🎓 FrootAI Workshops

Hands-on, facilitator-led workshops for teams adopting FrootAI. Each workshop includes prerequisites, a timed outline, and everything a facilitator needs to deliver the session.

Workshop Catalog

Note: Full slide decks and hands-on lab repos are coming soon. This README serves as the catalog and planning guide.

Workshop 1: Build Your First RAG Pipeline with FrootAI

Duration: 2 hours
Level: Intermediate
Detailed outline: 01-build-rag-pipeline.md

Prerequisites

• Azure subscription with Contributor access
• Azure OpenAI resource (GPT-4o deployed)
• Azure AI Search resource (Basic tier or higher)
• VS Code with FrootAI extension installed
• Node.js 18+ and Python 3.10+
• Basic familiarity with REST APIs and vector embeddings

Outline

1. Concepts (20 min) — What is RAG? Why not just fine-tune? Architecture overview.
2. Data Prep (20 min) — Load documents, chunk with FrootAI’s chunking strategy, generate embeddings.
3. Index Build (20 min) — Create Azure AI Search index with vector fields, push data.
4. Query Pipeline (20 min) — Build hybrid search (keyword + vector), configure semantic ranker.
5. Agent Integration (20 min) — Wire the pipeline into an agent via MCP, test via Copilot Chat.
6. Evaluation (20 min) — Run groundedness + relevance evals using FrootAI evaluation framework.

Materials Needed

• Play 01 (Enterprise RAG) from the Solution Plays library
• Sample PDF corpus (provided in lab repo)
• config/openai.json and config/search.json templates

Workshop 2: Set Up .github Agentic OS for Your Team

Duration: 1.5 hours
Level: Beginner

Prerequisites

• GitHub repository (public or private)
• VS Code with GitHub Copilot enabled
• Basic understanding of Markdown
• Admin access to the target repo

Outline

1. Why Agentic OS? (15 min) — The problem with undirected AI; how .github/ files steer Copilot.
2. copilot-instructions.md (15 min) — Write project-level instructions. Demo: before vs. after.
3. Instruction files (15 min) — Layer-specific rules (security, Azure coding, testing conventions).
4. agent.md (15 min) — Define agent persona, tools, guardrails, and escalation rules.
5. Custom prompts (15 min) — Create .github/prompts/ for repeatable slash commands.
6. Validation (15 min) — Test the setup end-to-end with Copilot Chat; iterate on instructions.

Materials Needed

• FrootAI .github/ template pack
• Sample copilot-instructions.md, agent.md, and instruction files
• Checklist of common instruction anti-patterns

Workshop 3: MCP Server: From Install to Production

Duration: 1 hour
Level: Beginner

Prerequisites

• Node.js 18+
• VS Code with Copilot or Claude Desktop
• npm access (for npx @abacloud/frootai-mcp)
• (Optional) Azure subscription for hosted deployment

Outline

1. What is MCP? (10 min) — Model Context Protocol explained. Tools vs. resources vs. prompts.
2. Quick Start (10 min) — npx @abacloud/frootai-mcp — verify tools show up in Copilot Chat.
3. Tool Deep Dive (15 min) — Walk through lookup_term, search_knowledge, get_architecture_pattern.
4. Configuration (10 min) — Custom config, hosted vs. stdio mode, environment variables.
5. Production Deployment (10 min) — Deploy to Azure Container Apps, configure Managed Identity.
6. Q&A (5 min) — Open floor.

Materials Needed

• MCP Server npm package
• mcp.json configuration template
• Sample queries for each of the 16 tools

Workshop 4: AI Cost Optimization with FrootAI

Duration: 1.5 hours
Level: Intermediate

Prerequisites

• Azure subscription with Cost Management Reader role
• Azure OpenAI resource with usage history (1+ week)
• VS Code with FrootAI extension
• Basic understanding of Azure pricing (pay-as-you-go vs. reserved)

Outline

1. The Cost Problem (15 min) — Why AI workloads surprise teams. Token math and GPU pricing.
2. Cost Visibility (20 min) — Set up cost dashboards, tag strategy, budget alerts.
3. Token Optimization (20 min) — Prompt compression, caching strategies, model routing (GPT-4o vs. GPT-4o-mini).
4. Infrastructure Right-Sizing (15 min) — PTU vs. pay-as-you-go, auto-shutdown for dev, reserved capacity.
5. FrootAI Cost Agent (15 min) — Demo the AI Cost Optimization play analyzing a real subscription.
6. Action Plan (5 min) — Each attendee writes their top-3 cost actions with expected savings.

Materials Needed

• Play 06 (AI Cost Optimization Advisor) from the Solution Plays library
• Cost Optimization module from the cost-optimization/ knowledge base
• Sample Azure Cost Management export (CSV, provided)

Facilitator Notes

• Format: Each workshop works both in-person and virtual (screen-share + breakout rooms).
• Pacing: Times are approximate. Allocate buffer for Q&A between sections.
• Recording: Request permission before recording. Share recordings internally only.
• Feedback: Use a post-workshop survey (template in workshops/templates/ — coming soon).

Contributing

Want to propose a new workshop topic? Open an issue with the workshop label or submit a PR following the outline template in 01-build-rag-pipeline.md.