Context management and agent memory
AI-500 · Lesson 6 of 16 · 47:28
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the AI-500 exam (Microsoft Certified: Multi-Agent AI Solutions Expert). In this lesson: how context accumulates across agent chains, long histories, and tool results, retrieval with hybrid search, injection within token budgets, compaction with rolling windows, important turns, and validated digests between agents, memory storage with embeddings and a partition key per traveler, Foundry's managed memory (preview), session management, and memory security, compliance, and lifecycle: auditing, deletion on request, decay, pruning, and consolidation. Includes three original practice questions.
Preview the topics
- By the time a request reaches Contoso’s payment agent, its context holds the full chat and every earlier agent’s notes, and costs keep rising. What should the orchestrator pass instead?
- Contoso wants its assistant to remember travelers across sessions, but storing full transcripts is costly and risky. What should the memory strategy store?
- A traveler asks Contoso to erase everything its assistant remembers about them. What must the memory design support?