The OpenAI Python SDK supports Azure OpenAI Service through the AzureOpenAI class.
The Azure API shape differs from the core OpenAI API shape, which means that the static types for responses and parameters won’t always be correct.
Basic Usage
With API Key
from openai import AzureOpenAI
# API key is read from AZURE_OPENAI_API_KEY environment variable
client = AzureOpenAI(
api_version="2023-07-01-preview",
azure_endpoint="https://example-endpoint.openai.azure.com",
)
completion = client.chat.completions.create(
model="deployment-name", # e.g. gpt-35-instant
messages=[
{
"role": "user",
"content": "How do I output all files in a directory using Python?",
},
],
)
print(completion.to_json())
With Azure Deployment
You can specify an azure_deployment to include it in the base URL:
from openai import AzureOpenAI
client = AzureOpenAI(
api_version="2023-07-01-preview",
azure_endpoint="https://example-resource.azure.openai.com/",
azure_deployment="deployment-name", # e.g. gpt-35-instant
)
completion = client.chat.completions.create(
model="<ignored>", # Model parameter is ignored when using azure_deployment
messages=[
{
"role": "user",
"content": "How do I output all files in a directory using Python?",
},
],
)
print(completion.to_json())
The azure_deployment parameter is not supported with Assistants APIs.
Authentication Methods
Azure OpenAI supports multiple authentication methods:
API Key
The simplest method uses an API key:
from openai import AzureOpenAI
client = AzureOpenAI(
api_key="your-azure-api-key", # or use AZURE_OPENAI_API_KEY env var
api_version="2023-07-01-preview",
azure_endpoint="https://example-endpoint.openai.azure.com",
)
Azure Active Directory (Recommended)
For production deployments, use Azure AD authentication:
from azure.identity import DefaultAzureCredential, get_bearer_token_provider
from openai import AzureOpenAI
token_provider = get_bearer_token_provider(
DefaultAzureCredential(),
"https://cognitiveservices.azure.com/.default"
)
client = AzureOpenAI(
api_version="2023-07-01-preview",
azure_endpoint="https://example-endpoint.openai.azure.com",
azure_ad_token_provider=token_provider,
)
Azure AD Token (Static)
You can also pass a static Azure AD token:
from openai import AzureOpenAI
client = AzureOpenAI(
api_version="2023-07-01-preview",
azure_endpoint="https://example-endpoint.openai.azure.com",
azure_ad_token="your-azure-ad-token", # or use AZURE_OPENAI_AD_TOKEN env var
)
Configuration
Required Parameters
The following parameters are required:
api_version: Azure API version (or OPENAI_API_VERSION environment variable)azure_endpoint OR base_url: Your Azure endpoint (or AZURE_OPENAI_ENDPOINT environment variable)
Environment Variables
The SDK automatically reads these environment variables:
export AZURE_OPENAI_API_KEY="your-azure-key"
export AZURE_OPENAI_ENDPOINT="https://example-endpoint.openai.azure.com"
export OPENAI_API_VERSION="2023-07-01-preview"
export AZURE_OPENAI_AD_TOKEN="your-ad-token" # Optional
Available Options
All options from the base OpenAI client are available, plus:
| Parameter | Description |
|---|
azure_endpoint | Your Azure endpoint URL |
azure_deployment | Model deployment name to include in base URL |
api_version | Azure API version |
azure_ad_token | Static Azure AD token |
azure_ad_token_provider | Function that returns an Azure AD token |
Async Client
Use AsyncAzureOpenAI for async operations:
import asyncio
from openai import AsyncAzureOpenAI
async def main():
client = AsyncAzureOpenAI(
api_version="2023-07-01-preview",
azure_endpoint="https://example-endpoint.openai.azure.com",
)
completion = await client.chat.completions.create(
model="deployment-name",
messages=[
{
"role": "user",
"content": "Say this is a test",
}
],
)
print(completion.to_json())
asyncio.run(main())
API Version
Azure OpenAI uses versioned APIs. See the Azure REST API versioning docs for available versions.
Currently recommended version: 2023-07-01-preview
Deployment Names
In Azure OpenAI, you deploy models with custom deployment names. Use your deployment name in the model parameter:
completion = client.chat.completions.create(
model="my-gpt-4-deployment", # Your Azure deployment name
messages=[...],
)
Learn More
