In a multi-agent research workflow on Azure, the Researcher agent accumulates long web-tool traces that must be handed to a Writer agent without exceeding context limits or losing key citations. Which context management design should you implement?
Select an answer to reveal the explanation.
Short Explanation
Context management is accumulate, then compact — not hoard forever or erase everything. Extract findings and citations into a structured handoff, summarize noisy traces, and inject that package into the Writer. Option D. Dumping full traces forever (A) blows the window. Dropping citations (B) destroys grounded writing. Resetting after every tool call (C) causes amnesia mid-research.
Full Explanation
AI-500 covers context accumulation, retrieval, injection, and compaction for single- and multi-agent systems. Option D preserves decision-critical content (findings, citations) while controlling token growth via compaction and structured handoff.
Option A leads to context overflow and degraded attention on critical facts.
Option B harms groundedness and enables hallucination of sources.
Option C discards useful intermediate state required for multi-step research quality.
Implementation notes: maintain a working memory object (claims, URLs, confidence), compress tool stdout, and pass an explicit inter-agent message schema rather than free-form chat dumps.
Exam tip: Compaction keeps signal; it is not the same as deleting memory.