Enabling the Plugin — Semantic Kernel Plugin

Package Reference

<PackageReference Include="BizFirst.Octopus.SemanticKernelPlugin" Version="*" />
<!-- Pulls in Microsoft.SemanticKernel transitively -->

Plugin Registration

builder.Services.AddOctopus(config =>
{
    config.AddPlugin<SqlServerPlugin>();        // Must be before SemanticKernelPlugin
    config.AddPlugin<SemanticKernelPlugin>();   // Requires IAgentStore from SqlServerPlugin
    config.AddPlugin<ChatbotUIPlugin>();
});

appsettings.json Configuration

{
  "SemanticKernelPlugin": {
    "DefaultProvider": "AzureOpenAI",

    "AzureOpenAI": {
      "ApiKeyCredentialId":    10,
      "Endpoint":              "https://myoai.openai.azure.com/",
      "ChatDeployment":        "gpt-4o",
      "EmbeddingDeployment":   "text-embedding-3-large"
    },

    "OpenAI": {
      "ApiKeyCredentialId":    11,
      "ChatModel":             "gpt-4o",
      "EmbeddingModel":        "text-embedding-3-large"
    },

    "Planner": {
      "Enabled":               true,
      "MaxSteps":              10,
      "AllowLoops":            false
    },

    "Memory": {
      "BridgeToOctopusStore":  true,
      "CollectionPrefix":      "sk_"
    }
  }
}

Config Properties

Property	Type	Description
`DefaultProvider`	string	`AzureOpenAI` or `OpenAI` — which provider the SK kernel uses by default
`AzureOpenAI.ApiKeyCredentialId`	int	Credential ID for the Azure OpenAI API key
`AzureOpenAI.Endpoint`	string	Azure OpenAI resource endpoint URL
`AzureOpenAI.ChatDeployment`	string	Deployment name for the chat completion model
`AzureOpenAI.EmbeddingDeployment`	string	Deployment name for the embedding model
`Planner.Enabled`	bool	Whether to activate the SK stepwise planner
`Planner.MaxSteps`	int	Maximum steps the planner may execute (safety cap)
`Memory.BridgeToOctopusStore`	bool	If true, SK memory calls go through `ISemanticMemoryStore`

OnStartAsync — Kernel Construction

public async Task OnStartAsync(IServiceProvider sp, CancellationToken ct)
{
    var config = sp.GetRequiredService<IOptions<SemanticKernelPluginConfig>>().Value;
    var creds  = sp.GetRequiredService<ICredentialResolver>();

    // Resolve API key at startup
    var apiKey = await creds.GetPasswordAsync(
        config.AzureOpenAI.ApiKeyCredentialId);

    // Build the SK kernel
    var kernelBuilder = Kernel.CreateBuilder();

    if (config.DefaultProvider == "AzureOpenAI")
    {
        kernelBuilder.AddAzureOpenAIChatCompletion(
            deploymentName: config.AzureOpenAI.ChatDeployment,
            endpoint:       config.AzureOpenAI.Endpoint,
            apiKey:         apiKey);

        kernelBuilder.AddAzureOpenAITextEmbeddingGeneration(
            deploymentName: config.AzureOpenAI.EmbeddingDeployment,
            endpoint:       config.AzureOpenAI.Endpoint,
            apiKey:         apiKey);
    }
    else
    {
        kernelBuilder.AddOpenAIChatCompletion(config.OpenAI.ChatModel, apiKey);
        kernelBuilder.AddOpenAITextEmbeddingGeneration(config.OpenAI.EmbeddingModel, apiKey);
    }

    var kernel = kernelBuilder.Build();

    // Register kernel as singleton for other components
    var kernelRegistry = sp.GetRequiredService<ISKKernelRegistry>();
    kernelRegistry.Register("default", kernel);
}

← Overview Next: LLM Routing via SK →