Implementation:Langchain ai Langchain AzureOpenAI LLM

Knowledge Sources	Langchain_ai_Langchain
Domains	LLM, Azure OpenAI, Text Completion
Last Updated	2026-02-11 00:00 GMT

Overview

AzureOpenAI is a LangChain LLM class for text completion using Azure-hosted OpenAI models, extending BaseOpenAI with Azure-specific endpoint and authentication configuration.

Description

The AzureOpenAI class extends BaseOpenAI from the langchain-openai package to provide Azure-specific LLM text completion capabilities. This is distinct from Azure OpenAI chat models -- it targets the completions API (e.g., gpt-3.5-turbo-instruct). The class supports Azure endpoints, deployment names, API versioning, and multiple authentication methods including API keys, Azure AD tokens, and AD token providers (both sync and async). It creates openai.AzureOpenAI and openai.AsyncAzureOpenAI client instances pointed at the completions endpoint. It includes backward-compatibility validation for the transition from openai_api_base to azure_endpoint. The class is LangChain-serializable and integrates with LangSmith tracing.

Usage

Import this class when you need text completion (not chat completion) from Azure-hosted OpenAI models in LangChain chains, agents, or direct generation workflows.

Code Reference

Source Location

• Repository: Langchain_ai_Langchain
• File: libs/partners/openai/langchain_openai/llms/azure.py
• Lines: 1-233

Signature

class AzureOpenAI(BaseOpenAI):
    azure_endpoint: str | None = Field(...)
    deployment_name: str | None = Field(default=None, alias="azure_deployment")
    openai_api_version: str | None = Field(alias="api_version", ...)
    openai_api_key: SecretStr | None = Field(alias="api_key", ...)
    azure_ad_token: SecretStr | None = Field(...)
    azure_ad_token_provider: Callable[[], str] | None = None
    azure_ad_async_token_provider: Callable[[], Awaitable[str]] | None = None
    openai_api_type: str | None = Field(...)
    validate_base_url: bool = True

Import

from langchain_openai import AzureOpenAI

I/O Contract

Inputs

Name	Type	Required	Description
azure_endpoint	None	No	Azure endpoint URL (e.g., https://example-resource.azure.openai.com/). Read from AZURE_OPENAI_ENDPOINT env var.
deployment_name	None	No	Azure model deployment name. Alias: azure_deployment.
openai_api_version	None	No	API version. Read from OPENAI_API_VERSION env var.
openai_api_key	None	No	API key. Read from AZURE_OPENAI_API_KEY or OPENAI_API_KEY env vars.
azure_ad_token	None	No	Azure AD token. Read from AZURE_OPENAI_AD_TOKEN env var.
azure_ad_token_provider	Callable[[], str] | None	No	Function returning an Azure AD token, invoked on every sync request.
azure_ad_async_token_provider	Callable[[], Awaitable[str]] | None	No	Async function returning an Azure AD token, invoked on every async request.
openai_api_type	None	No	API type. Read from OPENAI_API_TYPE env var. Default: "azure".
validate_base_url	bool	No	Whether to validate and transform legacy base URL. Default: True.
model_name	str	No	Model name (inherited from BaseOpenAI).

Outputs

Name	Type	Description
_call return	str	Generated completion text (inherited from BaseOpenAI).
generate return	LLMResult	Structured generation results (inherited from BaseOpenAI).

Properties

Property	Return Type	Description
_llm_type	str	Returns "azure".
_identifying_params	Mapping[str, Any]	Returns deployment_name plus parent identifying params.
_invocation_params	dict[str, Any]	Includes model set to deployment_name plus parent invocation params.
lc_attributes	dict[str, Any]	Returns openai_api_type and openai_api_version for tracing.
lc_secrets	dict[str, str]	Maps openai_api_key to AZURE_OPENAI_API_KEY and azure_ad_token to AZURE_OPENAI_AD_TOKEN.

Class Methods

Method	Return Type	Description
get_lc_namespace()	list[str]	Returns ["langchain", "llms", "openai"].
is_lc_serializable()	bool	Returns True, indicating this model can be serialized by LangChain.

Usage Examples

Basic Usage

from langchain_openai import AzureOpenAI

llm = AzureOpenAI(
    model_name="gpt-3.5-turbo-instruct",
    azure_deployment="my-deployment",
    azure_endpoint="https://my-resource.openai.azure.com/",
    # api_key="your-api-key",  # Or set AZURE_OPENAI_API_KEY env var
    # api_version="2024-02-01",  # Or set OPENAI_API_VERSION env var
)

# Generate text
response = llm.invoke("Tell me a joke about programming.")
print(response)

With Azure AD Authentication

from langchain_openai import AzureOpenAI
from azure.identity import DefaultAzureCredential

credential = DefaultAzureCredential()

llm = AzureOpenAI(
    model_name="gpt-3.5-turbo-instruct",
    azure_deployment="my-deployment",
    azure_endpoint="https://my-resource.openai.azure.com/",
    azure_ad_token_provider=lambda: credential.get_token(
        "https://cognitiveservices.azure.com/.default"
    ).token,
)

Related Pages

• Requires langchain-openai and openai packages
• Extends langchain_openai.llms.base.BaseOpenAI