Blazor DatAInsightsComposer Overview
The DatAInsightsComposer for Blazor lets users describe and create a data visualization in natural language. The component processes the user prompt through an AI service and renders other Telerik Blazor data components, such as Grid, Chart, or Gauge. The host app supplies dataset metadata and can customize how the component processes data.
The DatAInsightsComposer component abstracts the following processes:
The DatAInsightsComposer enables developers to plug into the above processes and customize them.
Installation
The Telerik DatAInsightsComposer component depends on a separate NuGet package Telerik.UI.for.Blazor.DatAInsightsComposer. Install the package alongside Telerik.UI.for.Blazor.
After package installation:
-
Register the composer service with
AddTelerikBlazorDatAInsightsComposer().Program.cs
C#// Telerik UI for Blazor builder.Services.AddTelerikBlazor(); // Telerik DatAInsightsComposer builder.Services.AddTelerikBlazorDatAInsightsComposer(); -
Import the
DatAInsightsComposernamespace in_Imports.razoror in specific.razorfiles._Imports.razor
RAZOR@* Telerik UI for Blazor *@ @using Telerik.Blazor @using Telerik.Blazor.Components @* Telerik DatAInsightsComposer *@ @using Telerik.Blazor.Components.DatAInsightsComposer
Creating Blazor DatAInsightsComposer
- Register an
IChatClientimplementation in the app dependency injection container, or provide a custom AI service transport. - Create one or more
DatasetDescriptorobjects with field metadata. Optionally, also populate the data items initially. - Set the
<TelerikDatAInsightsComposer>Datasetsparameter to anIEnumerable<DatasetDescriptor>.
Basic DatAInsightsComposer
<TelerikDatAInsightsComposer Datasets="@DAICDatasets"
@bind-Prompt="@DAICPrompt" />
@code {
private string DAICPrompt { get; set; } = "Show all data in tabular format";
private readonly static string[] ProductNames = new string[] { "Phones", "Tablets", "Laptops" };
private List<DatasetDescriptor> DAICDatasets { get; set; } = new()
{
new DatasetDescriptor()
{
Id = "sales-orders",
Label = "Sales Orders",
Description = "Sales orders by category, product, and country",
Fields = new List<DatasetFieldDescriptor>()
{
new() { Name = "Category", Type = DatasetFieldType.String },
new() { Name = "Product", Type = DatasetFieldType.String },
new() { Name = "Country", Type = DatasetFieldType.String },
new() { Name = "OrderDate", Label = "Order Date", Type = DatasetFieldType.Date },
new() { Name = "OrderValue", Label = "Order Value", Type = DatasetFieldType.Number }
}
}
};
protected override void OnInitialized()
{
int dataItemCount = 100;
int categoryCount = 3;
int productCount = 10 + 1;
int countryCount = 5;
Random rnd = Random.Shared;
List<DAICModel> data = new();
for (int i = 1; i <= dataItemCount; i++)
{
int productNumber = Random.Shared.Next(1, productCount);
int categoryNumber = productNumber % categoryCount + 1;
int countryNumber = i % countryCount + 1;
data.Add(new DAICModel()
{
Category = $"Category {categoryNumber}",
Product = $"Product {productNumber}",
Country = $"Country {countryNumber}",
OrderDate = DateOnly.FromDateTime(DateTime.Today.AddDays(-rnd.Next(1, 31)).AddMonths(-rnd.Next(1, 12)).AddYears(-rnd.Next(0, 5))),
OrderValue = rnd.Next(12345, 98765)
});
}
DAICDatasets.First(x => x.Id == "sales-orders").Data = data;
base.OnInitialized();
}
public class DAICModel
{
public string Category { get; set; } = string.Empty;
public string Product { get; set; } = string.Empty;
public string Country { get; set; } = string.Empty;
public DateOnly OrderDate { get; set; }
public decimal OrderValue { get; set; }
}
}
User Interface
By default, the initial UI consists of the following elements:
- Title
- Message box that shows AI-powered information or clarification questions below the title.
- PromptBox with two different placeholders that encourage the user to type their prompt and then refine it.
- Datasets panel inside the PromptBox that shows all datasets with their fields in a dropdown.
- Datasets list that shows all datasets as chips below the PromptBox.
- Generate button to submit the user prompt. While the AI response is in progress, the same button can abort the request.
You can customize the title, placeholders, and datasets information visibiliy through the respective DatAInsightsComposer parameters.
Once the user submits a prompt, the DatAInsightsComposer generates a UI component in a new panel below the PromptBox that includes:
- Title that is generated, based on the user prompt.
- The generated UI component (Grid, Chart, Gauge).
- Confirm and Discard buttons to accept or reject the generated component and its configuration.
- History button to show all prompts from the current session in a dropdown and restore the respective generated components.
You can customize the title and generated UI component settings in the DatAInsightsComposer OnGenerate event.
Datasets
The DatAInsightsComposer supports multiple datasets and the AI model selects the most appropriate one, depending on the user prompt and the dataset metadata. The AI model does not receive the data items from the datasets. Learn mode about how to define datasets and provide data.
Data Processing
The DatAInsightsComposer applies generated filtering, sorting, paging, grouping, and aggregations to data, based on the user prompts. The AI model suggests the data operations to perform, but has no access to the data itself. You can handle the DatAInsightsComposer OnDataProcess event to process data manually or load it on demand.
Visualization Components
The DatAInsightsComposer can generate a Grid, a Chart, or a Gauge. The app can configure component settings declaratively in the DatAInsightsComposer definition, or at runtime in the DatAInsightsComposer OnGenerate event.
AI Integration
The DatAInsightsComposer initiates AI service calls, including system prompt, context provision, and instructions about the expected response format.
By default, the DatAInsightsComposer uses the built-in DatAInsightsComposerAIService that implements the IDatAInsightsComposerAIService interface and relies on a registered Microsoft.Extensions.AI.IChatClient in the app. You can inherit and customize the DatAInsightsComposerAIService, or implement your own IDatAInsightsComposerAIService. In either case, register the new AI service after the default one:
Program.cs
// Default Telerik DatAInsightsComposer AI service
builder.Services.AddTelerikBlazorDatAInsightsComposer();
// Custom Telerik DatAInsightsComposer AI service
builder.Services.AddScoped<IDatAInsightsComposerAIService, MyDatAInsightsAIService>();
State
The DatAInsightsComposer allows the app to get or set the component state programmatically for greater flexibility.
Security
See the Security article to find out how the DatAInsightsComposer protects the application and data security, and what the app can or should contribute in this regard.
Events
Handle prompt, generation, data processing, history, confirm, and remove events for greater app flexibility. See the DatAInsightsComposer events article for details.
DatAInsightsComposer API
See the DatAInsightsComposer API Reference for the complete list of parameters, methods, and events.
DatAInsightsComposer Reference
The DatAInsightsComposer exposes methods for programmatic operation. To call them, define a reference with the @ref directive. Blazor populates component references in OnAfterRenderAsync, so the reference is not available earlier.
Define a reference to the DatAInsightsComposer
<TelerikDatAInsightsComposer @ref="@DatAInsightsComposerRef" />
@code {
private TelerikDatAInsightsComposer? DatAInsightsComposerRef;
}
Next Steps
- Provide Data to the DatAInsightsComposer
- Configure Generated Components
- Implement Best Security Practices
- Handle DatAInsightsComposer events