What we will build

A fully working AI Agent–powered Customer Support App running on Azure:

• Frontend — React 18 + Vite + TypeScript on Azure App Service
• Backend — Python 3.11 + FastAPI on Azure App Service
• AI — Microsoft Foundry Agent (gpt-4o-mini) with three function tools: create_ticket, lookup_ticket, list_open_tickets
• Data — Azure Database for PostgreSQL Flexible + Azure Blob Storage
• Identity — Managed Identity end-to-end, no API keys
• IaC — Bicep, deployed with Azure Developer CLI (azd up)

Three deployment paths covered

1. Local — uvicorn + npm run dev for inner-loop coding.
2. azd up — one-shot provisioning + deployment of the whole stack.
3. GitHub Actions — CI per layer (backend-ci, frontend-ci, infra-ci) plus a deploy-azd workflow that runs azd up from CI using OIDC federated credentials — no client secrets in the repo.

Production hardening focus

A dedicated section on Azure Key Vault: why secrets must never live in .env, how to wire @Microsoft.KeyVault(SecretUri=…) references into App Service config, and why Managed Identity + RBAC is the modern alternative to access policies.

Repo

The full source code, Bicep templates, GitHub Actions workflows, and demo script are open-sourced at:

github.com/MMAUG-ORG/build-ai-app-with-MSFoundry-Azure

Recording

Build AI App with Microsoft Foundry & Azure - Part 1

This Part 1 recording is from the May 9, 2026 technical hands-on lab. During the live session, the backend app failed, so participants did not get the end-to-end application flow demo. Part 2 of the video will capture the completed app-flow demo and will be added to this page when it is available.