Skip to content
Guides & References

Azure OpenAI for Iris: keep network config AI inside your Azure tenant

Run Iris on an Azure OpenAI deployment in your own tenant and region, under your Microsoft agreement. Design, setup notes, pros and cons.

Iris: AI CenterReference designUpdated
On this page

Azure OpenAI for Iris: keep network config AI inside your Azure tenant

In this design, Iris’s features run on a model deployed in your own Azure subscription. Configs go to Azure, in the region you chose, under the agreement your organisation already has with Microsoft. For many enterprises that’s the difference between a quick yes from security and a six-month review.

What it looks like

flowchart LR
    R["rConfig server"] -- "HTTPS" --> A["Azure OpenAI resource<br/>your subscription, your region"]
    A --> D["Your model deployment"]
    R -.-> L[("rConfig audit trail<br/>metadata only")]
    A -.-> M[("Azure monitoring<br/>and quotas")]
Component Role
rConfig Builds each request and records metadata in its audit trail
Azure OpenAI resource Your endpoint, keys, network rules and quotas
Model deployment The model you chose, under a deployment name you set

Setup notes

The provider setup page has the steps. Three details trip people up:

  • The endpoint goes in the form. From V8.4.2, paste the resource root from Keys and Endpoint into Endpoint. Azure providers saved before V8.4.2 without an endpoint keep reading AZURE_OPENAI_URL from .env.
  • Use the deployment name in Deployment name, not the model ID.
  • Network path. rConfig’s server makes the call, so its DNS and firewall decide whether it reaches a public Azure endpoint or a private one.

Pros

  • Data stays in your tenant and region, under an agreement you’ve already reviewed.
  • Access, quotas, network rules and monitoring are managed in Azure, alongside your other workloads.
  • Model quality close to the public API, without running GPUs.

Cons

  • Model availability varies by region, and new models can arrive later than on the public API.
  • Per-token cost on your Azure bill, with Azure quotas to manage.
  • Still a cloud service: not suitable where configs must not leave the site at all.

Frequently asked questions

Why use Azure OpenAI instead of OpenAI directly with rConfig?

Azure OpenAI runs in your own Azure subscription and region, under your existing Microsoft agreement, with access controlled in Azure. Many organisations have already approved Azure for sensitive data, which shortens the security review.

What model name do I enter in rConfig for Azure?

The deployment name you created in Azure, not the model family name.

Can rConfig reach Azure OpenAI over a private endpoint?

rConfig calls whatever endpoint URL you configure, so if your network routes the Azure OpenAI hostname to a private endpoint, rConfig uses it. Check name resolution and routing from the rConfig server itself.

Type to search.