WorkflowCore Azure AI Foundry Sample

This sample demonstrates how to use the WorkflowCore.AI.AzureFoundry extension to build AI-powered, agentic workflows.

Features Demonstrated

  1. Simple Chat Completion - Conversational LLM chat with persistent conversation
  2. Agent with Tools - Autonomous agent that uses tools (weather, calculator) to answer questions
  3. Human-in-the-Loop Review - AI generates content, human approves/modifies before continuing

Prerequisites

Setup

  1. Copy the environment file: 
    cp .env.example .env
  2. Edit .env with your Azure AI credentials: 
    AZURE_AI_ENDPOINT=https://your-resource.services.ai.azure.com
AZURE_AI_API_KEY=your-api-key-here
AZURE_AI_DEFAULT_MODEL=gpt-4o

    Get your endpoint and API key from the Azure Portal:

    • Navigate to your Azure AI Foundry resource
    • Go to Keys and Endpoint
    • Copy the endpoint and one of the keys

Running the Sample

cd src/samples/WorkflowCore.Sample.AzureFoundry
dotnet run

You’ll see an interactive menu:

=== WorkflowCore Azure AI Foundry Sample ===

Choose a workflow to run:
1. Simple Chat Completion
2. Agent with Tools (Agentic Loop)
3. Human-in-the-Loop Review
Q. Quit

Enter choice:

Sample Workflows

1. Simple Chat Completion

A conversational chat loop where you can have a multi-turn conversation with the LLM.

Enter choice: 1
Type 'quit' to exit the conversation.

You: What is the capital of France?
Assistant: The capital of France is Paris.

You: What's the population?
Assistant: Paris has a population of approximately 2.1 million in the city proper...

You: quit

Workflow code:

builder
    .StartWith(context => ExecutionResult.Next())
    .ChatCompletion(cfg => cfg
        .SystemPrompt("You are a helpful assistant")
        .UserMessage(data => data.UserMessage)
        .OutputTo(data => data.Response));

2. Agent with Tools (Agentic Loop)

An autonomous agent that can use tools to accomplish tasks. The agent decides when and how to use tools based on your request.

Available tools:

Enter choice: 2
Available tools: weather (get weather for a city), calculator (do math)
Type 'quit' to exit the conversation.

You: What's the weather in Seattle?
Agent: The current weather in Seattle is partly cloudy with a temperature of 31°C (87°F) and a humidity of 84%.

You: What is 25 * 4 + 10?
Agent: 25 × 4 + 10 = 110

You: What's the weather in Tokyo and convert the temperature from Celsius to Fahrenheit
Agent: The weather in Tokyo is sunny with a temperature of 28°C. Converting to Fahrenheit: (28 × 9/5) + 32 = 82.4°F

You: quit

How it works:

  1. You send a request
  2. The LLM analyzes your request and decides which tool(s) to use
  3. Tools are executed automatically
  4. Results are fed back to the LLM
  5. The LLM provides a final response using the tool results

Workflow code:

builder
    .StartWith(context => ExecutionResult.Next())
    .AgentLoop(cfg => cfg
        .SystemPrompt(@"You are a helpful assistant with access to tools.
            Use the weather tool to get weather information.
            Use the calculator tool for math operations.
            Always explain what you're doing.")
        .Message(data => data.UserRequest)
        .WithTool("weather")
        .WithTool("calculator")
        .MaxIterations(5)
        .AutoExecuteTools(true)
        .OutputTo(data => data.AgentResponse));

3. Human-in-the-Loop Review

Demonstrates workflows that pause for human approval. The AI generates content, then waits for a human to approve, reject, or modify it.

Enter choice: 3
Enter content to generate and review: Write a product description for wireless earbuds

AI Generated Content:
[AI generates a product description]

Enter your review decision:
1. Approve as-is
2. Approve with modifications
3. Reject

Enter decision: 1
Content approved: [approved content is stored]

Workflow code:

builder
    .StartWith(context => ExecutionResult.Next())
    .ChatCompletion(cfg => cfg
        .SystemPrompt("You are a marketing copywriter")
        .UserMessage(data => $"Write about: {data.Topic}")
        .OutputTo(data => data.GeneratedContent))
    .HumanReview(cfg => cfg
        .Content(data => data.GeneratedContent)
        .Reviewer(data => data.Reviewer)
        .OnApproved(data => data.ApprovedContent));

Creating Custom Tools

You can extend the agent’s capabilities by creating custom tools:

public class StockPriceTool : IAgentTool
{
    public string Name => "stock_price";
    
    public string Description => "Get the current stock price for a ticker symbol";
    
    public string ParametersSchema => @"{
        ""type"": ""object"",
        ""properties"": {
            ""ticker"": { 
                ""type"": ""string"", 
                ""description"": ""Stock ticker symbol (e.g., MSFT, AAPL)"" 
            }
        },
        ""required"": [""ticker""]
    }";

    private readonly IStockService _stockService;

    public StockPriceTool(IStockService stockService)
    {
        _stockService = stockService;
    }

    public async Task<ToolResult> ExecuteAsync(
        string toolCallId, 
        string arguments, 
        CancellationToken ct)
    {
        var args = JsonSerializer.Deserialize<StockArgs>(arguments);
        var price = await _stockService.GetPriceAsync(args.Ticker, ct);
        
        return ToolResult.Succeeded(toolCallId, Name, 
            JsonSerializer.Serialize(new { ticker = args.Ticker, price = price }));
    }
}

Register your tool:

services.AddSingleton<StockPriceTool>();
toolRegistry.Register(serviceProvider.GetRequiredService<StockPriceTool>());

Project Structure

WorkflowCore.Sample.AzureFoundry/
├── Program.cs                      # Entry point and service configuration
├── README.md                       # This file
├── .env.example                    # Environment variable template
├── Workflows/
│   ├── WorkflowData.cs            # Data classes for all workflows
│   ├── SimpleChatWorkflow.cs      # Simple LLM chat workflow
│   ├── AgentWithToolsWorkflow.cs  # Agentic workflow with tool calling
│   └── HumanReviewWorkflow.cs     # Human-in-the-loop workflow
└── Tools/
    ├── WeatherTool.cs             # Simulated weather API tool
    └── CalculatorTool.cs          # Mathematical calculator tool

Troubleshooting

“Resource not found” error

Make sure your endpoint is correct:

Authentication errors

  1. Verify your API key is correct
  2. Make sure the key has access to the resource
  3. Check that the model is deployed and accessible

Tool not being called

The LLM decides when to use tools based on your request. Try being more specific:

Learn More