAlias"	no
agent_alias_tags	Tag bedrock agent alias resource.	map(string)	null	no
agent_description	A description of agent.	string	null	no
agent_id	Agent identifier.	string	null	no
agent_name	The name of your agent.	string	"TerraformBedrockAgents"	no
api_schema_payload	String OpenAPI Payload.	string	null	no
api_schema_s3_bucket_name	A bucket in S3.	string	null	no
api_schema_s3_object_key	An object key in S3.	string	null	no
auth_type	The supported authentication type.	string	null	no
base_prompt_template	Defines the prompt template with which to replace the default prompt template.	string	null	no
bedrock_agent_version	Agent version.	string	null	no
blocked_input_messaging	Messaging for when violations are detected in text.	string	"Blocked input"	no
blocked_outputs_messaging	Messaging for when violations are detected in text.	string	"Blocked output"	no
bucket_owner_account_id	Bucket account owner ID for the S3 bucket.	string	null	no
collection_arn	The ARN of the collection.	string	null	no
collection_name	The name of the collection.	string	null	no
confluence_credentials_secret_arn	The ARN of an AWS Secrets Manager secret that stores your authentication credentials for your Confluence instance URL.	string	null	no
connection_string	The endpoint URL for your index management page.	string	null	no
crawl_filter_type	The crawl filter type.	string	null	no
crawler_scope	The scope that a web crawl job will be restricted to.	string	null	no
create_ag	Whether or not to create an action group.	bool	false	no
create_agent	Whether or not to deploy an agent.	bool	true	no
create_agent_alias	Whether or not to create an agent alias.	bool	false	no
create_confluence	Whether or not create a Confluence data source.	bool	false	no
create_default_kb	Whether or not to create the default knowledge base.	bool	false	no
create_guardrail	Whether or not to create a guardrail.	bool	false	no
create_kb	Whether or not to attach a knowledge base.	bool	false	no
create_kb_log_group	Whether or not to create a log group for the knowledge base.	bool	false	no
create_mongo_config	Whether or not to use MongoDB Atlas configuration	bool	false	no
create_opensearch_config	Whether or not to use Opensearch Serverless configuration	bool	false	no
create_pinecone_config	Whether or not to use Pinecone configuration	bool	false	no
create_prompt	Whether or not to create a prompt resource.	bool	false	no
create_prompt_version	Whether or not to create a prompt version.	bool	false	no
create_rds_config	Whether or not to use RDS configuration	bool	false	no
create_s3_data_source	Whether or not to create the S3 data source.	bool	true	no
create_salesforce	Whether or not create a Salesforce data source.	bool	false	no
create_sharepoint	Whether or not create a Share Point data source.	bool	false	no
create_web_crawler	Whether or not create a web crawler data source.	bool	false	no
credentials_secret_arn	The ARN of the secret in Secrets Manager that is linked to your database	string	null	no
custom_control	Custom control of action execution.	string	null	no
customer_encryption_key_arn	A KMS key ARN.	string	null	no
database_name	Name of the database.	string	null	no
default_variant	Name for a variant.	string	null	no
endpoint	Database endpoint	string	null	no
endpoint_service_name	MongoDB Atlas endpoint service name.	string	null	no
exclusion_filters	A set of regular expression filter patterns for a type of object.	list(string)	[]	no
existing_kb	The ID of the existing knowledge base.	string	null	no
filters_config	List of content filter configs in content policy.	list(map(string))	null	no
foundation_model	The foundation model for the Bedrock agent.	string	null	no
guardrail_description	Description of the guardrail.	string	null	no
guardrail_kms_key_arn	KMS encryption key to use for the guardrail.	string	null	no
guardrail_name	The name of the guardrail.	string	"TerraformBedrockGuardrail"	no
guardrail_tags	A map of tags keys and values for the knowledge base.	list(map(string))	null	no
host_type	The supported host type, whether online/cloud or server/on-premises.	string	null	no
host_url	The host URL or instance URL.	string	null	no
idle_session_ttl	How long sessions should be kept open for the agent.	number	600	no
inclusion_filters	A set of regular expression filter patterns for a type of object.	list(string)	[]	no
instruction	A narrative instruction to provide the agent as context.	string	""	no
kb_description	Description of knowledge base.	string	"Terraform deployed Knowledge Base"	no
kb_embedding_model_arn	The ARN of the model used to create vector embeddings for the knowledge base.	string	"arn:aws:bedrock:us-east-1::foundation-model/amazon.titan-embed-text-v1"	no
kb_log_group_retention_in_days	The retention period of the knowledge base log group.	number	0	no
kb_monitoring_arn	The ARN of the target for delivery of knowledge base application logs	string	null	no
kb_name	Name of the knowledge base.	string	"knowledge-base"	no
kb_role_arn	The ARN of the IAM role with permission to invoke API operations on the knowledge base.	string	null	no
kb_s3_data_source	The S3 data source ARN for the knowledge base.	string	null	no
kb_s3_data_source_kms_arn	The ARN of the KMS key used to encrypt S3 content	string	null	no
kb_state	State of knowledge base; whether it is enabled or disabled	string	"ENABLED"	no
kb_storage_type	The storage type of a knowledge base.	string	null	no
kb_tags	A map of tags keys and values for the knowledge base.	map(string)	null	no
kb_type	The type of a knowledge base.	string	null	no
kms_key_arn	KMS encryption key to use for the agent.	string	null	no
lambda_action_group_executor	ARN of Lambda.	string	null	no
managed_word_lists_config	A config for the list of managed words.	list(map(string))	null	no
max_length	The maximum number of tokens to generate in the response.	number	0	no
metadata_field	The name of the field in which Amazon Bedrock stores metadata about the vector store.	string	"AMAZON_BEDROCK_METADATA"	no
name_prefix	This value is appended at the beginning of resource names.	string	"BedrockAgents"	no
namespace	The namespace to be used to write new data to your pinecone database	string	null	no
override_lambda_arn	The ARN of the Lambda function to use when parsing the raw foundation model output in parts of the agent sequence.	string	null	no
parent_action_group_signature	Action group signature for a builtin action.	string	null	no
parser_mode	Specifies whether to override the default parser Lambda function.	string	null	no
pattern_object_filter_list	List of pattern object information.	list(object({ exclusion_filters = optional(list(string)) inclusion_filters = optional(list(string)) object_type = optional(string) }))	[]	no
pii_entities_config	List of entities.	list(map(string))	null	no
primary_key_field	The name of the field in which Bedrock stores the ID for each entry.	string	null	no
prompt_creation_mode	Specifies whether to override the default prompt template.	string	null	no
prompt_description	Description for a prompt resource.	string	null	no
prompt_name	Name for a prompt resource.	string	null	no
prompt_override	Whether to provide prompt override configuration.	bool	false	no
prompt_state	Specifies whether to allow the agent to carry out the step specified in the promptType.	string	null	no
prompt_tags	A map of tag keys and values for prompt resource.	map(string)	null	no
prompt_type	The step in the agent sequence that this prompt configuration applies to.	string	null	no
prompt_version_description	Description for a prompt version resource.	string	null	no
prompt_version_tags	A map of tag keys and values for a prompt version resource.	map(string)	null	no
rate_limit	Rate of web URLs retrieved per minute.	number	null	no
regexes_config	List of regex.	list(map(string))	null	no
resource_arn	The ARN of the vector store.	string	null	no
s3_inclusion_prefixes	List of S3 prefixes that define the object containing the data sources.	list(string)	null	no
salesforce_credentials_secret_arn	The ARN of an AWS Secrets Manager secret that stores your authentication credentials for your Salesforce instance URL.	string	null	no
seed_urls	A list of web urls.	list(object({url = string}))	[]	no
share_point_credentials_secret_arn	The ARN of an AWS Secrets Manager secret that stores your authentication credentials for your SharePoint site/sites.	string	null	no
share_point_domain	The domain of your SharePoint instance or site URL/URLs.	string	null	no
share_point_site_urls	A list of one or more SharePoint site URLs.	list(string)	[]	no
skip_resource_in_use	Specifies whether to allow deleting action group while it is in use.	bool	null	no
stop_sequences	A list of stop sequences.	list(string)	[]	no
table_name	The name of the table in the database.	string	null	no
tags	Tag bedrock agent resource.	map(string)	null	no
temperature	The likelihood of the model selecting higher-probability options while generating a response.	number	0	no
tenant_id	The identifier of your Microsoft 365 tenant.	string	null	no
text_field	The name of the field in which Amazon Bedrock stores the raw text from your data.	string	"AMAZON_BEDROCK_TEXT_CHUNK"	no
top_k	Sample from the k most likely next tokens.	number	50	no
top_p	Cumulative probability cutoff for token selection.	number	0.5	no
topics_config	List of topic configs in topic policy	list(object({ name = string examples = list(string) type = string definition = string }))	null	no
variants_list	List of prompt variants.	list(object({ name = optional(string) template_type = optional(string) model_id = optional(string) inference_configuration = optional(object({ text = optional(object({ max_tokens = optional(number) stop_sequences = optional(list(string)) temperature = optional(number) top_p = optional(number) })) })) template_configuration = optional(object({ text = optional(object({ input_variables = optional(list(object({ name = optional(string) }))) text = optional(string) text_s3_location = optional(object({ bucket = optional(string) key = optional(string) version = optional(string) })) })) })) }))	null	no
vector_field	The name of the field where the vector embeddings are stored	string	"bedrock-knowledge-base-default-vector"	no
vector_index_name	The name of the vector index.	string	"bedrock-knowledge-base-default-index"	no
words_config	List of custom word configs.	list(map(string))	null	no

Outputs

Name	Description
bedrock_agent	The Amazon Bedrock Agent if it is created.
cloudwatch_log_group	The name of the CloudWatch log group for the knowledge base. If no log group was requested, value will be null
datasource_identifier	The unique identifier of the data source.
default_collection	Opensearch default collection value.
default_kb_identifier	The unique identifier of the default knowledge base that was created. If no default KB was requested, value will be null
mongo_kb_identifier	The unique identifier of the MongoDB knowledge base that was created. If no MongoDB KB was requested, value will be null
opensearch_kb_identifier	The unique identifier of the OpenSearch knowledge base that was created. If no OpenSearch KB was requested, value will be null
pinecone_kb_identifier	The unique identifier of the Pinecone knowledge base that was created. If no Pinecone KB was requested, value will be null
rds_kb_identifier	The unique identifier of the RDS knowledge base that was created. If no RDS KB was requested, value will be null
s3_data_source_arn	The Amazon Bedrock Data Source for S3.