.NET Console OpenAI Chat (SDK)

This repository contains a minimal .NET 9 console app that integrates the official OpenAI .NET SDK to run a chat session with:

• Streaming output (tokens print as they arrive — no waiting for full responses)
• Persistent conversation history (system, user, and assistant turns)

The sample targets C# 13 and .NET 9.

Features

• Uses the OpenAI NuGet package (v2.6.x) and the ChatClient API
• Streams assistant responses via CompleteChatStreamingAsync(...)
• Preserves full conversation context by appending UserChatMessage and AssistantChatMessage objects
• Exits the loop on a blank user input

Prerequisites

• .NET 9 SDK installed
• An OpenAI API key with access to the selected chat model
• Network access to the OpenAI API

Getting started

1. Clone/download this repo and open the solution.

2. Restore and build (Visual Studio or CLI):

dotnet restore
dotnet build -c Release

3. Set the OPENAI_API_KEY environment variable:

• Windows (PowerShell):

$env:OPENAI_API_KEY = "<your-api-key>"

• macOS/Linux (bash/zsh):

export OPENAI_API_KEY="<your-api-key>"

4. Run:

dotnet run --project OpenAI.Chat.SDK.csproj

You should see a prompt:

Type your message and press Enter. Submit a blank line to exit.
You: 

Type a message and watch the assistant reply stream in real time. Submit a blank line to exit.

How it works

The core logic lives in Program.cs:

• Create a client using an API key:
  • var client = new OpenAIClient(new ApiKeyCredential(apiKey));
• Choose a chat model and get a chat client:
  • var chat = client.GetChatClient("gpt-4o-mini");
• Maintain a List<ChatMessage> for history:
  • Start with a SystemChatMessage to set the assistant behavior.
  • Append each UserChatMessage and AssistantChatMessage after every turn.
• Stream responses:
  • await foreach (var update in chat.CompleteChatStreamingAsync(messages.ToArray())) { ... }
  • Print Text parts as they arrive and accumulate them into a buffer for history.

Changing the model

Update the line in Program.cs to a model you have access to:

var chat = client.GetChatClient("gpt-4o-mini");

See the OpenAI model catalog for available models and pricing.

Non‑streaming example (reference)

If you prefer to wait for the complete response (not used in this sample), the SDK supports a non‑streaming call:

var result = await chat.CompleteChatAsync(messages.ToArray());
var completion = result.Value; // ChatCompletion
var text = string.Join("\n",
    Enumerable.Range(0, completion.Content.Count)
              .Select(i => completion.Content[i].Text)
              .Where(t => !string.IsNullOrEmpty(t)));

Troubleshooting

• OPENAI_API_KEY not set: The app exits with an error message; set the environment variable and try again.
• 401/403 errors: Verify the API key and model access.
• Networking issues: Ensure outbound HTTPS access to the OpenAI API.
• Quotas/limits: Check your account usage and limits.

Security notes

• Do not commit secrets (API keys) to source control.
• Prefer user or machine‑level environment variables, or managed secret stores.

Tech stack

• .NET 9
• C# 13
• OpenAI NuGet package (2.6.x)

Scripts/commands summary

• Restore: dotnet restore
• Build: dotnet build
• Run: dotnet run --project OpenAI.Chat.SDK.csproj