agents/guide-rail-agent

Interfaces#

GuideRailAgentInput#

Input interface for GuideRail agents

GuideRail agents receive both input and expected output, allowing them to validate, transform, or control the flow of messages between other agents.

Extends#

• `Message`

Indexable#

[key: string]: unknown

Properties#

Property	Type	Description
input?	unknown	The input data to be processed This is the original message that would be sent to the target agent
output?	unknown	The expected output data This is what the target agent is expected to produce, allowing the GuideRail agent to validate or transform the data flow

GuideRailAgentOutput#

Output interface for GuideRail agents

GuideRail agents can either allow the process to continue or abort it with a reason. This provides a mechanism for enforcing rules, validating data, and controlling the execution flow of the agent system.

Extends#

• `Message`

Indexable#

[key: string]: unknown

Properties#

Property	Type	Description
abort?	boolean	Whether to abort the current process When true, the agent system should stop the current execution flow and prevent further processing based on this input/output pair Default false
reason?	string	Reason for aborting the process When abort is true, this provides a human-readable explanation for why the process was stopped

Type Aliases#

GuideRailAgent#

GuideRailAgent = `Agent`<GuideRailAgentInput, GuideRailAgentOutput>

GuideRail agent type definition

GuideRail agents act as validators, transformers, or controllers for the message flow between agents. They can enforce rules, perform safety checks, ensure data quality, or implement business logic validations.

Use GuideRail agents when you need to:

• Validate inputs or outputs against specific criteria
• Enforce security or safety policies
• Implement business rules that control agent interactions
• Monitor and audit agent behavior

Variables#

guideRailAgentOptions#

const guideRailAgentOptions: `AgentOptions`<any, GuideRailAgentOutput>