CoreTemplate.AI 1.3.0

Core.AI

Modular and extensible AI integration layer for .NET applications. Supports multiple AI providers like OpenRouter and Ollama via pluggable architecture and Semantic Kernel.

Version 1.3.0: Added Semantic Kernel Agent service and full provider configurability.

Features

• Plug-and-play architecture
• Provider support: OpenRouter, Ollama
• Semantic Kernel-based agent support
• Dynamic model selection and validation
• Streaming & prompt support
• Memory support (user-based history)
• Easily extendable with your own providers

Installation

dotnet add package Core.AI --version 1.3.0

Configuration

Add the following to your appsettings.json:

"AiSettings": {
  "Provider": "OpenRouter",
  "Model": "mistralai/mistral-7b-instruct"
},

"OpenRouterAI": {
  "ApiKey": "sk-xxx"
},
"AgentProfiles": [
  {
    "Name": "Default",
    "Context": "You are a helpful assistant.",
    "Temperature": 0.7,
    "Model": "gpt-3.5-turbo"
  },
  {
    "Name": "CodeHelper",
    "Context": "You are an expert .NET developer. Answer code questions precisely.",
    "Temperature": 0.5,
    "Model": "gpt-4"
  },
  {
    "Name": "TravelGuide",
    "Context": "You are a friendly travel guide helping users plan trips.",
    "Temperature": 0.9,
    "Model": "gpt-3.5-turbo"
  }
],

// --- AI Services ---
builder.Services.AddOptions<AISettings>()
    .Bind(builder.Configuration.GetSection("AiSettings"))
    .Validate(settings => Enum.IsDefined(typeof(AIProvider), settings.Provider),
        "Invalid AI provider configured in AiSettings.Provider");

builder.Services.AddSingleton(sp =>
    sp.GetRequiredService<IOptions<AISettings>>().Value);

// Providers
builder.Services.AddScoped<OpenRouterAiService>();
builder.Services.AddScoped<OllamaAiService>();
builder.Services.AddScoped<IAIService, AIServiceResolver>();

// Model Providers
builder.Services.AddScoped<OpenRouterModelProvider>();
builder.Services.AddScoped<OllamaModelProvider>();
builder.Services.AddScoped<AIModelProviderResolver>();

// Agent Service
builder.Services.AddScoped<IAgentService, SemanticKernelAgentService>();
builder.Services.AddSingleton<ChatHistoryStore>();
builder.Services.AddSingleton<AgentProfileProvider>();

Memory Support

User messages are stored in-memory with ChatHistoryStore, enabling contextful chat.

Usage Example

using Azure;
using Core.AI.Abstractions;
using Core.AI.Models;
using Microsoft.AspNetCore.Mvc;
using System.Text;

namespace CoreApp.WebAPI.Controllers
{
    [Route("api/agent")]
    [ApiController]
    public class AgentController : ControllerBase
    {
        private readonly IAgentService _agentService;

        public AgentController(IAgentService agentService)
        {
            _agentService = agentService;
        }

        // POST /api/agent/prompt
        [HttpPost("prompt")]
        public async Task<IActionResult> Prompt([FromBody] AgentPromptRequest request)
        {
            var result = await _agentService.ChatAsync(request.Prompt, request.Options, request.UserId);
            return Ok(new { result });
        }

        // POST /api/agent/stream
        [HttpPost("stream")]
        public async Task StreamPrompt([FromBody] AgentPromptRequest request)
        {
            Response.ContentType = "text/plain";

            await foreach (var chunk in _agentService.StreamChatAsync(request.Prompt, request.Options, request.UserId))
            {
                var buffer = Encoding.UTF8.GetBytes(chunk);
                await Response.Body.WriteAsync(buffer);
                await Response.Body.FlushAsync();
            }
        }
    }
}

POST /api/agent/prompt Content-Type: application/json

{
  "prompt": "What's an interesting fact about the universe?",
  "options": {
    "context": "You are a helpful astrophysicist AI.",
    "temperature": 0.7,
    "model": "mistralai/mistral-7b-instruct",
    "provider": "OpenRouter",
    "profile": "Default"
  },
  "userId": "user-abc-123"
}

Field Descriptions