Module Awso_bedrock_asyncSource

The request is denied because of missing access permissions.

Selective content guarding controls for enforced guardrails.

Model-specific information for the enforced guardrail configuration.

Account-level enforced guardrail input configuration.

Account enforced guardrail output configuration.

Input data configuration for the advanced prompt optimization job.

ARN of the advanced prompt optimization job.

Description of the advanced prompt optimization job.

ARN or ID of the advanced prompt optimization job.

Name of the advanced prompt optimization job.

Summary of an advanced prompt optimization job.

Output data configuration for the advanced prompt optimization job.

Information about the agreement availability

The evaluator model used in knowledge base evaluation job or in model evaluation job that use a model as judge. This model computes all evaluation related metrics.

Specifies the model configuration for the evaluator model. EvaluatorModelConfig is required for evaluation jobs that use a knowledge base or in model evaluation job that use a model as judge. This model computes all evaluation related metrics.

The location in Amazon S3 where your prompt dataset is stored.

Used to specify the name of a built-in prompt dataset and optionally, the Amazon S3 bucket where a custom prompt dataset is saved.

Defines the prompt datasets, built-in metric names and custom metric names, and the task type.

Defines the model you want to evaluate custom metrics in an Amazon Bedrock evaluation job.

Configuration of the evaluator model you want to use to evaluate custom metrics in an Amazon Bedrock evaluation job.

Defines the value for one rating in a custom metric rating scale.

Defines the value and corresponding definition for one rating in a custom metric rating scale.

The definition of a custom metric for use in an Amazon Bedrock evaluation job. A custom metric definition includes a metric name, prompt (instructions) and optionally, a rating scale. Your prompt must include a task description and input variables. The required input variables are different for model-as-a-judge and RAG evaluations. For more information about how to define a custom metric in Amazon Bedrock, see Create a prompt for a custom metrics (LLM-as-a-judge model evaluations) and Create a prompt for a custom metrics (RAG evaluations).

An array item definining a single custom metric for use in an Amazon Bedrock evaluation job.

Defines the configuration of custom metrics to be used in an evaluation job. To learn more about using custom metrics in Amazon Bedrock evaluation jobs, see Create a prompt for a custom metrics (LLM-as-a-judge model evaluations) and Create a prompt for a custom metrics (RAG evaluations).

The configuration details of an automated evaluation job. The EvaluationDatasetMetricConfig object is used to specify the prompt datasets, task type, and metric names.

Represents a logical statement that can be expressed both in formal logic notation and natural language, providing dual representations for better understanding and validation.

Represents a logical scenario where claims can be evaluated as true or false, containing specific logical assignments.

References a portion of the original input text that corresponds to logical elements.

Contains the logical translation of natural language input into formal logical statements, including premises, claims, and confidence scores.

References a specific automated reasoning policy rule that was applied during evaluation.

Identifies logical issues in the translated statements that exist independent of any policy rules, such as statements that are always true or always false.

Indicates that the claims are definitively true and logically implied by the premises, with no possible alternative interpretations.

Represents one possible logical interpretation of ambiguous input content.

Indicates that the input has multiple valid logical interpretations, requiring additional context or clarification.

Indicates that the input exceeds the processing capacity due to the volume or complexity of the logical information.

Indicates that the claims could be either true or false depending on additional assumptions not provided in the input.

Indicates that no relevant logical information could be extracted from the input for validation.

Indicates that the claims are logically false and contradictory to the established rules or premises.

Indicates that no valid claims can be made due to logical contradictions in the premises or rules.

Represents the result of an Automated Reasoning validation check, indicating whether the content is logically valid, invalid, or falls into other categories based on the policy rules.

An annotation for adding a new rule to an Automated Reasoning policy using a formal logical expression.

An annotation for adding a new rule to the policy by converting a natural language description into a formal logical expression.

Represents a formal logic rule in an Automated Reasoning policy. For example, rules can be expressed as if-then statements that define logical constraints.

A mutation operation that adds a new rule to the policy definition during the build process.

Represents a single value within a custom type definition, including its identifier and description.

An annotation for adding a new custom type to an Automated Reasoning policy, defining a set of possible values for variables.

Represents a custom user-defined viarble type in an Automated Reasoning policy. Types are enum-based and provide additional context beyond predefined variable types.

A mutation operation that adds a new custom type to the policy definition during the build process.

Represents a single value that can be added to an existing custom type in the policy.

An annotation for adding a new variable to an Automated Reasoning policy, which can be used in rule expressions.

Represents a variable in an Automated Reasoning policy. Variables represent concepts that can have values assigned during natural language translation.

A mutation operation that adds a new variable to the policy definition during the build process.

Represents a single line of text from a source document, annotated with its line number for precise referencing.

Represents a content element within an annotated chunk. This union type allows for different types of content elements to be included in document chunks, such as individual lines of text with their line numbers.

Represents a portion of a source document with line number annotations. Chunks help organize document content for easier navigation and reference.

An annotation for modifying an existing variable in an Automated Reasoning policy.

Represents a modification to a value within an existing custom type.

Represents a value to be removed from an existing custom type in the policy.

An annotation for managing values within custom types, including adding, updating, or removing specific type values.

An annotation for modifying an existing custom type in an Automated Reasoning policy.

An annotation for modifying an existing rule in an Automated Reasoning policy.

An annotation for updating the policy based on feedback about how it performed on specific test scenarios.

An annotation for updating the policy based on feedback about how specific rules performed during testing or real-world usage.

An annotation for processing and incorporating new content into an Automated Reasoning policy.

An annotation for removing a variable from an Automated Reasoning policy.

An annotation for removing a custom type from an Automated Reasoning policy.

An annotation for removing a rule from an Automated Reasoning policy.

Contains the various operations that can be performed on an Automated Reasoning policy, including adding, updating, and deleting rules, variables, and types.

Describes the location of a statement within a source document using line numbers.

Represents a single, indivisible statement extracted from a source document. Atomic statements are the fundamental units used to ground policy rules and variables to their source material.

Represents a single element in an Automated Reasoning policy definition, such as a rule, variable, or type definition.

Represents a message generated during a build step, providing information about what happened or any issues encountered.

Represents the planning phase of policy build workflow, where the system analyzes source content and determines what operations to perform.

A mutation operation that modifies an existing variable in the policy definition during the build process.

A mutation operation that modifies an existing custom type in the policy definition during the build process.

A mutation operation that modifies an existing rule in the policy definition during the build process.

A mutation operation that removes a variable from the policy definition during the build process.

A mutation operation that removes a custom type from the policy definition during the build process.

A mutation operation that removes a rule from the policy definition during the build process.

A container for various mutation operations that can be applied to an Automated Reasoning policy, including adding, updating, and deleting policy elements.

Provides context about what type of operation was being performed during a build step.

Represents a single step in the policy build process, containing context about what was being processed and any messages or results.

Represents a single entry in the policy build log, containing information about a specific step or event in the build process.

Contains detailed logging information about the policy build process, including steps taken, decisions made, and any issues encountered.

Represents a single entry in the asset manifest, describing one artifact produced by the build workflow.

A catalog of all artifacts produced by a build workflow, providing a comprehensive list of available assets including their types and identifiers.

Represents a source document that was processed during a build workflow. Contains the document content, metadata, and a hash for verification.

Represents a test scenario used to validate an Automated Reasoning policy, including the test conditions and expected outcomes.

Contains a comprehensive entity encompassing all the scenarios generated by the build workflow, which can be used to validate an Automated Reasoning policy.

Represents a generated test case, consisting of query content, guard content, and expected results.

Contains a comprehensive test suite generated by the build workflow, providing validation capabilities for automated reasoning policies.

References a specific atomic statement within a source document, used to link policy elements back to their source material.

Provides detailed fidelity analysis for a specific policy variable, including which source document statements support it and how accurate the variable definition is.

Provides detailed fidelity analysis for a specific policy rule, including which source document statements support it and how accurate the rule is.

Represents a source document that was analyzed during fidelity report generation, including the document's metadata and its content broken down into atomic statements.

A comprehensive analysis report that measures how accurately a generated policy represents the source documents. The report includes coverage and accuracy scores, detailed grounding information linking policy elements to source statements, and annotated document content.

Represents a set of rules that operate on completely separate variables, indicating they address different concerns or domains within the policy.

Associates a type name with a specific value name, used for referencing type values in rules and other policy elements.

Provides a comprehensive analysis of the quality and completeness of an Automated Reasoning policy definition, highlighting potential issues and optimization opportunities.

Contains the formal logic rules, variables, and custom variable types that define an Automated Reasoning policy. The policy definition specifies the constraints used to validate foundation model responses for accuracy and logical consistency.

Contains the various assets generated during a policy build workflow, including logs, quality reports, test cases, and the final policy definition.

Represents a source document used in the policy build workflow, containing the content and metadata needed for policy generation.

Contains content and instructions for repairing or improving an existing Automated Reasoning policy.

Configuration for generating a fidelity report, which can either analyze new documents or update an existing fidelity report with a new policy definition.

Defines the content and configuration for different types of policy build workflows.

Defines the source content for a policy build workflow, which can include documents, repair instructions, or other input materials.

Provides a summary of a policy build workflow, including its current status, timing information, and key identifiers.

Contains summary information about an Automated Reasoning policy, including metadata and timestamps.

Represents a test for validating an Automated Reasoning policy. tests contain sample inputs and expected outcomes to verify policy behavior.

Contains the results of testing an Automated Reasoning policy against various scenarios and validation checks.

Batch deletion error for an advanced prompt optimization job.

Successfully deleted advanced prompt optimization job.

Batch Delete Advanced Prompt Optimization Jobs Request

Input validation failed. Check your request parameters and retry the request.

The number of requests exceeds the limit. Resubmit your request later.

An internal server error occurred. Retry your request.

Batch Delete Advanced Prompt Optimization Jobs Response

A JSON array that provides the status of the evaluation jobs being deleted.

An evaluation job for deletion, and it’s current status.

Deletes a batch of evaluation jobs. An evaluation job can only be deleted if it has following status FAILED, COMPLETED, and STOPPED. You can request up to 25 model evaluation jobs be deleted in a single request.

The specified resource Amazon Resource Name (ARN) was not found. Check the Amazon Resource Name (ARN) and try your request again.

Error occurred because of a conflict while performing an operation.

Deletes a batch of evaluation jobs. An evaluation job can only be deleted if it has following status FAILED, COMPLETED, and STOPPED. You can request up to 25 model evaluation jobs be deleted in a single request.

Contains the document contained in the wrapper object, along with its attributes/fields.

Cancels a running Automated Reasoning policy build workflow. This stops the policy generation process and prevents further processing of the source documents.

Cancels a running Automated Reasoning policy build workflow. This stops the policy generation process and prevents further processing of the source documents.

S3 configuration for storing log data.

CloudWatch logging configuration.

Definition of the key/value pair for a tag.

Inference configuration for a model.

Configuration for a model used in advanced prompt optimization.

Create Advanced Prompt Optimization Job Request

The request contains more tags than can be associated with a resource (50 tags per resource). The maximum number of tags includes both existing tags and those included in your current request.

The number of requests exceeds the service quota. Resubmit your request later.

Create Advanced Prompt Optimization Job Response

Creates an Automated Reasoning policy for Amazon Bedrock Guardrails. Automated Reasoning policies use mathematical techniques to detect hallucinations, suggest corrections, and highlight unstated assumptions in the responses of your GenAI application. To create a policy, you upload a source document that describes the rules that you're encoding. Automated Reasoning extracts important concepts from the source document that will become variables in the policy and infers policy rules.

Creates an Automated Reasoning policy for Amazon Bedrock Guardrails. Automated Reasoning policies use mathematical techniques to detect hallucinations, suggest corrections, and highlight unstated assumptions in the responses of your GenAI application. To create a policy, you upload a source document that describes the rules that you're encoding. Automated Reasoning extracts important concepts from the source document that will become variables in the policy and infers policy rules.

Creates a test for an Automated Reasoning policy. Tests validate that your policy works as expected by providing sample inputs and expected outcomes. Use tests to verify policy behavior before deploying to production.

Creates a test for an Automated Reasoning policy. Tests validate that your policy works as expected by providing sample inputs and expected outcomes. Use tests to verify policy behavior before deploying to production.

Creates a new version of an existing Automated Reasoning policy. This allows you to iterate on your policy rules while maintaining previous versions for rollback or comparison purposes.

Creates a new version of an existing Automated Reasoning policy. This allows you to iterate on your policy rules while maintaining previous versions for rollback or comparison purposes.

Deploys a custom model for on-demand inference in Amazon Bedrock. After you deploy your custom model, you use the deployment's Amazon Resource Name (ARN) as the modelId parameter when you submit prompts and generate responses with model inference. For more information about setting up on-demand inference for custom models, see Set up inference for a custom model. The following actions are related to the CreateCustomModelDeployment operation: GetCustomModelDeployment ListCustomModelDeployments DeleteCustomModelDeployment

Deploys a custom model for on-demand inference in Amazon Bedrock. After you deploy your custom model, you use the deployment's Amazon Resource Name (ARN) as the modelId parameter when you submit prompts and generate responses with model inference. For more information about setting up on-demand inference for custom models, see Set up inference for a custom model. The following actions are related to the CreateCustomModelDeployment operation: GetCustomModelDeployment ListCustomModelDeployments DeleteCustomModelDeployment

The Amazon S3 data source of the model to import.

The data source of the model to import.

Creates a new custom model in Amazon Bedrock. After the model is active, you can use it for inference. To use the model for inference, you must purchase Provisioned Throughput for it. You can't use On-demand inference with these custom models. For more information about Provisioned Throughput, see Provisioned Throughput. The model appears in ListCustomModels with a customizationType of imported. To track the status of the new model, you use the GetCustomModel API operation. The model can be in the following states: Creating - Initial state during validation and registration Active - Model is ready for use in inference Failed - Creation process encountered an error Related APIs GetCustomModel ListCustomModels DeleteCustomModel

Creates a new custom model in Amazon Bedrock. After the model is active, you can use it for inference. To use the model for inference, you must purchase Provisioned Throughput for it. You can't use On-demand inference with these custom models. For more information about Provisioned Throughput, see Provisioned Throughput. The model appears in ListCustomModels with a customizationType of imported. To track the status of the new model, you use the GetCustomModel API operation. The model can be in the following states: Creating - Initial state during validation and registration Active - Model is ready for use in inference Failed - Creation process encountered an error Related APIs GetCustomModel ListCustomModels DeleteCustomModel

The Amazon S3 location where the results of your evaluation job are saved.

Configuration for the Amazon Bedrock foundation model used for reranking vector search results. This specifies which model to use and any additional parameters required by the model.

Specifies a field to be used during the reranking process in a Knowledge Base vector search. This structure identifies metadata fields that should be considered when reordering search results to improve relevance.

Configuration for selectively including or excluding metadata fields during the reranking process. This allows you to control which metadata attributes are considered when reordering search results.

Configuration for how metadata should be used during the reranking process in Knowledge Base vector searches. This determines which metadata fields are included or excluded when reordering search results.

Configuration for using Amazon Bedrock foundation models to rerank Knowledge Base vector search results. This enables more sophisticated relevance ranking using large language models.

Configuration for reranking vector search results to improve relevance. Reranking applies additional relevance models to reorder the initial vector search results based on more sophisticated criteria.

Specifies the name of the metadata attribute/field to apply filters. You must match the name of the attribute/field in your data source/document metadata.

Specifies the filters to use on the metadata attributes/fields in the knowledge base data sources before returning results.

Defines the schema for a metadata attribute used in Knowledge Base vector searches. Metadata attributes provide additional context for documents and can be used for filtering and reranking search results.

Configuration for implicit filtering in Knowledge Base vector searches. Implicit filtering allows you to automatically filter search results based on metadata attributes without requiring explicit filter expressions in each query.

The configuration details for returning the results from the knowledge base vector search.

Contains configuration details for retrieving information from a knowledge base.

The configuration details for retrieving information from a knowledge base.

The configuration details for transforming the prompt.

The configuration details for the model to process the prompt prior to retrieval and response generation.

The template for the prompt that's sent to the model for response generation.

The configuration details for text generation using a language model via the RetrieveAndGenerate function.

Contains configuration details of the inference for knowledge base retrieval and response generation.

The configuration details for the guardrail.

The configuration details for response generation based on retrieved text chunks.

Contains configuration details for retrieving information from a knowledge base and generating responses.

The response generation configuration of the external source wrapper object.

The unique wrapper object of the document from the S3 location.

The unique external source of the content contained in the wrapper object.

The configuration of the external source wrapper object in the retrieveAndGenerate function.

Contains configuration details for a knowledge base retrieval and response generation.

The configuration details for retrieving information from a knowledge base and generating responses.

A summary of a RAG source used for a retrieve-only Knowledge Base evaluation job where you provide your own inference response data.

A summary of a RAG source used for a retrieve-and-generate Knowledge Base evaluation job where you provide your own inference response data.

A summary of a RAG source used for a Knowledge Base evaluation job where you provide your own inference response data.

Contains configuration details for retrieval of information and response generation.

A summary of a model used for a model evaluation job where you provide your own inference response data.

Contains performance settings for a model.

Contains the ARN of the Amazon Bedrock model or inference profile specified in your evaluation job. Each Amazon Bedrock model supports different inferenceParams. To learn more about supported inference parameters for Amazon Bedrock models, see Inference parameters for foundation models. The inferenceParams are specified using JSON. To successfully insert JSON as string make sure that all quotations are properly escaped. For example, "temperature":"0.25" key value pair would need to be formatted as \"temperature\":\"0.25\" to successfully accepted in the request.

Defines the models used in the model evaluation job.

The configuration details of the inference model for an evaluation job. For automated model evaluation jobs, only a single model is supported. For human-based model evaluation jobs, your annotator can compare the responses for up to two different models.

Contains SageMakerFlowDefinition object. The object is used to specify the prompt dataset, task type, rating method and metric names.

In a model evaluation job that uses human workers you must define the name of the metric, and how you want that metric rated ratingMethod, and an optional description of the metric.

Specifies the custom metrics, how tasks will be rated, the flow definition ARN, and your custom prompt datasets. Model evaluation jobs use human workers only support the use of custom prompt datasets. To learn more about custom prompt datasets and the required format, see Custom prompt datasets. When you create custom metrics in HumanEvaluationCustomMetric you must specify the metric's name. The list of names specified in the HumanEvaluationCustomMetric array, must match the metricNames array of strings specified in EvaluationDatasetMetricConfig. For example, if in the HumanEvaluationCustomMetric array your specified the names "accuracy", "toxicity", "readability" as custom metrics then the metricNames array would need to look like the following ["accuracy", "toxicity", "readability"] in EvaluationDatasetMetricConfig.

The configuration details of either an automated or human-based evaluation job.

Creates an evaluation job.

Creates an evaluation job.

Request a model access agreement for the specified model.

Request a model access agreement for the specified model.

A word to configure for the guardrail.

The managed word list to configure for the guardrail.

Contains details about the word policy to configured for the guardrail.

The tier that your guardrail uses for denied topic filters. Consider using a tier that balances performance, accuracy, and compatibility with your existing generative AI workflows.

Details about topics for the guardrail to identify and deny.

Contains details about topics that the guardrail should identify and deny.

The regular expression to configure for the guardrail.

The PII entity to configure for the guardrail.

Contains details about PII entities and regular expressions to configure for the guardrail.

The system-defined guardrail profile that you're using with your guardrail. Guardrail profiles define the destination Amazon Web Services Regions where guardrail inference requests can be automatically routed. Using guardrail profiles helps maintain guardrail performance and reliability when demand increases. For more information, see the Amazon Bedrock User Guide.

The filter configuration details for the guardrails contextual grounding filter.

The policy configuration details for the guardrails contextual grounding policy.

The tier that your guardrail uses for content filters. Consider using a tier that balances performance, accuracy, and compatibility with your existing generative AI workflows.

Contains filter strengths for harmful content. Guardrails support the following content filters to detect and filter harmful user inputs and FM-generated outputs. Hate – Describes language or a statement that discriminates, criticizes, insults, denounces, or dehumanizes a person or group on the basis of an identity (such as race, ethnicity, gender, religion, sexual orientation, ability, and national origin). Insults – Describes language or a statement that includes demeaning, humiliating, mocking, insulting, or belittling language. This type of language is also labeled as bullying. Sexual – Describes language or a statement that indicates sexual interest, activity, or arousal using direct or indirect references to body parts, physical traits, or sex. Violence – Describes language or a statement that includes glorification of or threats to inflict physical pain, hurt, or injury toward a person, group or thing. Content filtering depends on the confidence classification of user inputs and FM responses across each of the four harmful categories. All input and output statements are classified into one of four confidence levels (NONE, LOW, MEDIUM, HIGH) for each harmful category. For example, if a statement is classified as Hate with HIGH confidence, the likelihood of the statement representing hateful content is high. A single statement can be classified across multiple categories with varying confidence levels. For example, a single statement can be classified as Hate with HIGH confidence, Insults with LOW confidence, Sexual with NONE confidence, and Violence with MEDIUM confidence. For more information, see Guardrails content filters.

Contains details about how to handle harmful content.

Configuration settings for integrating Automated Reasoning policies with Amazon Bedrock Guardrails.

Creates a guardrail to block topics and to implement safeguards for your generative AI applications. You can configure the following policies in a guardrail to avoid undesirable and harmful content, filter out denied topics and words, 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. In addition to the above policies, you can also configure the messages to be returned to the user if a user input or model response is in violation of the policies defined in the guardrail. For more information, see Amazon Bedrock Guardrails in the Amazon Bedrock User Guide.

Creates a guardrail to block topics and to implement safeguards for your generative AI applications. You can configure the following policies in a guardrail to avoid undesirable and harmful content, filter out denied topics and words, 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. In addition to the above policies, you can also configure the messages to be returned to the user if a user input or model response is in violation of the policies defined in the guardrail. For more information, see Amazon Bedrock Guardrails in the Amazon Bedrock User Guide.

Creates a version of the guardrail. Use this API to create a snapshot of the guardrail when you are satisfied with a configuration, or to compare the configuration with another version.

Creates a version of the guardrail. Use this API to create a snapshot of the guardrail when you are satisfied with a configuration, or to compare the configuration with another version.

Contains information about the model or system-defined inference profile that is the source for an inference profile..

Creates an application inference profile to track metrics and costs when invoking a model. To create an application inference profile for a foundation model in one region, specify the ARN of the model in that region. To create an application inference profile for a foundation model across multiple regions, specify the ARN of the system-defined inference profile that contains the regions that you want to route requests to. For more information, see Increase throughput and resilience with cross-region inference in Amazon Bedrock. in the Amazon Bedrock User Guide.

Creates an application inference profile to track metrics and costs when invoking a model. To create an application inference profile for a foundation model in one region, specify the ARN of the model in that region. To create an application inference profile for a foundation model across multiple regions, specify the ARN of the system-defined inference profile that contains the regions that you want to route requests to. For more information, see Increase throughput and resilience with cross-region inference in Amazon Bedrock. in the Amazon Bedrock User Guide.

The configuration of a virtual private cloud (VPC). For more information, see Protect your data using Amazon Virtual Private Cloud and Amazon Web Services PrivateLink.

Specifies the configuration for a Amazon SageMaker endpoint.

Specifies the configuration for the endpoint.

Creates an endpoint for a model from Amazon Bedrock Marketplace. The endpoint is hosted by Amazon SageMaker.

Contains details about an endpoint for a model from Amazon Bedrock Marketplace.

Creates an endpoint for a model from Amazon Bedrock Marketplace. The endpoint is hosted by Amazon SageMaker.

Copies a model to another region so that it can be used there. For more information, see Copy models to be used in other regions in the Amazon Bedrock User Guide.

Copies a model to another region so that it can be used there. For more information, see Copy models to be used in other regions in the Amazon Bedrock User Guide.

Information about a validator.

Array of up to 10 validators.

A mapping of a metadata key to a value that it should or should not equal.

Rules for filtering invocation logs. A filter can be a mapping of a metadata key to a value that it should or should not equal (a base filter), or a list of base filters that are all applied with AND or OR logical operators

A storage location for invocation logs.

Settings for using invocation logs to customize a model.

S3 Location of the training data.

S3 Location of the output data.

Hyperparameters for controlling the reinforcement fine-tuning training process, including learning settings and evaluation intervals.

Configuration for using an AWS Lambda function to grade model responses during reinforcement fine-tuning training.

Configuration for the grader used in reinforcement fine-tuning to evaluate model responses and provide reward signals.