╔══════════════════════════════════════════════════════════════════════════════╗ ║ ║ ║ 🤖 AI AGENTS DEMO - PROGRESSIVE AI INTEGRATION IN .NET ║ ║ ║ ║ Presented at: Cyber RGV - October 8, 2025 ║ ║ By: Josh Morales (SelAromDotNet) ║ ║ ║ ║ 📚 Complete Presentation Resources: ║ ║ https://link.content360.io/dotnet2025oct ║ ║ ║ ║ 💬 Join Rio Grande Valley .NET User Group: ║ ║ https://rgvdevs.net ║ ║ Collaborate, learn, and join more sessions like this! ║ ║ ║ ╚══════════════════════════════════════════════════════════════════════════════╝

# AI Agents Demo - Progressive AI Integration in .NET

A comprehensive demonstration showcasing the evolution of AI integration patterns in .NET, from raw HTTP API calls to sophisticated stateful agents.

## 🎯 What You'll Learn

This repository demonstrates five progressive approaches to integrating AI into .NET applications:

1. **Demo 1: Direct API Calls** — The "Hard Way" - Raw HTTP requests to AI providers
2. **Demo 2: Semantic Kernel** — Provider abstraction and prompt orchestration
3. **Demo 3: Extensions.AI with DI** — Configuration-driven provider switching
4. **Demo 4: AutoGen** — Tool-enabled collaboration with function calling
5. **Demo 5: Agent Framework** — Stateful agents with persistent memory and tools

## 🚀 Quick Start

### Prerequisites
- .NET 8.0 SDK or later
- OpenAI API Key
- (Optional) Google Gemini API Key (used in Demo 1, simulated in Demo 3)

### Setup

1. Clone the repository
   ```bash
   git clone https://github.com/selaromdotnet/dotnet-ai-demo.git
   cd AIAgentsDemo

2. Configure API keys in each demo project

   # Copy the example config and edit it (do not check into source control!)
   cp appsettings.example.json appsettings.json
   # Windows: copy appsettings.example.json appsettings.json

3. Run any demo

   dotnet run --project src/Demo1.DirectAPI
   dotnet run --project src/Demo2.SemanticKernel
   dotnet run --project src/Demo3.ExtensionsAI
   dotnet run --project src/Demo4.AutoGen
   dotnet run --project src/Demo5.AgentFramework

📚 Demo Overview

Demo 1: Direct API Calls (The "Hard Way")

Location: src/Demo1.DirectAPI/

• Raw HTTP requests to OpenAI and Gemini using HttpClient
• Provider-specific JSON structures and authentication
• Pain Point: Each provider requires custom code
• Limitation: Doesn't scale—manual work for every new provider

See Demo 1 README

---

Demo 2: Semantic Kernel (The "Smart Way")

Location: src/Demo2.SemanticKernel/

• Uses Microsoft Semantic Kernel for provider abstraction
• Creates reusable prompt functions and orchestration
• Two-step AI workflow: analysis → response generation
• Limitation: Provider choice configured in code (requires rebuild)

See Demo 2 README

---

Demo 3: Configuration + DI (The "Enterprise Way")

Location: src/Demo3.ExtensionsAI/

• Dependency Injection with Microsoft.Extensions.AI
• Configuration-driven provider selection via appsettings.json
• Runtime provider switching without code changes
• A/B testing, cost optimization, disaster recovery patterns
• Limitation: AI hallucinates data it cannot access

See Demo 3 README

---

Demo 4: AutoGen (The "Collaborative Way")

Location: src/Demo4.AutoGen/

• Agent with C# tool/function calling via AutoGen
• Real data access eliminates hallucinations
• Function middleware for tool registration
• Limitation: Stateless by default—no conversation memory

See Demo 4 README

---

Demo 5: Agent Framework (The "Stateful Way")

Location: src/Demo5.AgentFramework/

• Stateful agents with AgentThread for persistent memory
• Multi-turn conversations with context retention
• Auto tool invocation with dependency injection
• Thread serialization for conversation persistence
• Strongly-typed function returns for reliable AI parsing
• Achievement: Production-ready pattern addressing all prior limitations

See Demo 5 README

🎨 Features

• ✨ Colored console output for better visualization
• 🔧 Standalone projects—run any demo independently
• ⚙️ Consistent appsettings.json schema across demos
• 📖 Educational narratives highlighting tradeoffs and limitations
• 🚀 Progressive complexity from basics to production patterns

📦 Package Dependencies

• Microsoft.SemanticKernel 1.65.0 (Demos 2, 3, 5)
• Microsoft.SemanticKernel.Connectors.OpenAI 1.65.0 (Demos 2, 3, 5)
• Microsoft.Extensions.Configuration 8.0.x (All demos)
• Microsoft.Extensions.Hosting 8.0.1 (Demo 3)
• Microsoft.Extensions.AI 9.0.0-preview.9.24507.7 (Demo 3)
• Microsoft.Extensions.AI.OpenAI 9.0.0-preview.9.24507.7 (Demo 3)
• AutoGen 0.2.3 (Demo 4)
• AutoGen.OpenAI 0.2.3 (Demo 4)
• Microsoft.Agents.AI (Demo 5)

🤝 Contributing

Contributions welcome! Please open a Pull Request with your improvements.

📄 License

MIT License

⭐ Support

If you find this helpful, please give it a star on GitHub!

---

Happy Learning! 🎉

From the Rio Grande Valley .NET Community