Radzen Blazor Components v12.1: AI Chat with Tool Calling, AIPrompt, Smart Paste, and More

Radzen Blazor Components v12.1 is out, and it is all about AI. The AIChat component can now do real work in your app: call your C# code, ask for permission before it changes anything, and answer with charts and grids instead of plain text. Three new components bring AI to prompts, text fields and forms, and the DataGrid, Scheduler and Chart get assistants of their own. All of it works with the model provider you already use.

If you're new here: Radzen Blazor is a free, open-source, MIT-licensed library of over 145 native Blazor components. Here is what's new in v12.1.

AIChat does things, not just talksLink to this section

Until now, a chat in your app could only answer from what the model already knew. In v12.1, RadzenAIChat can call your own C# methods. Hand it a few methods and it works out when to use them. Ask "Which customers are from Germany?" and the assistant queries your data, shows what it did as a small card, and answers with the real list.

Left: a Northwind assistant RadzenAIChat answering Which customers are from Germany? with a Reasoning block, a completed FindCustomers tool card and a table of German customers. Right: an Inventory assistant that called GetProduct and UpdateProductPrice to raise the price of Chai by 10%, next to a product grid showing Chai at $19.80

It asks before it actsLink to this section

Reading data is one thing, changing it is another. Mark a method as requiring approval and the chat stops before running it. "Raise the price of Chai by 10%" shows a card with exactly what is about to happen and Approve and Reject buttons. The user stays in control, and the assistant carries on once they decide.

Try the tool calling demo →

Try the approval demo →

Answers as real UILink to this section

An answer no longer has to be a paragraph. Ask for the 2023 sales by month and the chat can reply with a RadzenChart. Ask for a customer's orders and get a RadzenDataGrid. You decide how each result is rendered, using the same components as the rest of your app, so the answer is something users can sort, hover and click.

A Sales assistant RadzenAIChat answering Show me the 2023 sales by month with a RadzenChart column chart of revenue rendered inside the chat, followed by a table of the same figures

Try the generative UI demo →

Reasoning, sources, files and memoryLink to this section

A few more things AIChat picks up in this release:

  • Reasoning. Models that think before they answer show their reasoning in a collapsible block above the reply.
  • Sources. Answers can list where they came from, whether that is a web search, Perplexity, or your own search method.
  • Attachments. Users can drop in images, text files and PDFs and ask questions about them.
  • Stop and usage. A stop button interrupts a long answer, and the token count of each reply can be shown.
  • Memory. Conversations can be saved and restored, in memory out of the box or in your own database.

See all AIChat demos →

Three new AI componentsLink to this section

AIPromptLink to this section

Not every AI task needs a conversation. RadzenAIPrompt is a prompt box for one-shot jobs: summarize this, fix the grammar, turn it into bullet points. Point it at a piece of text, offer a few suggested prompts, and let users refine the result with one click: shorter, more formal, or translated. The output streams in as it is written, with copy and regenerate built in.

InlineAIPromptLink to this section

RadzenInlineAIPrompt puts that same assistant inside any text field. A small button opens it in a popup, and the result replaces the field's text, is added to it, or is thrown away. Select part of the text first and only the selection is rewritten.

Left: RadzenAIPrompt next to a Text to work on textarea, with suggested prompts and a streamed bullet point summary followed by Shorter, Longer, Formal and German commands. Right: a product form whose Description field was rewritten by RadzenInlineAIPrompt into a persuasive product description

Try the AIPrompt demo →

Try the InlineAIPrompt demo →

SmartPasteButtonLink to this section

Everyone has retyped an email signature into a form. RadzenSmartPasteButton ends that. Copy any text with the details in it, click Smart paste, and the whole form fills itself: names, emails, addresses, dates, amounts, even dropdown choices. It reads your form model to know which fields exist, and you can add a short description to any field to guide it.

A customer note from Maria Anders at Alfreds Futterkiste next to a form that RadzenSmartPasteButton filled in: company, contact, email, phone, address, city, country, customer since date, credit limit and segment

Try the SmartPasteButton demo →

AI in the DataGrid, Scheduler and ChartLink to this section

The demos have a new AI section with assistants for the components you already use:

  • DataGrid. Tell the grid what you want in plain English: "Show orders shipped to Germany sorted by freight." It filters, sorts and groups, and every step is visible and can be undone. There is also an AI column, semantic search, and AI highlighting.
  • Scheduler. "Book a dentist visit on Friday at 15:00 for an hour." The assistant proposes the appointment and adds it once you approve.
  • Chart. Ask a chart to summarize the trend, find the quarter that grew the most, or forecast the next one.

A RadzenDataGrid of orders filtered to Ship country Germany and sorted by Freight, next to a Grid assistant chat that called FilterColumn and SortByColumn after the user asked Show orders shipped to Germany sorted by freight

There are also AI assistants for the HtmlEditor and the MarkdownEditor.

Explore the AI demos →

Any model, your choiceLink to this section

Every AI component in v12.1 is built on Microsoft.Extensions.AI, the standard abstraction for AI in .NET. Configure a provider once and all of them use it: OpenAI, Azure OpenAI, Ollama, any OpenAI-compatible endpoint, or any other client that speaks the standard. Embeddings are supported too, which is what powers semantic search in the DataGrid demo.

One safeguard worth knowing: in Blazor WebAssembly apps, everything in the browser is visible to users, so the AI service now refuses an API key there and points you to a server-side proxy instead. Your key stays on the server, where it belongs.

Also in v12.1Link to this section

  • AIChat, AIPrompt and Chat render Markdown while the answer streams, not only after it ends.
  • Lists, code, block quotes and other Markdown elements are styled in RadzenMarkdown and the MarkdownEditor design view.
  • The demos have a new AI section in the navigation and a redesigned AI overview page.

The full list is in the changelog.

Upgrading from v12.0Link to this section

One change can affect existing Blazor WebAssembly apps.

Get started with v12.1Link to this section

That's Radzen Blazor Components v12.1: free, open source, and MIT-licensed.

Update your Radzen.Blazor NuGet package and try every AI component in the live demos. The complete list of changes is in the v12.1.0 release notes.

If this helps with your Blazor projects, drop us a star on GitHub, and tell us in the forums which AI feature you'd like us to build next.

Explore the AI components

Try Radzen Blazor Studio

View plans and pricing

© 2016-2026 Radzen Ltd. All Rights Reserved.
Designed and developed with ❤️ in Radzen Blazor Studio.
IP Geolocation by DB-IP

Select theme: