Azure OpenAI¶

Use the azure-openai provider to call Azure OpenAI deployments through the v1 API. It supports chat and embeddings. Chat requests default to Responses; Chat Completions is also available.

Get started¶

Install Republic using the quickstart, set REPUBLIC_AZURE_OPENAI_API_KEY, set REPUBLIC_AZURE_OPENAI_RESOURCE to your resource name, and replace DEPLOYMENT with a deployment name:

import asyncio

import republic


async def main():
    model = republic.get_model("azure-openai:DEPLOYMENT")
    response = await model.chat("Say hello in one sentence.")
    print(response.text)


asyncio.run(main())

Authentication¶

Republic reads REPUBLIC_AZURE_OPENAI_API_KEY when the provider is created and sends the key in the api-key header. You can pass a key loaded by your application as api_key=; see API-key authentication. For Microsoft Entra ID, pass an auth= object that sends Authorization: Bearer with a token for the https://ai.azure.com/.default scope.

Model support¶

The model name is the deployment name. Republic builds the endpoint https://RESOURCE.openai.azure.com/openai/v1 from the resource name, read from resource= on AzureOpenAI or REPUBLIC_AZURE_OPENAI_RESOURCE. A full api_base= or REPUBLIC_AZURE_OPENAI_API_BASE, such as a services.ai.azure.com endpoint, takes precedence. Creating the provider without a resource or base URL raises ValueError.

from republic.providers import AzureOpenAI

model = AzureOpenAI(resource="my-resource").get_model("DEPLOYMENT")

Select api_format="chat" for deployments without Responses support. Use get_embedding_model() for embedding deployments.

See configuration for endpoint overrides and request options.