An architect must design memory for a multi-agent consulting workspace. Agents need the last few turns for immediate coherence, a team-shared board of intermediate findings during a multi-hour engagement, and durable client preferences that survive across sessions for the same tenant. Which memory architecture best fits?
Select an answer to reveal the explanation.
Short Explanation
You need three tiers: short-term session context, a shared team board for the engagement, and long-term tenant-scoped preferences with lifecycle rules. A model-context-only approach cannot hold multi-hour team state or cross-session preferences. A single global forever table without isolation is a compliance bomb, and zero memory destroys coherence.
Full Explanation
The multi-tier memory design is correct because it maps to the three distinct needs in the scenario: short-term session context for immediate coherence, shared team state for intermediate findings during a multi-hour engagement, and long-term semantic memory for durable client preferences with tenant isolation and lifecycle policies. Architecture skills require short-term and long-term memory designs including context sharing, plus multi-tier persistence for session state, shared team state, and long-term semantic memory with lifecycle and tenant isolation.
The model-context-only approach is incorrect because pure in-window context cannot reliably provide multi-hour shared boards or cross-session preferences.
The single global forever table approach is incorrect because lacking tenant isolation and using unbounded retention violates multi-tenant security and lifecycle governance.
The zero-memory approach is incorrect because eliminating memory removes required coherence and personalization rather than solving the design problem.