AzureAIChatCompletionsModel

This will help you getting started with AzureAIChatCompletionsModel chat models. For detailed documentation of all AzureAIChatCompletionsModel features and configurations head to the API reference

The AzureAIChatCompletionsModel class uses the Azure AI Foundry SDK. AI Foundry has several chat models including AzureOpenAI, Cohere, Llama, Phi-3/4, and DeepSeek-R1 to name a few. You can find information about their latest models and their costs, context windows, and supported input types in the Azure docs.

Overview

Integration details

Class	Package	Local	Serializable	JS support	Package downloads	Package latest
AzureAIChatCompletionsModel	langchain-azure-ai	❌	✅	✅

Model features

Tool calling	Structured output	JSON mode	Image input	Audio input	Video input	Token-level streaming	Native async	Token usage	Logprobs
✅	✅	✅	✅	❌	❌	✅	✅	✅	✅

Setup

To access AzureAIChatCompletionsModel models you'll need to create an Azure account, get an API key, and install the langchain-azure-ai integration package.

Credentials

Head to the Azure docs to see how to create your deployment and generate an API key. Once your model is deployed you click the 'get endpoint' button in AI Foundry. This will show you your endpoint and api key. Once you've done this set the AZURE_INFERENCE_CREDENTIAL and AZURE_INFERENCE_ENDPOINT environment variables:

import getpass
import os

if not os.getenv("AZURE_INFERENCE_CREDENTIAL"):
    os.environ["AZURE_INFERENCE_CREDENTIAL"] = getpass.getpass(
        "Enter your AzureAIChatCompletionsModel API key: "
    )

if not os.getenv("AZURE_INFERENCE_ENDPOINT"):
    os.environ["AZURE_INFERENCE_ENDPOINT"] = getpass.getpass(
        "Enter your model endpoint: "
    )

If you want to get automated tracing of your model calls you can also set your LangSmith API key by uncommenting below:

# os.environ["LANGCHAIN_TRACING_V2"] = "true"
# os.environ["LANGCHAIN_API_KEY"] = getpass.getpass("Enter your LangSmith API key: ")

Installation

The LangChain AzureAIChatCompletionsModel integration lives in the langchain-azure-ai package:

%pip install -qU langchain-azure-ai

Instantiation

Now we can instantiate our model object and generate chat completions:

from langchain_azure_ai.chat_models import AzureAIChatCompletionsModel

llm = AzureAIChatCompletionsModel(
    model_name="gpt-4",
    temperature=0,
    max_tokens=None,
    timeout=None,
    max_retries=2,
)

API Reference:AzureAIChatCompletionsModel

Invocation

messages = [
    (
        "system",
        "You are a helpful assistant that translates English to French. Translate the user sentence.",
    ),
    ("human", "I love programming."),
]
ai_msg = llm.invoke(messages)
ai_msg

AIMessage(content="J'adore programmer.", additional_kwargs={}, response_metadata={'model': 'gpt-4o-2024-05-13', 'token_usage': {'input_tokens': 31, 'output_tokens': 4, 'total_tokens': 35}, 'finish_reason': 'stop'}, id='run-c082dffd-b1de-4b3f-943f-863836663ddb-0', usage_metadata={'input_tokens': 31, 'output_tokens': 4, 'total_tokens': 35})

print(ai_msg.content)

J'adore programmer.

Chaining

We can chain our model with a prompt template like so:

from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate(
    [
        (
            "system",
            "You are a helpful assistant that translates {input_language} to {output_language}.",
        ),
        ("human", "{input}"),
    ]
)

chain = prompt | llm
chain.invoke(
    {
        "input_language": "English",
        "output_language": "German",
        "input": "I love programming.",
    }
)

API Reference:ChatPromptTemplate

AIMessage(content='Ich liebe Programmieren.', additional_kwargs={}, response_metadata={'model': 'gpt-4o-2024-05-13', 'token_usage': {'input_tokens': 26, 'output_tokens': 5, 'total_tokens': 31}, 'finish_reason': 'stop'}, id='run-01ba6587-6ff4-4554-8039-13204a7d95db-0', usage_metadata={'input_tokens': 26, 'output_tokens': 5, 'total_tokens': 31})

API reference

For detailed documentation of all AzureAIChatCompletionsModel features and configurations head to the API reference: https://python.langchain.com/api_reference/azure_ai/chat_models/langchain_azure_ai.chat_models.AzureAIChatCompletionsModel.html

Related

• Chat model conceptual guide
• Chat model how-to guides