Universal.OpenAI.ChatGpt.Client 1.0.6

Universal.OpenAI.ChatGpt.Client

A .NET Standard client for ChatGPT Responses.

This package targets the ChatGPT Responses backend. It is not a client for the OpenAI Platform Responses API. For that API and the broader OpenAI Platform surface, use Universal.OpenAI.Client.

Access token

using Universal.OpenAI.ChatGpt.Client;
using Universal.OpenAI.ChatGpt.Client.Responses;

using var client = new ChatGptResponsesClient(accessToken, accountId);

var response = await client.CreateResponseAsync(new CreateResponseRequest
{
    Model = Models.Gpt56Luna,
    Input = new Output[]
    {
        new MessageOutput(MessageRoles.User, "Explain this code.")
    }
});

Console.WriteLine(response.GetOutputText());

For multi-turn requests, set a stable PromptCacheKey on each request. The client sends the same value in the session-id header, which the ChatGPT backend uses for cache affinity. ResponsesClientOptions.SessionId supplies a fallback header value when the request has no cache key. The backend may report cache reads and writes in Usage.InputTokensDetails; a matching key and prompt prefix do not guarantee a hit.

Read the account plan and usage windows:

using Universal.OpenAI.ChatGpt.Client.Wham;

using var wham = new WhamClient(httpClient, credentials);
Usage usage = await wham.GetUsageAsync();
Console.WriteLine($"{usage.PlanType}: {usage.RateLimit?.PrimaryWindow?.UsedPercent}% used");

RateLimitResetCreditsDetails resetCredits = await wham.ListRateLimitResetCreditsAsync();

List the models currently available to the account:

using Universal.OpenAI.ChatGpt.Client.Models;

using var models = new ModelsClient(httpClient, credentials);
ModelsResponse catalog = await models.ListModelsAsync();

Responses.Models.Gpt6Astra, Gpt61Sol, Gpt6Sol, and Gpt6Luna provide the GPT-6 model identifiers. Availability depends on the ChatGPT account's rollout and can be checked with ModelsClient.

Authentication file

Use an existing auth.json file for automatic token refresh:

using Universal.OpenAI.ChatGpt.Client;
using Universal.OpenAI.ChatGpt.Client.Responses;

using var httpClient = new HttpClient();
using var credentials = new AuthenticationFileCredentialProvider(
    authenticationFilePath,
    httpClient);
using var client = new ChatGptResponsesClient(httpClient, credentials);

var response = await client.CreateResponseAsync(new CreateResponseRequest
{
    Model = Models.Gpt56Luna,
    Input = new Output[] { new MessageOutput(MessageRoles.User, "Explain this code.") }
});

Console.WriteLine(response.GetOutputText());

Browser authentication

Create an authorization session, open its URL, and exchange the callback code:

using Universal.OpenAI.ChatGpt.Client;
using Universal.OpenAI.ChatGpt.Client.Responses;

using var httpClient = new HttpClient();
var oauth = new OAuthClient(httpClient);
var session = oauth.CreateAuthorizationSession();

Console.WriteLine(session.AuthorizationUri);

// callbackUri is the URI received by the configured local callback endpoint.
var code = oauth.ReadAuthorizationCode(callbackUri, session);
var tokens = await oauth.ExchangeCodeAsync(code, session);

using var credentials = new AuthenticationFileCredentialProvider(
    authenticationFilePath,
    httpClient);
credentials.Save(tokens);

using var client = new ChatGptResponsesClient(httpClient, credentials);