ReasoningAzureOpenAIChatClient<T>

Models & Types · AiDotNet.Agentic.Models.Connectors

An Azure OpenAI connector that applies the same reasoning-model request rules as ReasoningOpenAIChatClient.

Inheritance: ChatClientBase<T> → OpenAIChatClient<T> → ReasoningOpenAIChatClient<T> → ReasoningAzureOpenAIChatClient<T>

For Beginners

Use this instead of the plain Azure connector when your Azure deployment might be one of the newer reasoning models. It fixes the request settings those models refuse, exactly as the OpenAI version does, and behaves identically to the plain connector for every other model. Name your Azure deployment after the model it runs (for example o3-mini) so the library can tell what it is.

How It Works

Azure hosts the same o-series and GPT-5 weights behind a per-deployment URL and an api-key header, and enforces exactly the same parameter rules as the public endpoint. That is precisely why reasoning-model detection keys off the model id rather than the base URL: a check on the endpoint would classify this client wrongly, while the deployment name resolves through the same profile registry the OpenAI connector uses.

One Azure-specific caveat is worth knowing: on Azure the model id this client sees is your deployment name, not the underlying model. A deployment of o3-mini named o3-mini is recognised automatically; one named reasoning-prod is not, because nothing in that name says what it runs. Give the deployment a name that starts with the model family, or supply a registry whose profile claims your naming convention.

The class is left open for derivation, like ReasoningOpenAIChatClient, so a gateway that changes only the endpoint shape or the authentication header can subclass it rather than duplicating the reasoning rules.

Constructors

Constructor Summary
ReasoningAzureOpenAIChatClient(string apiKey, string deploymentName, string resourceEndpoint, string apiVersion, HttpClient httpClient, ReasoningModelOptions options) Initializes a reasoning-aware Azure OpenAI chat client.