Building a retrieval-augmented generation (RAG) app: the moving parts
The core components of a RAG application, and where teams usually go wrong when moving from a demo to production.
· 1 min read
By Srinivasa Rao G · · 1 min read
A hub-and-spoke topology is the most common starting point for Azure networking. A central hub virtual network hosts shared services, and workload spokes peer to it.
Each spoke holds one workload or environment, such as a production application or a dev/test subscription. Spokes peer to the hub but not directly to each other, which keeps traffic paths predictable and auditable.
| Decision | Good default |
|---|---|
| Address space | Plan non-overlapping ranges up front, with room to grow |
| Subscriptions | One per workload or environment |
| Routing | User-defined routes sending egress through the hub firewall |
| Policy | Azure Policy assigned at a management group, not per subscription |
az group create --name rg-hub-eus --location eastus
az network vnet create \
--resource-group rg-hub-eus \
--name vnet-hub-eus \
--address-prefixes 10.0.0.0/22 \
--subnet-name AzureFirewallSubnet \
--subnet-prefixes 10.0.0.0/26
Treat this as a skeleton: define naming, tagging and policy before the first production workload lands, because they are far harder to retrofit later.
The core components of a RAG application, and where teams usually go wrong when moving from a demo to production.
· 1 min read