azure-ai-inference

name	azure-ai-inference
description	Use when calling Azure-hosted AI models via the Azure.AI.Inference SDK. Covers chat completions, embeddings, streaming, model selection, and Azure AI model catalog integration. USE FOR: calling Azure AI model catalog models, Azure OpenAI chat completions, generating embeddings from Azure-hosted models, streaming chat responses, working with Mistral/Cohere/Llama models on Azure DO NOT USE FOR: provider-agnostic AI abstractions (use microsoft-extensions-ai), building agent workflows with tools (use agent-framework), on-device ML inference (use onnx), training custom ML models (use mlnet)
license	MIT
metadata	{"displayName":"Azure.AI.Inference","author":"Tyler-R-Kendrick","version":"1.0.0"}
compatibility	claude, copilot, cursor
references	[{"title":"Azure AI Model Inference API Documentation","url":"https://learn.microsoft.com/azure/ai-studio/reference/reference-model-inference-api"},{"title":"Azure SDK for .NET GitHub Repository","url":"https://github.com/Azure/azure-sdk-for-net"},{"title":"Azure.AI.Inference NuGet Package","url":"https://www.nuget.org/packages/Azure.AI.Inference"}]

Azure.AI.Inference

Overview

Azure.AI.Inference is the .NET SDK for calling models deployed in the Azure AI model catalog (including OpenAI, Mistral, Cohere, Meta Llama, and others) through a unified API. It provides ChatCompletionsClient for chat/text generation and EmbeddingsClient for vector embeddings, both supporting streaming, tool calling, and structured output with Azure credential integration.

NuGet Packages

dotnet add package Azure.AI.Inference
dotnet add package Azure.Identity   # For managed identity / DefaultAzureCredential

Chat Completions

using Azure;
using Azure.AI.Inference;

var endpoint = new Uri("https://my-model.eastus.models.ai.azure.com");
var credential = new AzureKeyCredential(Environment.GetEnvironmentVariable("AZURE_AI_KEY")!);

var client = new ChatCompletionsClient(endpoint, credential);

var requestOptions = new ChatCompletionsOptions
{
    Messages =
    {
        new ChatRequestSystemMessage("You are a helpful coding assistant specializing in C#."),
        new ChatRequestUserMessage("Write a method to validate an email address using regex.")
    },
    Temperature = 0.7f,
    MaxTokens = 1024
};

ChatCompletions response = await client.CompleteAsync(requestOptions);
Console.WriteLine(response.Choices[0].Message.Content);
Console.WriteLine($"Tokens used: {response.Usage.TotalTokens}");

Streaming Chat Completions

var requestOptions = new ChatCompletionsOptions
{
    Messages =
    {
        new ChatRequestSystemMessage("You are a technical writer."),
        new ChatRequestUserMessage()
    },
    Temperature = 
};

  (StreamingChatCompletionsUpdate update 
    client.CompleteStreamingAsync(requestOptions))
{
     (update.ContentUpdate  { } content)
    {
        Console.Write(content);
    }
}
Console.WriteLine();