Solution Plays Overview
Solution Plays are pre-built, tested, production-ready AI solution blueprints. Each play provides everything you need to go from zero to deployed: Copilot agents, configuration files, infrastructure-as-code, evaluation pipelines, and documentation — all wired together by the FAI Protocol.
:::info Not Boilerplate
Plays aren't starter templates. They're opinionated, WAF-aligned implementations that encode the lessons from hundreds of enterprise AI deployments. You customize the config/ files; the architecture and safety patterns are built in.
:::
What's in a Play?
Every solution play follows the 4-kit structure:
solution-play-NN/
├── .github/ ← DevKit: Copilot brain
│ ├── copilot-instructions.md ← under 150 lines, domain knowledge
│ ├── agents/ ← builder / reviewer / tuner
│ │ ├── builder.agent.md
│ │ ├── reviewer.agent.md
│ │ └── tuner.agent.md
│ ├── instructions/ ← WAF-aligned guidance
│ ├── skills/ ← 150+ line action skills
│ └── hooks/ ← SessionStart guardrails
├── config/ ← TuneKit: customer-tunable params
│ ├── openai.json ← model, temperature, max_tokens
│ └── guardrails.json ← safety thresholds
├── infra/ ← Infra: Azure Bicep (AVM modules)
│ ├── main.bicep
│ └── modules/
├── evaluation/ ← Eval: quality pipeline
│ └── eval-config.json
└── spec/ ← SpecKit: metadata + docs
└── fai-manifest.json ← FAI Protocol wiring
| Kit | Purpose | Who Edits |
|---|---|---|
| DevKit | Copilot agents, instructions, skills, hooks | Platform team |
| TuneKit | AI parameters, guardrails, thresholds | AI team / customer |
| SpecKit | Play metadata, documentation, wiring | Auto-generated |
| Infra | Azure infrastructure (Bicep + AVM) | Platform / DevOps |
Categories
| Category | Plays | Description |
|---|---|---|
| RAG | 01, 09, 21 | Retrieval-Augmented Generation patterns |
| Agents | 03, 07, 22, 23 | Single-agent to swarm orchestration |
| Infrastructure | 02, 11, 14 | Landing zones, gateways, platforms |
| Security | 10 | Content moderation, safety pipelines |
| Voice | 04 | Speech-to-text, LLM, text-to-speech |
| Document | 06, 15 | OCR, extraction, multi-modal processing |
| Copilot | 08, 16 | Copilot Studio bots, Teams extensions |
| Observability | 17 | AI monitoring, metrics, dashboards |
| Fine-Tuning | 13 | LoRA/QLoRA workflows |
| Edge | 19 | On-device AI, ONNX, edge deployment |
| Real-Time | 20 | Streaming analytics, event-driven AI |
Featured Plays
| # | Name | Category | Complexity | Key Azure Services |
|---|---|---|---|---|
| 01 | Enterprise RAG | RAG | Intermediate | OpenAI, AI Search, Blob Storage |
| 02 | AI Landing Zone | Infrastructure | Intermediate | VNet, Private Endpoints, Key Vault |
| 03 | Deterministic Agent | Agents | Intermediate | OpenAI (temp=0), Structured Output |
| 04 | Call Center Voice AI | Voice | Advanced | Speech Services, OpenAI, Event Hub |
| 05 | IT Ticket Resolution | Agents | Intermediate | OpenAI, ServiceNow, Logic Apps |
| 06 | Document Intelligence | Document | Intermediate | Document Intelligence, OpenAI |
| 07 | Multi-Agent Service | Agents | Advanced | OpenAI, Semantic Kernel |
| 08 | Copilot Studio Bot | Copilot | Beginner | Copilot Studio, Dataverse |
| 09 | AI Search Portal | RAG | Intermediate | AI Search, OpenAI, App Service |
| 10 | Content Moderation | Security | Intermediate | Content Safety, OpenAI |
See the full Play Catalog for all 23+ plays.
How to Browse
| Method | Command / Link |
|---|---|
| Website | frootai.dev/solution-plays |
| VS Code | Install FrootAI extension → Solution Plays panel |
| CLI | npx frootai list plays |
| MCP Server | npx frootai-mcp@latest → list_community_plays tool |
How to Initialize
# Scaffold any play by number + name
npx frootai scaffold 01-enterprise-rag
# Scaffold with specific options
npx frootai scaffold 02-ai-landing-zone --region eastus2
# List available plays
npx frootai list plays
After scaffolding:
- Review
.github/copilot-instructions.mdfor domain context - Customize
config/openai.jsonandconfig/guardrails.json - Deploy infrastructure with
az deployment group create --template-file infra/main.bicep - Use
@builderagent in Copilot Chat to implement
The Agent Triad
Every play includes three specialized agents that follow the Build → Review → Tune workflow:
| Agent | Role | Model | Focus |
|---|---|---|---|
| Builder | Implements the solution | GPT-4o | Code generation, architecture |
| Reviewer | Reviews for quality and security | GPT-4o-mini | OWASP, WAF compliance, bugs |
| Tuner | Optimizes for production | GPT-4o-mini | Config validation, cost, performance |
@builder → implements code and architecture
↓
@reviewer → checks security, quality, WAF alignment
↓
@tuner → validates config, thresholds, production readiness
WAF Alignment
All plays are aligned to the 6 Well-Architected Framework pillars:
- Reliability — retry, circuit breaker, health checks, fallbacks
- Security — Managed Identity, Key Vault, RBAC, private endpoints
- Cost Optimization — model routing, token budgets, right-sizing
- Operational Excellence — CI/CD, IaC, observability, incident response
- Performance Efficiency — caching, streaming, async, model routing
- Responsible AI — content safety, groundedness, fairness, transparency
For detailed WAF patterns, see T3: Production Patterns and T2: Responsible AI.
Next Steps
- Browse the full Play Catalog
- Try Workshop: Build a RAG Pipeline (Play 01)
- Try Workshop: AI Landing Zone (Play 02)
- Learn the foundations in the Learning Hub