Implement advanced multi-agent orchestration patterns in Microsoft Foundry

Why it matters

Parallel specialists only help when code, not the model, decides what counts as enough evidence and what happens when one of them fails.

What to practise

Complete a Python hub that creates versioned prompt agents in Microsoft Foundry, chooses sequential or parallel execution from declared dependencies, and fans out live Responses calls with asyncio.gather behind a bounded semaphore. You write the quorum check that separates a missing optional spoke from a missing required one, and the fan-in that blocks the supervisor when quorum fails. Run a normal input and an optional-failure input, and compare quorum.missing_agents, accepted_evidence and the supervisor’s caveat. In the portal traces, check that spoke calls overlap and the supervisor starts after them.

How it fits

The quorum and partial-failure outcomes are 2.4.2’s aggregation policy. Spokes that never see each other’s output are 2.4.3’s controlled hub. The semaphore is a small case of 2.4.9’s admission limits, and the optional sequential route is 2.4.1’s dependency gate.