AI / Azure / Agent

Customer Service Agent

A Microsoft Foundry customer-service demo described by its README: a React workbench connects through a FastAPI BFF to a Hosted Agent, with Azure Container Apps, Entra identity, Azure Table Storage sessions, and ticket-style handoff framing the end-to-end study.

Public code / learning case / README-based description; not running on this site

README

Architecture reading map (based on repository description)

  1. 01React workbench
  2. 02FastAPI BFF
  3. 03Azure Container Apps
  4. 04Foundry Hosted Agent
  5. 05Azure AI Search / Table Storage

What to study

  • User preferences need per-user isolation
  • RAG citations should be structured and traceable
  • Synthetic read-only orders are not a real customer-service system

Architecture review dimensions

  • Business problem: what user task does the README actually name?
  • Compute: which runtime boundary is stated, and which remains unknown?
  • Data: what source, retention, and evaluation evidence is public?
  • Network: what ingress, private access, or DNS claim is explicit versus absent?
  • Identity: where would RBAC, user isolation, and secrets be reviewed?
  • Observability and cost: what would need logs, alerts, budgets, and measured load before production?

Boundaries

Not running on this site; it includes no real human dispatch, production endpoint, customer data, or availability claim.