Play 21
Agentic RAG
Autonomous retrieval — the agent decides when, what, and where to search.
Autonomous RAG where the AI agent controls retrieval end-to-end. Unlike standard RAG (fixed pipeline: query → search → generate), the agent decides when to search, what to search for, which sources to query (AI Search, Bing, SQL, custom APIs), iterates on results if insufficient, and synthesizes across multiple knowledge bases with proper citations. Self-evaluates response quality (groundedness ≥0.95) before returning. Semantic caching reduces repeat query costs by >60%.
Architecture Pattern
Agentic RAG: agent-controlled retrieval, multi-source, semantic caching
Azure Services
DevKit (.github Agentic OS)
- agent.md — root orchestrator with builder→reviewer→tuner handoffs
- 3 agents — Agentic RAG Builder (gpt-4o), Reviewer (gpt-4o-mini), Tuner (gpt-4o-mini)
- 3 skills — deploy (110 lines), evaluate (104 lines), tune (104 lines)
- 4 prompts — /deploy, /test, /review, /evaluate with agent routing
- .vscode/mcp.json — FrootAI MCP with OpenAI + AI Search inputs + envFile
TuneKit (AI Config)
- config/openai.json — temp=0.1, seed=42, structured output
- config/search.json — multi-source ranking, semantic cache TTL
- config/guardrails.json — groundedness ≥0.95, abstention on low confidence
- evaluation/eval.py — Groundedness >0.95, Coherence >0.90, Relevance >0.90
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.