Fantastic Agent

A lightweight C# library for parallel LLM orchestration with native tool calling support. Run multiple providers simultaneously and compare their responses in real-time.

Why Fantastic Agent?

• Native tool calling - Simple reflection-based function registration that works across providers
• Event-driven streaming - Real-time response chunks from all providers
• Multi-provider support - Swap between Ollama, Claude, OpenAI without rewriting code
• Unopinionated architecture - Build agents your way without framework lock-in

Quick Start

// Create a thread with Ollama
OllamaThread ollama = new OllamaThread(
    "localhost", 
    11434, 
    OllamaModel, 
    "You are a helpful assistant."
);

// Register function tools
ollama.DeclareFunctionTool(typeof(WebSearchProviders).GetMethod("BraveSearch"));

// Subscribe to events
ollama.AssistantReplyChunkReceived += (s, e) => Console.Write(e.Message);

// Send a message and get streaming response
await ollama.SendMessageAsync("Search for the latest AI news");

Demo

The WPF demo shows three LLM providers (Ollama, Claude, ChatGPT) responding to the same Arabic query in parallel, each making tool calls to BraveSearch and streaming results independently.

Current Status

• Native LLM support with tool calling
• Multi-provider parallel execution
• Event-driven streaming
• Usage Monitoring
• NuGet package at https://www.nuget.org/packages/fantasticagent/
• Full agent orchestration (figuring it out)

Supported Providers

• Ollama
• Claude (Anthropic)
• OpenAI/ChatGPT
• Gemini
• HugginFace, and Groq (as a direct inheritance from OpenAI responses api)

License

[MIT]