The model endpoint your procurement team already approved.
For many enterprises the AI question was settled when Azure OpenAI was approved: the data agreement exists, the region is chosen, the billing runs through an existing subscription. NopsAI treats that as a provider like any other, and adds the governance layer around it.
Azure OpenAI inside a governed run.
How Azure OpenAI participates once it is connected — and what the platform does with what it returns.
Deployment-based configuration
An Azure OpenAI profile takes your resource base URL, the deployment name as the model, and a credential reference. NopsAI builds the chat completions URL from the base, so both bare-resource and versioned base URLs work.
All three fields are required
Validation rejects a profile missing the base URL, the deployment or the credential. A half-configured provider fails at configuration time rather than at the first production run.
Scoped per environment
allowed_scopes decides which environments may select the profile, so a production deployment with a stricter data agreement can be reserved for production scopes.
Usage attributed to a subject
AI usage events record the effective subject alongside the profile, which is what lets you answer who drove spend against an Azure deployment rather than only how much was spent.
The controls that make it safe to leave connected.
An integration is easy to add and hard to bound. These are the parts that decide whether it stays reviewable six months later.
Keys are references, never values
The API key lives in the encrypted credential registry with versioning, rotation and consumer access logs. Configuration in Git holds credential://system/llm/... and nothing else, so the repository is safe to review widely.
Region stays your decision
NopsAI calls the base URL you configure. Data residency for inference is whatever your Azure resource is set to; the platform adds no routing of its own.
Content boundaries are enforced upstream
What may be sent is decided by the pipeline's content sharing settings before the request is built, so an approved provider does not imply approval to send it everything in the workspace.
Profile selection is reviewed
Adding or changing a profile is a change to system configuration in Git, with the same review and promotion path as any other platform change.
Owned by Git, reviewed like code.
Configuration is a reviewable file rather than a form someone filled in once. Credentials appear as references; the values live in the encrypted registry.
profiles:
- name: azure-standard
provider: azure-openai
model: gpt-4o-prod
base_url: https://your-resource.openai.azure.com/openai/v1
credential_ref: credential://system/llm/azure-openai
timeout_seconds: 60
max_tokens: 4096
allowed_scopes: ["prod"]Workflows where Azure OpenAI shows up.
The same integration, seen from the workflow rather than from the connection.
Production incident at 2 a.m.
At 2 a.m. the hard part is finding what changed and deciding whether rollback is safe — fast, without losing the evidence trail.
Preparing a release
Scope mismatch, unmerged PRs, failed gates and an outdated runbook can all cause a late surprise.
Monthly status report to leadership
The report takes hours, invites copy-paste mistakes, and is often stale by the time leadership reads it.
What teams ask about Azure OpenAI.
Do I use the deployment name or the model name?
The deployment name. Azure OpenAI addresses models by the deployment you created, so that value goes in the model field while base_url points at your resource.
Does NopsAI need the full versioned endpoint?
No. You can supply the bare resource URL or one already ending in /openai/v1, and NopsAI builds the chat completions path from it in both cases.
Can different teams use different Azure deployments?
Yes. Define one profile per deployment with its own allowed_scopes, then have each team's pipelines reference the profile they are entitled to. The scope check happens before the run is dispatched.
Map your Azure OpenAI workflow end to end.
Bring one real workflow. We will map its trigger, tools, approvals, runtime boundary and evidence requirements with you.