🎯 Planning & Design Patterns with Azure OpenAI (Responses API) (.NET)

📋 Learning Objectives

This notebook demonstrates enterprise-grade planning and design patterns for building intelligent agents using the Microsoft Agent Framework in .NET with Azure OpenAI (Responses API). You'll learn to create agents that can decompose complex problems, plan multi-step solutions, and execute sophisticated workflows with .NET's enterprise features.

⚙️ Prerequisites & Setup

Development Environment:

.NET 9.0 SDK or higher
Visual Studio 2022 or VS Code with C# extension
An Azure subscription with an Azure OpenAI resource and a model deployment
The Azure CLI — sign in with az login

Required Dependencies:

<PackageReference Include="Microsoft.Extensions.AI" Version="9.9.0" />
<PackageReference Include="Azure.AI.OpenAI" Version="2.1.0" />
<PackageReference Include="Azure.Identity" Version="1.13.1" />
<PackageReference Include="DotNetEnv" Version="3.1.1" />

Environment Configuration (.env file):

AZURE_OPENAI_ENDPOINT=https://<your-resource>.openai.azure.com
AZURE_OPENAI_DEPLOYMENT=gpt-4o-mini

Running the Code

This lesson includes a .NET Single File App implementation. To run it:

# Make the file executable (Linux/macOS)
chmod +x 07-dotnet-agent-framework.cs

# Run the application
./07-dotnet-agent-framework.cs

Or use the dotnet run command:

dotnet run 07-dotnet-agent-framework.cs

Code Implementation

The complete implementation is available in 07-dotnet-agent-framework.cs, which demonstrates:

Loading environment configuration with DotNetEnv
Configuring the Azure OpenAI client for the Responses API
Defining structured data models (Plan and TravelPlan) with JSON serialization
Creating an AI agent with structured output using JSON schema
Executing planning requests with type-safe responses

Key Concepts

Structured Planning with Type-Safe Models

The agent uses C# classes to define the structure of planning outputs:

public class Plan
{
    [JsonPropertyName("assigned_agent")]
    public string? Assigned_agent { get; set; }

    [JsonPropertyName("task_details")]
    public string? Task_details { get; set; }
}

public class TravelPlan
{
    [JsonPropertyName("main_task")]
    public string? Main_task { get; set; }

    [JsonPropertyName("subtasks")]
    public IList<Plan> Subtasks { get; set; }
}

JSON Schema for Structured Outputs

The agent is configured to return responses matching the TravelPlan schema:

ChatClientAgentOptions agentOptions = new(name: AGENT_NAME, instructions: AGENT_INSTRUCTIONS)
{
    ChatOptions = new()
    {
        ResponseFormat = ChatResponseFormatJson.ForJsonSchema(
            schema: AIJsonUtilities.CreateJsonSchema(typeof(TravelPlan)),
            schemaName: "TravelPlan",
            schemaDescription: "Travel Plan with main_task and subtasks")
    }
};

Planning Agent Instructions

The agent acts as a coordinator, delegating tasks to specialized sub-agents:

FlightBooking: For booking flights and providing flight information
HotelBooking: For booking hotels and providing hotel information
CarRental: For booking cars and providing car rental information
ActivitiesBooking: For booking activities and providing activity information
DestinationInfo: For providing information about destinations
DefaultAgent: For handling general requests

Expected Output

When you run the agent with a travel planning request, it will analyze the request and generate a structured plan with appropriate task assignments to specialized agents, formatted as JSON conforming to the TravelPlan schema.

3.6 KiB Raw Permalink Blame History