Tool that automatically preloads relevant memories into the LLM request.

Unlike RecallMemoryTool which requires explicit agent invocation, this tool injects memories during request preprocessing based on the user's query.

This tool does not expose a function declaration to the LLM - it only processes the LLM request to inject context.

Hierarchy (view full)

Constructors

Properties

name: string

Name of the tool

description: string

Description of the tool

isLongRunning: boolean

Whether the tool is a long running operation, which typically returns a resource id first and finishes the operation later.

shouldRetryOnFailure: boolean

Whether the tool execution should be retried on failure

maxRetryAttempts: number

Maximum retry attempts

baseRetryDelay: number = 1000

Base delay for retry in ms (will be used with exponential backoff)

maxRetryDelay: number = 10000

Maximum delay for retry in ms

logger: Logger = ...

Accessors

Methods