preloadMemoryContext

function preloadMemoryContext(
   memoryService,
   session,
   userInput,
maxTokens): Promise<string | null>;

Preloads relevant memories into the system prompt before each LLM call.

This is NOT a tool the LLM calls. It is a Runtime-level middleware that:

1. Takes the user’s latest message as a search query
2. Searches long-term memory for relevant context
3. Formats matching memories as a markdown section
4. Truncates the output to fit within the allocated token budget

The maxTokens parameter is mandatory. If the formatted output exceeds maxTokens, memories are dropped in lowest-relevance-first order until the output fits.

Parameters

Parameter	Type
memoryService	MemoryService
session	Session
userInput	string
maxTokens	number

Returns

Promise<string | null>