- •The Apply Guardrail action applies established policies and restrictions to ensure that the AI models and applications comply with the security, ethics, and reliability requirements.
The following snippet is a sample to build a requestBody object:
<root xmlns:m="urn:informatica:ae:xquery:json2xml:meta-data">
<source>INPUT</source>
<content m:isArray="true">
<text>
<text> How can I treat a runny nose?</text>
</text>
</content>
</root>
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Apply Guardrail action, see the
Amazon Bedrock documentation.
- •The Associate Agent Knowledge Base associates a knowledge base with a Secure Agent. If a knowledge base is associated and the indexState is set to Enabled, the Secure Agent queries the knowledge base for information to augment its response to the user.
The following snippet is a sample to build a requestBody object:
<root>
<description> Agent description </description>
<knowledgeBaseId> KnowledgeBaseId </knowledgeBaseId>
<knowledgeBaseState>ENABLED</knowledgeBaseState>
</root>
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Associate Agent Knowledge Base action, see the
Amazon Bedrock documentation.
- •The Converse action sends messages to the specified Amazon Bedrock model using a consistent interface that is compatible with all message-supporting models. This enables you to write code once and reuse it with different models.
The following snippet is a sample to build a requestBody object:
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Converse action, see the
Amazon Bedrock documentation.
- •The Create Agent creates an agent that orchestrates interactions between foundation models, data sources, software applications, user conversations, and APIs for customer assistance.
To create and use Agent, use the actions in the following order:
- 1Create Agent
- 2Associate Agent Knowledge Base
- 3Create Agent Action Group
- 4Update Agent
- 5Prepare Agent
- 6Create Agent Alias
The following snippet is a sample to build a requestBody object:
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Create Agent action, see the
Amazon Bedrock documentation.
- •The Create Agent Action Group creates an action group for a Secure Agent. An action group represents the actions that a Secure Agent can carry out for the customer by defining the APIs that a Secure Agent can call and the logic for calling them.
The following snippet is a sample to build a requestBody object:
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Create Agent Action Group action, see the
Amazon Bedrock documentation.
- •The Create Agent Alias creates an alias for a Secure Agent that can be used to deploy the agent.
The following snippet is a sample to build a requestBody object:
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Create Agent Alias action, see the
Amazon Bedrock documentation.
- •The Create Data Source action creates a data source connector for a knowledge base.
The following snippet is a sample to build a requestBody object:
<root xmlns:m="urn:informatica:ae:xquery:json2xml:meta-data">
<dataDeletionPolicy>DELETE</dataDeletionPolicy>
<dataSourceConfiguration>
<s3Configuration>
<bucketArn>arn:aws:s3:::BUCKET</bucketArn>
<inclusionPrefixes m:isArray="true"> Lorem_ipsum.pdf </inclusionPrefixes>
</s3Configuration>
<type>S3</type>
</dataSourceConfiguration>
<name>Source</name>
</root>
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Create Data Source action, see the
Amazon Bedrock documentation.
- •The Create Knowledge Base action creates a knowledge base of data sources that the LLMs can use to query for information.
To create and use the Knowledge Base, use the actions in the following order::
- 1Create Knowledge Base
- 2Create Data Source
- 3Start Ingestion Job
- 4Get List of Ingestion Jobs
- 5Retrieve and Generate Knowledge Base
The following snippet is a sample to build a requestBody object:
<root>
<knowledgeBaseConfiguration>
<type>VECTOR</type>
<vectorKnowledgeBaseConfiguration>
<embeddingModelArn>arn:aws:bedrock:us-east-1::foundation-model/amazon.titan-embed-text-v1</embeddingModelArn>
</vectorKnowledgeBaseConfiguration>
</knowledgeBaseConfiguration>
<name> KnowledgeBaseName </name>
<roleArn>arn:aws:iam::800000000000:role/service-role/AmazonBedrockExecutionRoleForKnowledgeBase_ROLE</roleArn>
<storageConfiguration>
<opensearchServerlessConfiguration>
<collectionArn>arn:aws:aoss:us-east-1: 800000000000:collection/COLLECTION</collectionArn>
<fieldMapping>
<metadataField>metadata</metadataField>
<textField>text</textField>
<vectorField>name</vectorField>
</fieldMapping>
<vectorIndexName>index</vectorIndexName>
</opensearchServerlessConfiguration>
<type>OPENSEARCH_SERVERLESS</type>
</storageConfiguration>
</root>
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Create Knowledge Base action, see the
Amazon Bedrock documentation.
- •The Delete Agent Memory action deletes memory from the specified memory identifier. You can delete all stored sessions when memory is enabled for your Secure Agent.
After a session ends, the Secure Agent summarizes the conversation and stores the conclusion in its memory. This process might take some time.
For more information about configuring the parameters for the
Delete Agent Memory action, see the
Amazon Bedrock documentation.
- •The Get Agent Memory action returns the sessions stored in the Secure Agent's memory. You can view these stored sessions when memory is enabled for your agent.
After a session ends, the Secure Agent summarizes the conversation and stores the conclusion in its memory. This process might take some time.
For more information about configuring the parameters for the
Get Agent Memory action, see the
Amazon Bedrock documentation.
- •Use the Get Foundation Models List action to get a list of the available models.
For more information about using the request parameter URI, see the
Amazon Bedrock documentation.
- •The Get List Ingestion Jobs gets a list of ingestion jobs for a data source and provides information about each job.
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Get List Ingestion Jobs action, see the
Amazon Bedrock documentation.
- •The Invoke Agent action in Amazon Bedrock Agents allow you to build and configure autonomous agents within your application. These agents assist your end users in completing actions based on organizational data and user input. Assign a new value to sessionId each time you start a new dialog with the agent. Set a unique identifier in the memoryId field within the request body when memory is enabled for the Secure Agent.
Secure Agent responses are provided in the base64 format. To decode the response, set decodeOutput to true. There is no option to enable trace for tracking the agent's reasoning process.
The following snippet is a sample to build a requestBody object:
<root>
<inputText>Send email to employee Alex </inputText>
<memoryId>Memory_01</memoryId>
</root>
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Invoke Agent action, see the
Amazon Bedrock documentation.
- •The Invoke Flow action runs a flow alias to process the inputs that you specify and returns the output from each node.
The finalResponse contains the value document in the FlowOutputContent object.
The following snippet is a sample to build a requestBody object:
<root xmlns:m="urn:informatica:ae:xquery:json2xml:meta-data">
<inputs m:isArray="true">
<nodeName>FlowInputNode</nodeName>
<nodeOutputName>document</nodeOutputName>
<content>
<document>Some text</document>
</content>
</inputs>
</root>
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Invoke Flow action, see the
Amazon Bedrock documentation.
- •The Invoke Model action invokes the specified Amazon Bedrock model to run inference using the prompt and parameters provided in the request body. It uses model inference to generate text, images, and embeddings. This action does not support Guardrail to call models. The prompt and inference parameters in the requestBody vary for each model.
The following snippet is a sample requestBody object:
For more information about the inference parameters for foundation models, see the
Amazon Bedrock documentation.
If fields in the request body need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Invoke Model action, see the
Amazon Bedrock documentation.
- •The Invoke Model with Guardrail action invokes the specified Amazon Bedrock model to run inference using the prompt and parameters provided in the request body. It uses model inference to generate text, images, and embeddings, and applies a guardrail. The prompt and inference parameters in the requestBody vary for each model.
The following snippet is a sample requestBody object:
For more information about the inference parameters for foundation models, see the
Amazon Bedrock documentation.
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Invoke Model with Guardrail action, see the
Amazon Bedrock documentation.
- •The Prepare Agent creates a draft version of the Secure Agent for internal testing.
For more information about configuring the parameters for the
Prepare Agent action, see the
Amazon Bedrock documentation.
- •The Retrieve and Generate Knowledge Base action generates responses based on the queried knowledge base results. The response includes only sources relevant to the query.
The following snippet is a sample to build a requestBody object:
<root>
<input>
<text>Give me the email of the employee Alex</text>
</input>
<retrieveAndGenerateConfiguration>
<knowledgeBaseConfiguration>
<knowledgeBaseId> KnowledgeBaseId </knowledgeBaseId>
<modelArn>anthropic.claude-v2:1</modelArn>
</knowledgeBaseConfiguration>
<type>KNOWLEDGE_BASE</type>
</retrieveAndGenerateConfiguration>
</root>
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Retrieve and Generate Knowledge Base action, see the
Amazon Bedrock documentation.
- •The Retrieve Knowledge Base action queries a knowledge base to fetch information.
The following snippet is a sample to build a requestBody object:
<root>
<retrievalQuery>
<text> Give me the email of the employee Alex</text>
</retrievalQuery>
</root>
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Retrieve Knowledge Base action, see the
Amazon Bedrock documentation.
- •The Start Ingestion Job starts an ingestion job to add a data source to a knowledge base.
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Start Ingestion Job action, see the
Amazon Bedrock documentation.
- •The Update Agent updates the configuration of a Secure Agent.
The following snippet is a sample to build a requestBody object:
If fields in the requestBody need to be in different formats, you must add additional namespaces as shown in the following sample code:
For more information about configuring the parameters for the
Update Agent action, see the
Amazon Bedrock documentation.