createMemoryEngineTool

createMemoryEngineTool(storageCtx: StorageOperationsContext, embeddingOptions: MemoryEngineEmbeddingOptions, searchOptions?: Partial<MemoryEngineSearchOptions>, callbacks?: object): ToolConfig

Defined in: src/lib/memoryEngine/tool.ts:93

Creates a memory engine tool for use with chat completions.

The tool allows the LLM to search through past conversation messages using semantic similarity. Messages must have embeddings stored to be searchable.

Parameters

Parameter	Type	Description
`storageCtx`	`StorageOperationsContext`	Storage operations context for database access
`embeddingOptions`	`MemoryEngineEmbeddingOptions`	Options for embedding generation
`searchOptions?`	`Partial`<`MemoryEngineSearchOptions`>	Default search options (can be overridden per-call)
`callbacks?`	`object`	‐
`callbacks.onRetrieve?`	(`conversationIds`: `string`[]) => `void`	Called after retrieval with the conversation IDs that were actually returned to the LLM.

Returns

ToolConfig

A ToolConfig that can be passed to chat completion tools

Example


const tool = createMemoryEngineTool(
  storageCtx,
  { getToken: () => getIdentityToken() },
  { limit: 5, minSimilarity: 0.4 }
);
 
// Use with chat completion
const result = await sendMessage({
  messages: [...],
  tools: [tool],
});