Build an intelligent FAQ assistant using Azure SQL vector search, RAG, Foundry Agents, Microsoft Fabric, and MCP.
By the end of this workshop, you will have:
- Semantic search — Query FAQs by meaning, not just keywords
- RAG pipeline — Ground AI responses in your data (no hallucinations)
- AI Agent — Orchestrate the workflow via Microsoft Foundry
- Analytics — Mirror SQL data into Fabric with Power BI reports
- MCP integration — Expose your database as an AI tool
| # | Module | Duration | What You'll Do |
|---|---|---|---|
| 0 | Setup & Prerequisites | 15 min | Verify tools, login, test connectivity |
| 1 | AI-Enhanced Querying | 20 min | Embeddings + retrieval scoring |
| 2 | Copilot SQL Dev | 15 min | Generate & refine SQL with AI |
| 3 | RAG Implementation | 25 min | Build grounded AI responses |
| 4 | Foundry Agents | 25 min | MCP tools + dev tunnels |
| 5 | Fabric Analytics | 20 min | Mirror, model, report |
| 6 | Data API Builder | 20 min | Database as MCP tool |
Total: ~2.5 hours
# 1. Clone this repo
git clone https://github.com/MSLab513/MS-LAB513-Student.git
cd MS-LAB513-Student
# 2. Login to Azure (credentials from instructor)
az login --tenant <tenant-id>
az account set --subscription <subscription-id>
# 3. Open the Student Guide and start Module 0Your instructor will provide:
- Tenant ID and Subscription ID
- SQL Server FQDN (e.g.,
faq-ai-assistant-xxxxx.database.windows.net) - OpenAI Endpoint (e.g.,
https://ai-xxxxx.openai.azure.com/) - Environment name (e.g.,
lab513-user01)
| Setting | Value |
|---|---|
| Region | swedencentral |
| SQL Auth | Microsoft Entra ID (no passwords needed) |
| AI Model | o4-mini |
| API Version | 2025-01-01-preview |
| Embeddings | text-embedding-ada-002 (1536 dims) |
| Issue | Quick Fix |
|---|---|
| Can't login to Azure | Ask instructor to verify your account |
| SQL "Login failed" | Use Entra ID auth, not SQL auth |
| OpenAI "Not found" | Check API version is 2025-01-01-preview |
| Fell behind on a module | Ask instructor for checkpoint |