Class ChatClientProvider

Namespace: IntentSharp.Extensions.AI

Assembly: IntentSharp.Extensions.AI.dll

Runs IntentSharp on any Microsoft.Extensions.AI Microsoft.Extensions.AI.IChatClient: Azure OpenAI, OpenAI, Ollama, the Agent Framework and every other backend that implements the abstraction. The chat client only carries each request and its reply; the checks on the result, further attempts, the judge and memory work as they do with the built-in providers.

var client = new IntentClientBuilder()
    .UseChatClient(chatClient)
    .Build();

public sealed class ChatClientProvider : IAIProvider

Inheritance: object

ChatClientProvider

Implements: IAIProvider

Inherited Members: object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.ReferenceEquals(object, object)

object.ToString()

Extension Methods: IntentExtensions.RunIntentAsync(object?, IntentRuntime, string, CancellationToken)

IntentExtensions.RunIntentAsync(object?, string, CancellationToken)

IntentExtensions.RunIntentAsync<T>(object?, IntentRuntime, string, CancellationToken)

IntentExtensions.RunIntentAsync<T>(object?, string, CancellationToken)

AIExtensions.AIAsync(object?, IntentClient, string, AIRequestOptions?, CancellationToken)

AIExtensions.AIAsync<T>(object?, IntentClient, string, AIRequestOptions?, CancellationToken)

AIExtensions.AIClassifyAsync<TEnum>(object?, IntentClient, string?, AIRequestOptions?, CancellationToken)

AIExtensions.AIConvertAsync<T>(object?, IntentClient, AIRequestOptions?, CancellationToken)

AmbientAIExtensions.AIAsync(object?, string, AIRequestOptions?, CancellationToken)

AmbientAIExtensions.AIAsync<T>(object?, string, AIRequestOptions?, CancellationToken)

AmbientAIExtensions.AIClassifyAsync<TEnum>(object?, string?, AIRequestOptions?, CancellationToken)

AmbientAIExtensions.AIConvertAsync<T>(object?, AIRequestOptions?, CancellationToken)

Remarks

Each AIRequest becomes a system message (when SystemPrompt is set) followed by the request's messages, and Microsoft.Extensions.AI.ChatOptions carrying the model, output token ceiling, temperature and response format (ForJsonSchema(JsonElement, string, string) for a JSON schema, Microsoft.Extensions.AI.ChatResponseFormat.Json for any JSON object, none for text). Request tags are never sent.

Failures thrown by the chat client are translated to the IntentSharp provider exceptions so the client's fallback rules apply unchanged: an HTTP status found on the exception (StatusCode, or a public Status or StatusCode property as on the OpenAI and Azure SDK exceptions) decides the type, and a transport failure or a timeout inside the chat client is transient. This provider does not retry by itself; wrap the chat client in one that retries, or rely on the fallback provider.

The caller owns the chat client and the embedding generator. Disposing the IntentClient does not dispose them.

Constructors

ChatClientProvider(IChatClient, IEmbeddingGenerator<string, Embedding<float>>?, string?, string?, ILogger?): Wraps the chat client, and optionally an embedding generator, as a provider. The name and the default model come from the chat client's metadata unless they are given here.

Fields

DefaultName: The provider name used when neither the constructor nor the client's metadata supplies one.

Properties

ChatClient: The wrapped chat client.

DefaultModel: Model used when a request does not name one.

Name: Stable provider name used in results, rate limiting and logs (for example "OpenAI").

SupportsEmbeddings: True when EmbedAsync(string, CancellationToken) is implemented.

Methods

CompleteAsync(AIRequest, CancellationToken): Sends a chat-style request and returns the model's reply.