Skip to content

Commit

Permalink
.Net: Concept sample showing different options for calling functions …
Browse files Browse the repository at this point in the history
…with multiple parameters (#8653)

### Motivation and Context

Concept sample showing different options for calling functions with
multiple parameters

### Description

<!-- Describe your changes, the overall approach, the underlying design.
These notes will help understanding how your code works. Thanks! -->

### Contribution Checklist

<!-- Before submitting this PR, please make sure: -->

- [ ] The code builds clean without any errors or warnings
- [ ] The PR follows the [SK Contribution
Guidelines](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md)
and the [pre-submission formatting
script](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md#development-scripts)
raises no violations
- [ ] All unit tests pass, and I have added new tests where possible
- [ ] I didn't break anyone 😄
  • Loading branch information
markwallace-microsoft authored Sep 11, 2024
1 parent e44817d commit 5a3eda5
Show file tree
Hide file tree
Showing 3 changed files with 216 additions and 1 deletion.
3 changes: 2 additions & 1 deletion .github/_typos.toml
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,8 @@ extend-exclude = [
"test_code_tokenizer.py",
"*response.json",
"test_content.txt",
"serializedChatHistoryV1_15_1.json"
"serializedChatHistoryV1_15_1.json",
"MultipleFunctionsVsParameters.cs"
]

[default.extend-words]
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,213 @@
// Copyright (c) Microsoft. All rights reserved.

using System.ComponentModel;
using System.Text.Json;
using Microsoft.Extensions.DependencyInjection;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Connectors.OpenAI;

namespace FunctionCalling;

/// <summary>
/// This sample shows different options for calling functions with multiple parameters.
/// The scenario is to search for invoices by customer name, purchase order, or vendor number.
///
/// The first sample uses multiple functions, one for each search criteria. One issue is that
/// as the number of functions increases then the reliability of the AI model to select the correct
/// function may decrease. To help avoid this issue, you can try filtering which functions are advertised
/// to the AI model e.g. if your application has come context information which indicates a purchase order
/// is available then you can filter out the customer name and vendor number functions.
///
/// The second sample uses a single function that takes an object with all search criteria. In this case some
/// of the search criteria are optional. Again as the number of parameters increases then the reliability of the
/// AI model may decrease. One advantage of this approach is that if the AI model can extra multiple search criteria
/// for the users ask then your plugin can use this information to provide more reliable results.
///
/// For both options care should be taken to validate the parameters that the AI model provides. E.g. the customer
/// name could be wrong or the purchase order could be invalid. It is worth catching these errors and responding the
/// AI model with a message that explains what has gone wrong to see how it responds. It may be able to retry the search
/// and get a successful response on the second attempt. Or it may decide to revert pack to the human in the loop to ask
/// for more information.
/// </summary>
public class MultipleFunctionsVsParameters(ITestOutputHelper output) : BaseTest(output)
{
/// <summary>
/// Shows how to use multiple Search By functions to search for invoices by customer name, purchase order, or vendor number.
/// </summary>
[Fact]
public async Task InvoiceSearchBySampleAsync()
{
// Create a kernel with OpenAI chat completion
IKernelBuilder kernelBuilder = Kernel.CreateBuilder();
kernelBuilder.Services.AddSingleton<IAutoFunctionInvocationFilter>(
new AutoFunctionInvocationFilter(this.Output));
kernelBuilder.AddOpenAIChatCompletion(
modelId: TestConfiguration.OpenAI.ChatModelId,
apiKey: TestConfiguration.OpenAI.ApiKey);
kernelBuilder.Plugins.AddFromType<InvoiceSearchBy>();
Kernel kernel = kernelBuilder.Build();

await InvokePromptsAsync(kernel);
}

/// <summary>
/// Shows how to use a single Search function to search for invoices by customer name, purchase order, or vendor number.
/// </summary>
[Fact]
public async Task InvoiceSearchSampleAsync()
{
// Create a kernel with OpenAI chat completion
IKernelBuilder kernelBuilder = Kernel.CreateBuilder();
kernelBuilder.Services.AddSingleton<IAutoFunctionInvocationFilter>(
new AutoFunctionInvocationFilter(this.Output));
kernelBuilder.AddOpenAIChatCompletion(
modelId: TestConfiguration.OpenAI.ChatModelId,
apiKey: TestConfiguration.OpenAI.ApiKey);
kernelBuilder.Plugins.AddFromType<InvoiceSearch>();
Kernel kernel = kernelBuilder.Build();

await InvokePromptsAsync(kernel);
}

/// <summary>Invoke the various prompts we want to test.</summary>
private async Task InvokePromptsAsync(Kernel kernel)
{
OpenAIPromptExecutionSettings settings = new() { ToolCallBehavior = ToolCallBehavior.AutoInvokeKernelFunctions };
Console.WriteLine("Prompt: Show me the invoices for customer named Contoso Industries.");
Console.WriteLine(await kernel.InvokePromptAsync("Show me the invoices for customer named Contoso Industries.", new(settings)));
Console.WriteLine("----------------------------------------------------");
Console.WriteLine("Prompt: Show me the invoices for purchase order PO123.");
Console.WriteLine(await kernel.InvokePromptAsync("Show me the invoices for purchase order PO123.", new(settings)));
Console.WriteLine("----------------------------------------------------");
Console.WriteLine("Prompt: Show me the invoices for vendor number VN123.");
Console.WriteLine(await kernel.InvokePromptAsync("Show me the invoices for vendor number VN123.", new(settings)));
Console.WriteLine("----------------------------------------------------");
Console.WriteLine("Prompt: Show me the invoices for Contoso Industries.");
Console.WriteLine(await kernel.InvokePromptAsync("Show me the invoices for Contoso Industries.", new(settings)));
Console.WriteLine("----------------------------------------------------");
Console.WriteLine("Prompt: Show me the invoices for PO123.");
Console.WriteLine(await kernel.InvokePromptAsync("Show me the invoices for PO123.", new(settings)));
Console.WriteLine("----------------------------------------------------");
Console.WriteLine("Prompt: Show me the invoices for VN123.");
Console.WriteLine(await kernel.InvokePromptAsync("Show me the invoices for VN123.", new(settings)));
Console.WriteLine("----------------------------------------------------");
Console.WriteLine("Prompt: Zeigen Sie mir die Rechnungen für Contoso Industries.");
Console.WriteLine(await kernel.InvokePromptAsync("Zeigen Sie mir die Rechnungen für Contoso Industries.", new(settings)));
Console.WriteLine("----------------------------------------------------");
}

/// <summary>Shows available syntax for auto function invocation filter.</summary>
private sealed class AutoFunctionInvocationFilter(ITestOutputHelper output) : IAutoFunctionInvocationFilter
{
public async Task OnAutoFunctionInvocationAsync(AutoFunctionInvocationContext context, Func<AutoFunctionInvocationContext, Task> next)
{
var functionName = context.Function.Name;
var arguments = context.Arguments;

// Output the details of the function being called
output.WriteLine($"Function: {functionName} {JsonSerializer.Serialize(arguments)}");

// Calling next filter in pipeline or function itself.
await next(context);
}
}

/// <summary>
/// A plugin that provides methods to search for Invoices using different criteria.
/// </summary>
private sealed class InvoiceSearchBy
{
[KernelFunction]
[Description("Search for invoices by customer name.")]
public IEnumerable<Invoice> SearchByCustomerName([Description("The customer name.")] string customerName)
{
return
[
new Invoice { CustomerName = customerName, PurchaseOrder = "PO123", VendorNumber = "VN123" },
new Invoice { CustomerName = customerName, PurchaseOrder = "PO124", VendorNumber = "VN124" },
new Invoice { CustomerName = customerName, PurchaseOrder = "PO125", VendorNumber = "VN125" },
];
}

[KernelFunction]
[Description("Search for invoices by purchase order.")]
public IEnumerable<Invoice> SearchByPurchaseOrder([Description("The purchase order. Purchase orders begin with a PN prefix.")] string purchaseOrder)
{
return
[
new Invoice { CustomerName = "Customer1", PurchaseOrder = purchaseOrder, VendorNumber = "VN123" },
new Invoice { CustomerName = "Customer2", PurchaseOrder = purchaseOrder, VendorNumber = "VN124" },
new Invoice { CustomerName = "Customer3", PurchaseOrder = purchaseOrder, VendorNumber = "VN125" },
];
}

[KernelFunction]
[Description("Search for invoices by vendor number")]
public IEnumerable<Invoice> SearchByVendorNumber([Description("The vendor number. Vendor numbers begin with a VN prefix.")] string vendorNumber)
{
return
[
new Invoice { CustomerName = "Customer1", PurchaseOrder = "PO123", VendorNumber = vendorNumber },
new Invoice { CustomerName = "Customer2", PurchaseOrder = "PO124", VendorNumber = vendorNumber },
new Invoice { CustomerName = "Customer3", PurchaseOrder = "PO125", VendorNumber = vendorNumber },
];
}
}

/// <summary>
/// A plugin that provides methods to search for Invoices using different criteria.
/// </summary>
private sealed class InvoiceSearch
{
[KernelFunction]
[Description("Search for invoices by customer name or purchase order or vendor number.")]
public IEnumerable<Invoice> Search([Description("The invoice search request. It must contain either a customer name or a purchase order or a vendor number")] InvoiceSearchRequest searchRequest)
{
return
[
new Invoice
{
CustomerName = searchRequest.CustomerName ?? "Customer1",
PurchaseOrder = searchRequest.PurchaseOrder ?? "PO123",
VendorNumber = searchRequest.VendorNumber ?? "VN123"
},
new Invoice
{
CustomerName = searchRequest.CustomerName ?? "Customer2",
PurchaseOrder = searchRequest.PurchaseOrder ?? "PO124",
VendorNumber = searchRequest.VendorNumber ?? "VN124"
},
new Invoice
{
CustomerName = searchRequest.CustomerName ?? "Customer3",
PurchaseOrder = searchRequest.PurchaseOrder ?? "PO125",
VendorNumber = searchRequest.VendorNumber ?? "VN125"
},
];
}
}

/// <summary>
/// Represents an invoice.
/// </summary>
private sealed class Invoice
{
public string CustomerName { get; set; }
public string PurchaseOrder { get; set; }
public string VendorNumber { get; set; }
}

/// <summary>
/// Represents an invoice search request.
/// </summary>
[Description("The invoice search request.")]
private sealed class InvoiceSearchRequest
{
[Description("Optional, customer name.")]
public string? CustomerName { get; set; }
[Description("Optional, purchase order. Purchase orders begin with a PN prefix.")]
public string? PurchaseOrder { get; set; }
[Description("Optional, vendor number. Vendor numbers begin with a VN prefix.")]
public string? VendorNumber { get; set; }
}
}
1 change: 1 addition & 0 deletions dotnet/samples/Concepts/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,7 @@ dotnet test -l "console;verbosity=detailed" --filter "FullyQualifiedName=ChatCom
- [Gemini_FunctionCalling](https://github.com/microsoft/semantic-kernel/blob/main/dotnet/samples/Concepts/FunctionCalling/Gemini_FunctionCalling.cs)
- [OpenAI_FunctionCalling](https://github.com/microsoft/semantic-kernel/blob/main/dotnet/samples/Concepts/FunctionCalling/OpenAI_FunctionCalling.cs)
- [NexusRaven_HuggingFaceTextGeneration](https://github.com/microsoft/semantic-kernel/blob/main/dotnet/samples/Concepts/FunctionCalling/NexusRaven_FunctionCalling.cs)
- [MultipleFunctionsVsParameters](https://github.com/microsoft/semantic-kernel/blob/main/dotnet/samples/Concepts/FunctionCalling/MultipleFunctionsVsParameters.cs)

### Caching - Examples of caching implementations

Expand Down

0 comments on commit 5a3eda5

Please sign in to comment.