Play 22
Multi-Agent Swarm
Distributed agent teams — supervisor, pipeline, and debate patterns.
Deploys specialized AI agents that collaborate to solve complex problems no single agent can handle. Each agent has a distinct role, tools, and expertise. Implements three patterns: Supervisor (central orchestrator delegates to specialists), Pipeline (sequential handoff: research → analyze → write), and Debate (agents critique each other's outputs). Dapr provides state management and pub/sub messaging between agents. Loop prevention and max-iteration guards keep costs predictable.
Architecture Pattern
Multi-agent: supervisor, pipeline, debate patterns with shared state
Azure Services
DevKit (.github Agentic OS)
- agent.md — root orchestrator with builder→reviewer→tuner handoffs
- 3 agents — Swarm Builder (gpt-4o), Reviewer (gpt-4o-mini), Tuner (gpt-4o-mini)
- 3 skills — deploy (106 lines), evaluate (103 lines), tune (102 lines)
- 4 prompts — /deploy, /test, /review, /evaluate with agent routing
- .vscode/mcp.json — FrootAI MCP with OpenAI + Redis inputs + envFile
TuneKit (AI Config)
- config/openai.json — per-agent model params (supervisor=4o, workers=mini)
- config/agents.json — team topology, delegation rules, fallback chains
- config/guardrails.json — loop prevention, max 5 iterations, cost cap
- evaluation/eval.py — Handoff accuracy >90%, Loop rate <5%
Tuning Parameters
Machine evidence
FrootAI evidence lifecycle
This is an internal evidence maturity label, not third-party certification, accreditation, legal compliance, or a production guarantee. Missing or expired evidence demotes automatically; catalog claims cannot promote a play.
This play currently has design evidence only. A runnable scenario, endpoint evaluation, and build receipts are the next contiguous gates.
Repo Intelligence
v1A no-clone, revision-pinned map for agents and humans. Observed evidence is separated from inferred flow so the output stays useful without pretending to be a full call graph.