Azure OpenAI

Use OpenAI’s GPT-4o, GPT-4o-mini, o-series, and other models through Azure OpenAI Service. Same OpenAI capabilities, but authenticated with Azure credentials, billed through Azure, and deployed in your own Azure subscription with enterprise compliance.

azureOpenai() uses the openai SDK’s built-in AzureOpenAI class. All OpenAI features — tool calling, streaming, JSON mode, vision, reasoning — work identically to the direct openai() provider.

Setup

Install
Environment

The Azure OpenAI provider uses the same OpenAI SDK:

npm install openai

Set your Azure OpenAI credentials:

export AZURE_OPENAI_API_KEY="..."
export AZURE_OPENAI_ENDPOINT="https://<your-resource>.openai.azure.com"
export AZURE_OPENAI_DEPLOYMENT="gpt-4o"  # optional, can use modelId instead

Factory

import { azureOpenai } from "@agentium/core";

const model = azureOpenai("gpt-4o");

string

required

The model or deployment name. This is passed as the model parameter to the API.

AzureOpenAIConfig

Optional configuration. See Config below.

Supported Models

Model	Description
`gpt-4o`	Latest flagship model. Vision, audio, tool calling.
`gpt-4o-mini`	Smaller, faster, cost-effective.
`gpt-4-turbo`	High capability, larger context.
`o1-preview`	Reasoning-optimized model.
`o3-mini`	Fast reasoning model.

Deploying models on Azure

You must create a deployment for each model in the Azure AI Studio portal. The deployment name becomes the modelId (or you can use config.deployment). See Azure OpenAI docs for setup instructions.

Config

string

Azure OpenAI API key. Falls back to AZURE_OPENAI_API_KEY env var.

string

Azure OpenAI endpoint URL. Falls back to AZURE_OPENAI_ENDPOINT env var. Format: https://<resource-name>.openai.azure.com

string

Deployment name. Falls back to AZURE_OPENAI_DEPLOYMENT env var. If not set, modelId is used.

string

default:"2024-10-21"

Azure API version. Falls back to AZURE_OPENAI_API_VERSION env var.

Authentication Methods

Method 1: API Key (Recommended)

export AZURE_OPENAI_API_KEY="..."
export AZURE_OPENAI_ENDPOINT="https://my-resource.openai.azure.com"

const model = azureOpenai("gpt-4o");

Method 2: Explicit Config

const model = azureOpenai("gpt-4o", {
  apiKey: "...",
  endpoint: "https://my-resource.openai.azure.com",
  deployment: "my-gpt4o-deployment",
  apiVersion: "2024-10-21",
});

Tool Calling

Tool calling works identically to the direct OpenAI provider:

import { Agent, azureOpenai, defineTool } from "@agentium/core";
import { z } from "zod";

const agent = new Agent({
  name: "azure-assistant",
  model: azureOpenai("gpt-4o"),
  instructions: "You are a helpful assistant.",
  tools: [
    defineTool({
      name: "getWeather",
      description: "Get current weather for a city",
      parameters: z.object({ city: z.string() }),
      execute: async ({ city }) => `${city}: 22°C, Partly Cloudy`,
    }),
  ],
});

const result = await agent.run("What's the weather in London?");
console.log(result.text);

Reasoning Models (o-series)

Azure OpenAI supports the o-series reasoning models with the same reasoning config:

const agent = new Agent({
  name: "analyst",
  model: azureOpenai("o3-mini"),
  instructions: "Solve problems step by step.",
  reasoning: {
    effort: "high",
  },
});

const result = await agent.run("Calculate the compound interest on $10,000 at 5% for 10 years, compounded monthly.");
console.log(result.text);

Azure OpenAI GPT-4o supports images, audio, and files — same as the direct OpenAI provider:

import { Agent, azureOpenai, type ContentPart } from "@agentium/core";
import { readFileSync } from "node:fs";

const agent = new Agent({
  name: "vision-agent",
  model: azureOpenai("gpt-4o"),
  instructions: "Analyze images and documents.",
});

const imageData = readFileSync("chart.png").toString("base64");
const result = await agent.run([
  { type: "text", text: "What trends does this chart show?" },
  { type: "image", data: imageData, mimeType: "image/png" },
] as ContentPart[]);

Full Example

import { Agent, azureOpenai, CostTracker, defineTool } from "@agentium/core";
import { z } from "zod";

const costTracker = new CostTracker({
  pricing: {
    "gpt-4o": { promptPer1k: 0.0025, completionPer1k: 0.01 },
  },
});

const agent = new Agent({
  name: "enterprise-assistant",
  model: azureOpenai("gpt-4o", {
    endpoint: "https://my-company.openai.azure.com",
    apiVersion: "2024-10-21",
  }),
  instructions: "You are an enterprise assistant for internal use.",
  tools: [
    defineTool({
      name: "searchDocs",
      description: "Search internal documentation",
      parameters: z.object({ query: z.string() }),
      execute: async ({ query }) => `Found 3 results for "${query}"...`,
    }),
  ],
  costTracker,
  maxToolRoundtrips: 3,
});

const result = await agent.run("Find the onboarding docs for new engineers");
console.log(result.text);
console.log(`Cost: $${costTracker.getSummary().totalCost.toFixed(4)}`);

Azure OpenAI vs Direct OpenAI

Feature	`openai()` (Direct)	`azureOpenai()` (Azure)
Auth	OpenAI API key	Azure API key / Azure AD
Billing	OpenAI billing	Azure billing (consolidated)
Data residency	OpenAI servers	Your Azure region
VPC / Private	No	Yes (Private Endpoints)
Compliance	SOC2	SOC2, HIPAA, FedRAMP, GDPR
Prompt Caching	Yes	Yes
Models	All OpenAI models	Models you deploy

Environment Variables

Variable	Description
`AZURE_OPENAI_API_KEY`	Azure OpenAI API key
`AZURE_OPENAI_ENDPOINT`	Azure OpenAI endpoint URL
`AZURE_OPENAI_DEPLOYMENT`	Default deployment name
`AZURE_OPENAI_API_VERSION`	API version (default: `2024-10-21`)

Cross-References

OpenAI (Direct) — Direct OpenAI API with API key auth
Azure AI Foundry — Open-source models (Phi, Llama, Mistral) on Azure
Reasoning — O-series reasoning configuration
Multi-Modal — Provider support matrix

​Azure OpenAI

​Setup

​Factory

​Supported Models

​Config

​Authentication Methods

​Method 1: API Key (Recommended)

​Method 2: Explicit Config

​Tool Calling

​Reasoning Models (o-series)

​Multi-Modal Support

​Full Example

​Azure OpenAI vs Direct OpenAI

​Environment Variables

​Cross-References

Azure OpenAI

Setup

Factory

Supported Models

Config

Authentication Methods

Method 1: API Key (Recommended)

Method 2: Explicit Config

Tool Calling

Reasoning Models (o-series)

Multi-Modal Support

Full Example

Azure OpenAI vs Direct OpenAI

Environment Variables

Cross-References