Azure OpenAI

LogicGrid talks to Azure OpenAI over its native per-deployment URL layout, using the api-key header for authentication. The wire format is the same OpenAI Chat Completions schema, but the model you call is actually a deployment name you create in the Azure portal — it maps to an underlying model (gpt-4o, gpt-4o-mini, etc.).

• Resource & deployment management: Azure OpenAI Studio
• Models and regions: learn.microsoft.com/azure/ai-services/openai/concepts/models
• Pricing: azure.microsoft.com/pricing/details/cognitive-services/openai-service

Deployment name vs. model name

In Azure OpenAI you create a deployment in the portal that maps a model (e.g. gpt-4o) to a logical name (e.g. my-gpt-4o-deployment). LogicGrid's deploymentName argument is the logical name; pass the underlying model id as underlyingModel so the client can look up per-token pricing for cost tracking.

Use it

There are three equivalent ways to instantiate the Azure OpenAI client.

Option 1 — static factory (recommended)

using LogicGrid.Core.Llm;

var llm = LlmClientBase.AzureOpenAi(
    endpoint:        "https://my-resource.openai.azure.com",
    apiKey:          Environment.GetEnvironmentVariable("AZURE_OPENAI_KEY")!,
    deploymentName:  "my-gpt-4o-deployment",
    apiVersion:      "2024-10-21",
    underlyingModel: "gpt-4o");      // optional — used for cost lookup

Parameter	Type	Default	Notes
endpoint	string	(required)	Azure OpenAI resource URL, e.g. https://my-resource.openai.azure.com.
apiKey	string	(required)	Resource API key, sent as the api-key header.
deploymentName	string	(required)	The deployment name from the Azure portal — not the model id.
apiVersion	string	"2024-10-21"	Azure OpenAI API version. Older versions may not see newer deployments.
underlyingModel	string?	null	Optional model id (e.g. gpt-4o) for per-token pricing lookup.

Option 2 — direct construction

using LogicGrid.Core.Providers;

var llm = new AzureOpenAiClient(
    endpoint:        "https://my-resource.openai.azure.com",
    apiKey:          Environment.GetEnvironmentVariable("AZURE_OPENAI_KEY")!,
    deploymentName:  "my-gpt-4o-deployment",
    apiVersion:      "2024-10-21",
    underlyingModel: "gpt-4o");

Identical parameters to the factory. Use direct construction when you need an injected HttpClient explained below (for retries, proxies, or testing).

Option 3 — injected HttpClient

using LogicGrid.Core.Providers;

var llm = new AzureOpenAiClient(
    httpClient:      myHttpClient,
    endpoint:        "https://my-resource.openai.azure.com",
    deploymentName:  "my-gpt-4o-deployment",
    apiVersion:      "2024-10-21",
    underlyingModel: "gpt-4o");

This overload takes no apiKey — the caller is responsible for setting the api-key header on the supplied HttpClient (typically via IHttpClientFactory, a DelegatingHandler, or test fakes). Use it when auth is managed outside the client, or in unit tests with a mocked transport.

Tool calling

Azure OpenAI supports the same native tool-call protocol as OpenAI. The default PromptSchemaStrategy works everywhere; switch to NativeToolCallingStrategy when targeting gpt-4o and similar models. See Tool calling strategy.

Troubleshooting

• 401 — wrong api-key, or the key doesn't belong to the resource at the given endpoint.
• 404 — deploymentName doesn't exist in this resource. Create it in Azure OpenAI Studio.
• DeploymentNotFound despite a successful create — check apiVersion. Older API versions don't see newer deployments.