FabrCore.Core 0.6.7

FabrCore

An Orleans-based framework for building distributed AI agent systems in .NET.

FabrCore provides the building blocks for creating, hosting, and connecting to AI agents that run as Orleans grains. Agents are durable, scalable, and communicate through a structured message-passing architecture.

Architecture

┌─────────────────────────────────────────────────┐
│                  Your Application                │
├──────────────────┬──────────────────────────────┤
│   FabrCore.Client    │         FabrCore.Host            │
│  ClientContext   │   AgentGrain (Orleans)       │
│  ChatDock (UI)   │   API Controllers            │
│                  │   WebSocket Middleware        │
├──────────────────┴──────────────────────────────┤
│                   FabrCore.Sdk                       │
│  FabrCoreAgentProxy · TaskWorkingAgent · ChatClient  │
├─────────────────────────────────────────────────┤
│                  FabrCore.Core                        │
│  Interfaces · Data Models · Grain Abstractions   │
└─────────────────────────────────────────────────┘

Packages

Quick Start

1. Install packages

dotnet add package FabrCore.Host

FabrCore.Host pulls in FabrCore.Sdk and FabrCore.Core transitively. For client applications, add FabrCore.Client instead.

2. Create an agent

using FabrCore.Sdk;
using FabrCore.Core;

[AgentAlias("my-assistant")]
public class MyAssistantAgent : FabrCoreAgentProxy
{
    public override async Task<AgentMessage> OnMessage(AgentMessage message)
    {
        // Create an AI-powered chat agent
        var (agent, session) = await Host.CreateChatClientAgent(
            modelName: "AzureProd",
            instructions: "You are a helpful assistant."
        );

        var response = await agent.SendAsync(session, message.Text);
        return message.ToReply(response);
    }
}

3. Configure the host

var builder = WebApplication.CreateBuilder(args);

// Configure FabrCore with Orleans
builder.AddFabrCoreServer();

var app = builder.Build();
app.UseFabrCoreServer();
app.Run();

4. Configure model access

Copy FabrCore.json.example to FabrCore.json in your project root and fill in your API keys:

{
  "ModelConfigurations": [
    {
      "Name": "AzureProd",
      "Provider": "Azure",
      "Uri": "https://your-resource.cognitiveservices.azure.com/",
      "Model": "gpt-4.1-mini",
      "ApiKeyAlias": "AZURE_KEY"
    }
  ],
  "ApiKeys": [
    {
      "Alias": "AZURE_KEY",
      "Value": "your-api-key-here"
    }
  ]
}

Note: FabrCore.json is gitignored by default to prevent accidental secret commits.

Configuration

Orleans Clustering

FabrCore supports multiple Orleans clustering providers out of the box:

• Azure Storage — Microsoft.Orleans.Clustering.AzureStorage
• SQL Server (ADO.NET) — Microsoft.Orleans.Clustering.AdoNet
• Localhost — for development

Configure clustering through standard Orleans configuration in your host's appsettings.json.

WebSocket API

FabrCore includes WebSocket middleware for real-time agent communication. See the WebSocket documentation for protocol details and usage.

Client Library

The FabrCore.Client package provides ClientContext for connecting to agents and a ChatDock Blazor component for building chat UIs. See the Client documentation for details.

Building from Source

dotnet build src/FabrCore.sln

License

Licensed under the Apache License, Version 2.0.

See NOTICE for attribution requirements.

Contributing

Contributions are welcome! Please open an issue or pull request on GitHub.