LLM 同 AI
Memory Query
Use when the user asks to read from, write to, or query agent memory, or when
an agent needs to retrieve prior context, store results for downstream agents,
or search for related knowledge. Handles requests like "retrieve hypotheses
from memory", "store KPI targets for the next agent", "search memory for
stakeholder data", "add a semantic memory entry", or "query the vector store".
Covers which memory layer to use, correct tenant-scoped query patterns, and
the distinction between in-process MemorySystem and the persistent
packages/memory subsystem.