Terraform module for Amazon Bedrock Agent resources
| Name | Version |
|---|---|
| aws | ~> 5.94 |
| opensearch | ~> 2.3 |
| time | ~> 0.13 |
| Name | Version |
|---|---|
| terraform | ~> 1.3 |
| aws | ~> 5.94 |
| opensearch | ~> 2.3 |
| time | ~> 0.13 |
The following input variables are required:
Description: Name for the agent.
Type: string
Description: Name for the agent alias.
Type: string
Description: Model identifier for agent.
Type: string
Description: Name for the knowledgebase.
Type: string
Description: Description for the knowledgebase.
Type: string
Description: ARN of S3 bucket with data
Type:
object({
bucket_arn = string
bucket_owner_account_id = optional(string)
inclusion_prefixes = optional(set(string))
})Description: Name of OpenSearch Serverless Collection.
Type: string
The following input variables are optional (have default values):
Description: Description for the agent alias.
Type: string
Default: null
Description: Model identifier for agent.
Type: string
Default: "anthropic.claude-v2"
Description: Model identifier for Knowledgebase.
Type: string
Default: "amazon.titan-embed-text-v1"
Description: Data deletion policy for a data source. Valid values: RETAIN, DELETE
Type: string
Default: "RETAIN"
Description: n/a
Type:
object({
chunking_configuration = object({
chunking_strategy = string
fixed_size_chunking_configuration = optional(object({
max_tokens = number
overlap_percentage = optional(number)
}))
hierarchical_chunking_configuration = optional(object({
overlap_tokens = number
level_1 = object({ max_tokens = number })
level_2 = object({ max_tokens = number })
}))
semantic_chunking_configuration = optional(object({
breakpoint_percentile_threshold = number
buffer_size = number
max_token = number
}))
})
custom_transformation_configuration = optional(object({
intermediate_storage = string
transformation_function = string
}))
})Default:
{
"chunking_configuration": {
"chunking_strategy": "FIXED_SIZE",
"fixed_size_chunking_configuration": {
"max_tokens": 300,
"overlap_percentage": 20
},
"hierarchical_chunking_configuration": null,
"semantic_chunking_configuration": null
}
}Description: Additional ARNs of roles to access OpenSearch
Type: list(string)
Default: []
Description: Prompt template for pre-processing.
Type: string
Default: " You are a helpful assistant. Answer the following question using the context provided:\n Question: {question}\n Context: {context}\n Your response should be thoughtful, detailed, and relevant to the provided context.\n"
Description: Parser mode for pre-processing.
Type: string
Default: "DEFAULT"
Description: Prompt creation mode for pre-processing.
Type: string
Default: "OVERRIDDEN"
Description: Prompt state for pre-processing.
Type: string
Default: "ENABLED"
Description: Maximum number of tokens to allow in the generated response.
Type: number
Default: 512
Description: List of stop sequences that will stop generation.
Type: list(string)
Default:
[
"END"
]Description: Likelihood of the model selecting higher-probability options while generating a response.
Type: number
Default: 0.7
Description: Number of top most-likely candidates from which the model chooses the next token.
Type: number
Default: 50
Description: Top percentage of the probability distribution of next tokens, from which the model chooses the next token.
Type: number
Default: 0.9
Description: Prompt template for pre-processing.
Type: string
Default: " You are preparing the input. Extract relevant context and pre-process the following question:\n Question: {question}\n Context: {context}\n Pre-processing should focus on extracting the core information.\n"
Description: Parser mode for pre-processing.
Type: string
Default: "DEFAULT"
Description: Prompt creation mode for pre-processing.
Type: string
Default: "OVERRIDDEN"
Description: Prompt state for pre-processing.
Type: string
Default: "ENABLED"
Description: Maximum number of tokens to allow in the generated response.
Type: number
Default: 512
Description: List of stop sequences that will stop generation.
Type: list(string)
Default:
[
"END"
]Description: Likelihood of the model selecting higher-probability options while generating a response.
Type: number
Default: 0.7
Description: Number of top most-likely candidates from which the model chooses the next token.
Type: number
Default: 50
Description: Top percentage of the probability distribution of next tokens, from which the model chooses the next token.
Type: number
Default: 0.9
Description: Prompt template for orchestration.
Type: string
Default: " You are orchestrating the flow of the agent. Based on the question and context, determine the next steps in the process:\n Question: {question}\n Context: {context}\n Plan the next steps to follow the best strategy.\n"
Description: Parser mode for orchestration.
Type: string
Default: "DEFAULT"
Description: Prompt creation mode for orchestration.
Type: string
Default: "OVERRIDDEN"
Description: Prompt state for orchestration.
Type: string
Default: "ENABLED"
Description: Maximum number of tokens to allow in the generated response.
Type: number
Default: 512
Description: List of stop sequences that will stop generation.
Type: list(string)
Default:
[
"END"
]Description: Likelihood of the model selecting higher-probability options while generating a response.
Type: number
Default: 0.7
Description: Number of top most-likely candidates from which the model chooses the next token.
Type: number
Default: 50
Description: Top percentage of the probability distribution of next tokens, from which the model chooses the next token.
Type: number
Default: 0.9
Description: Prompt template for post-processing.
Type: string
Default: "You are performing post-processing. Review the agent's output and refine the response for clarity and relevance:\nResponse: {response}\nContext: {context}\nEnsure the output is polished and aligns with the context.\n"
Description: Parser mode for post-processing.
Type: string
Default: "DEFAULT"
Description: Prompt creation mode for post-processing.
Type: string
Default: "OVERRIDDEN"
Description: Prompt state for post-processing.
Type: string
Default: "DISABLED"
Description: Maximum number of tokens to allow in the generated response.
Type: number
Default: 512
Description: List of stop sequences that will stop generation.
Type: list(string)
Default:
[
"END"
]Description: Likelihood of the model selecting higher-probability options while generating a response.
Type: number
Default: 0.7
Description: Number of top most-likely candidates from which the model chooses the next token.
Type: number
Default: 50
Description: Top percentage of the probability distribution of next tokens, from which the model chooses the next token.
Type: number
Default: 0.9
Description: A map of tags to assign to the customization job and custom model.
Type: map(string)
Default: {}
| Name | Description |
|---|---|
| agent | Information about created Bedrock Agent |
| agent_alias | Information about created Bedrock Agent Alias |
| knowledge_base | Information about created Bedrock Knowledgebase |
| oss_collection | Information about created OpenSearch Serverless collection |
Copyright (c) 2024 Flaconi GmbH