Azure OpenAI

Azure OpenAI provides access to OpenAI's models through Microsoft's Azure cloud platform. You get the same embedding models as direct OpenAI access, but with Azure's enterprise features: regional deployments, private endpoints, managed identities, and compliance certifications.

If you're already running on Azure or need to meet specific compliance requirements, Azure OpenAI is the natural choice. The models are identical to OpenAI's—same quality, same dimensions—but the infrastructure is Azure's.

Setup

First, create an Azure OpenAI resource in the Azure portal and deploy an embedding model. Note your resource name and deployment name.

Install the Azure AI SDK package:

bun add @ai-sdk/azure

Set your credentials in the environment:

AZURE_OPENAI_API_KEY="..."
AZURE_RESOURCE_NAME="your-resource-name"

Configure the provider in your unrag.config.ts:

import { defineUnragConfig } from "./lib/unrag/core";

export const unrag = defineUnragConfig({
  // ...
  embedding: {
    provider: "azure",
    config: {
      model: "text-embedding-3-small",  // Your deployment name
      timeoutMs: 15_000,
    },
  },
} as const);

Configuration options

model specifies which deployment to use. This should match the deployment name you created in the Azure portal. If not set, the provider checks the AZURE_EMBEDDING_MODEL environment variable, then falls back to text-embedding-3-small.

timeoutMs sets the request timeout in milliseconds.

dimensions enables dimension truncation for embedding-3 models, same as with direct OpenAI access.

user is an optional end-user identifier for abuse detection.

embedding: {
  provider: "azure",
  config: {
    model: "my-embedding-deployment",
    dimensions: 1024,
    timeoutMs: 20_000,
  },
},

Environment variables

AZURE_OPENAI_API_KEY (required): Your Azure OpenAI API key.

AZURE_RESOURCE_NAME (required): The name of your Azure OpenAI resource.

AZURE_EMBEDDING_MODEL (optional): Overrides the deployment name specified in code.

# .env
AZURE_OPENAI_API_KEY="..."
AZURE_RESOURCE_NAME="my-openai-resource"
AZURE_EMBEDDING_MODEL="text-embedding-3-small"

Azure vs direct OpenAI

The models are the same. The difference is infrastructure and compliance.

Use Azure OpenAI when you need Azure's compliance certifications (SOC 2, HIPAA, etc.), want to keep all your services in Azure, need private network access, or are required to use Azure for organizational reasons.

Use direct OpenAI when you want the simplest setup, don't have Azure requirements, or prefer OpenAI's pricing and terms.