MemoryStack icon

MemoryStack

AI Memory Infrastructure for Agents & LLMs

Search Memories

AI-generated

Summary

Search memories using a natural language query with configurable search modes and filters.

Inputs

  • Query (required) — Natural language search query string used to find relevant memories.
  • Search Mode — Method to perform the search: 'Hybrid (Best)' combines vector and text search, 'Vector (Semantic)' uses semantic vector search, and 'Text (Keyword)' uses keyword text search. Defaults to 'Hybrid (Best)'.
  • Limit — Maximum number of memories to return, between 1 and 50. Defaults to 10.
  • Min Confidence — Optional minimum confidence score filter (0-1) to only return memories above this threshold. Defaults to 0.
  • User ID — Optional user identifier for B2B multi-tenant scenarios to scope the search to a specific user.

Output shape

a list of memory records matching the search query

The operation returns an array of memory objects in JSON format filtered and ranked according to the query and parameters. Pagination is controlled via the 'Limit' parameter. If 'Min Confidence' is set above 0, only memories with at least that confidence are included.

Discussion