Nabs.Launchpad.Core.Ai.Clients 10.0.234

Nabs.Launchpad.Core.Ai.Clients

Nabs.Launchpad.Core.Ai.Clients gives you production-ready IAIClientFactory implementations and DI extensions so you can switch between Azure Foundry, local OpenAI-compatible endpoints, and deterministic simulators with minimal code changes.

Installation

dotnet add package Nabs.Launchpad.Core.Ai.Clients

The Problem

// 1) AI client wiring duplicated across apps
var openAiClient = new OpenAIClient(new ApiKeyCredential(key), new OpenAIClientOptions { Endpoint = new Uri(url) });
var chat = openAiClient.GetChatClient(model).AsIChatClient();

// 2) Environment switching is brittle and scattered
if (env.IsDevelopment())
{
	// local endpoint
}
else
{
	// azure endpoint
}

// 3) Tests require live services or heavy mocks
// No deterministic queue/keyed response simulator for chat streaming behavior.

// 4) Missing response setup causes runtime failures deep in execution flow
// InvalidOperationException thrown too late, hard to diagnose.

Features

• Factory-based AI client resolution through IAIClientFactory.
• Azure Foundry integration with AzureFoundryProjectClientFactory and DefaultAzureCredential.
• Local OpenAI-compatible integration with LocalOpenAIClientFactory.
• Deterministic simulation with AIChatClientSimulator for streaming and keyed responses.
• DI convenience extensions via AddAzureFoundryClientFactory, AddLocalOpenAIClientFactory, and AddSimulatorClientFactory.

Usage

using Microsoft.Extensions.DependencyInjection;
using Nabs.Launchpad.Core.Ai;
using Nabs.Launchpad.Core.Ai.Clients;

var services = new ServiceCollection();

services.AddAzureFoundryClientFactory("https://my-project.services.ai.azure.com/api/projects/my-project");

using var provider = services.BuildServiceProvider();
var factory = provider.GetRequiredService<IAIClientFactory>();
var chatClient = factory.GetChatClient("gpt-4.1-mini");

using Microsoft.Extensions.DependencyInjection;
using Nabs.Launchpad.Core.Ai;
using Nabs.Launchpad.Core.Ai.Clients;

var services = new ServiceCollection();

services.AddSimulatorClientFactory(sim =>
{
	sim.AddTextResponse("First simulated response")
	   .AddKeyedTextResponse("status", "All systems green");
});

using var provider = services.BuildServiceProvider();
var factory = provider.GetRequiredService<IAIClientFactory>();
var chatClient = factory.GetChatClient("ignored-by-simulator");

Configuration

• Azure Foundry: pass your AI project URL to AddAzureFoundryClientFactory(string aiProjectUrl).
• Local endpoint: pass your local OpenAI-compatible URL to AddLocalOpenAIClientFactory(string aiProjectUrl).
• Simulator: configure queued or keyed responses in AddSimulatorClientFactory(Action<AIChatClientSimulator>).

Project Structure

• ProjectExtensions - dependency injection extension methods for registering factory implementations.
• AzureFoundryProjectClientFactory - creates IChatClient instances from Azure AI Project OpenAI clients.
• LocalOpenAIClientFactory - creates local OpenAI-compatible IChatClient instances.
• SimulatorClientFactory - returns an AIChatClientSimulator implementation of IChatClient.
• AIChatClientSimulator - deterministic test client with queued and keyed streaming responses.

License

Copyright (c) Net Advantage Business Solutions.