An enterprise multi-agent assistant needs (1) turn-by-turn dialogue context, (2) shared scratch state for the agent team during a case, and (3) long-term semantic memory of resolved cases—with tenant isolation. Which memory architecture should you specify?
Select an answer to reveal the explanation.
Short Explanation
You need multi-tier memory: session context, shared case state, and long-term semantic memory with tenant isolation. That’s the sticky notes, the team whiteboard, and the filing cabinet—each with a different lifespan and lock. It lets the agent team remember without leaking across customers.
Full Explanation
Architecture skills include short- and long-term memory designs and multi-tier state persistence for session state, shared team state, and long-term semantic memory with lifecycle and tenant isolation. The multi-tier state architecture matches that three-tier model with isolation and retention policies.
The single global vector store with no session boundaries is incorrect: it creates a data-leakage and compliance failure because every agent can see all tenants’ data.
The stateless agents with no memory components approach is incorrect: memory is a core multi-agent capability when governed properly; responsible AI does not ban memory, it governs it.
The store all history only in the model’s weights via continuous fine-tuning after every user message approach is incorrect: per-message fine-tuning is impractical, costly, and not a substitute for explicit memory systems.
Exam tip: Map session, shared team, and long-term semantic memory plus tenant isolation language directly to multi-tier memory architecture answers.