Integrate Azure OpenAI Models into .NET Apps

This .NET guide describes a client library for Azure-hosted language services and compatible endpoints. It covers installing the package, configuring a top-level client, authentication, chat completions, streaming, conversation history, JSON Schema responses, reasoning options, search integration, embeddings, image generation, audio transcription and speech, function calls, error handling, and related libraries.

Use this reference when building a .NET application that calls Azure-hosted chat, embedding, image, or audio capabilities through the documented client hierarchy. It also includes a compatibility path for direct service access and references migration and quickstart resources. The source is scoped to this client integration; it is not a general guide to unrelated .NET programming. Package versions and service capabilities should be checked against the official documentation linked by the source before implementation.

Inputs include the Azure endpoint, service deployment name, messages, generation options, schemas, search configuration, or media files. Authentication examples include an API key credential and Microsoft Entra ID through DefaultAzureCredential, with Entra ID recommended for production. Chat requests can be synchronous or asynchronous, accept options such as output token count, temperature, top-p, and penalties, and return content and usage information. Streaming examples iterate over response updates. Multi-turn use keeps prior messages and appends the returned response.

A namespace import from the source is:

using Azure.Identity;

Structured output examples provide a strict JSON Schema and parse the returned text with System.Text.Json. Search examples connect Azure AI Search for retrieval-augmented generation, and embedding examples can send multiple texts and inspect vector dimensions. The media sections show generating an image with size, quality, and style settings, transcribing audio with a selected response format and language, and generating speech with a voice, speed, and MP3 output.

The guide covers Azure Identity, Azure AI Search, .NET client libraries, NuGet installation, JSON Schema, and service endpoints. Function calling defines a tool name, description, and parameter schema; after a response indicates tool calls, the application inspects function names and arguments. Validate each function name and its arguments before executing the function.

Operational recommendations include reusing client instances, applying exponential backoff for rate limits, streaming long responses, setting suitable timeouts, using structured outputs for consistent formats, and tracking token usage. Error examples distinguish rate limiting and bad requests from other service errors. Related packages listed include Azure Identity and Azure Search Documents.

.NET developers integrating Azure language services can use this reference to choose client operations, configure credentials, handle responses, and connect search or media workflows.