Website AI Assistant Azure Function

This is an Azure Function for a website AI assistant. It provides an endpoint for generating responses based on visitor's input.

The response can be a Prediction or your own custom response (eg. data from database or other source) based on the Prediction.

Packages	Version	Downloads
WebsiteAIAssistant
WebsiteAIAssistant.AzureFunction

Endpoint

Out of the box, the Function provides a single endpoint:

GET /ai/{input}

Response

The Function returns a Prediction by default.

But, you can also implement a Post Prediction Service, in which you can return any response you want.

Eg. you can return information about the predicted category from database or other source.

Just implement the IPostPredictionService interface and register it in the DI container as Scoped.

public interface IPostPredictionService
{
    Task<object> HandlePredictionAsync(HttpRequest request, ModelInput input, Prediction prediction);
}

For example, you can return a Response object with database results (for eg.) or just a string message.

Negative

Integration

You create your model and save it as a .zip file, and then just provide the path to load the model in the Function settings.

Add the Nuget package or a reference to the WebsiteAIAssistant.AzureFunction project in your ASP.NET Core application.

Then, in your Program.cs, add the following lines to register the Function:

//Website AI Assistant
//Optional: register a custom post-prediction service to handle the prediction results
builder.Services.AddScoped<IPostPredictionService, PostPredictionService>();
//Optional: register a custom logger to log the assistant's operations
builder.Services.AddSingleton<IWebsiteAIAssistantLogger, WebsiteAIAssistantLogger>();
builder.Services.AddWebsiteAIAssistant(settings =>
{
    // Path to load model
    string modelPath = Path.Combine(Environment.CurrentDirectory, "SampleWebsite-AI-Model.zip");
    settings.AIModelLoadFilePath = modelPath;

    settings.NegativeConfidenceThreshold = 0.70f;
    settings.NegativeLabel = -1f;
});

Tests

You can browse the Tests to see how to call the Function.

Sample

You can find a sample Function App using the library here.