Table of Contents
1. What Is Ant Design X?
Ant Design X (@ant-design/x) is a React component library for AI-driven interfaces launched by the Ant Design team. It is not a replacement for Ant Design 5, but rather a new component system built on top of the existing design system, specifically designed for scenarios such as AI chat, intelligent assistants, and Agent consoles.
On April 30, 2026, Ant Design X released version v2.7.0, and its ecosystem now encompasses 5 core packages:
| Package Name | Purpose | Latest Version |
|---|---|---|
@ant-design/x | Core AI interface component library | 2.7.0 |
@ant-design/x-markdown | Streaming Markdown renderer | 2.7.0 |
@ant-design/x-sdk | AI data stream management SDK | 2.7.0 |
@ant-design/x-card | A2UI dynamic card rendering | 2.7.0 |
@ant-design/x-skill | AI coding assistant skill library | 2.7.0 |
2. Design Philosophy: The RICH Interaction Paradigm
The core design philosophy of Ant Design X is the RICH paradigm, which breaks down AI interaction into four stages:
RICH interaction flowβββ R β Role (role intent)β βββ Welcome welcome componentβ βββ Prompts prompt setβββ I β Intention (intent awakening)β βββ Sender input boxβ βββ Attachment attachmentβ βββ Suggestion quick suggestionβββ C β Conversation (conversation communication)β βββ Bubble conversation bubbleβ βββ Conversations conversation listβ βββ Notification notificationβββ H β Hybrid UI (hybrid interface) βββ Think thinking process βββ ThoughtChain thought chain βββ Actions action bar βββ FileCard file card βββ Sources cited sources βββ Mermaid diagram rendering βββ Folder file tree βββ CodeHighlighter code highlightingRICH is not a linear process, but rather composable interaction units. You can use components from a single stage independently, or combine them into a complete AI application.
3. Hybrid UI: Merging GUI with Natural Language
The Hybrid UI (HUI) concept proposed by Ant Design X attempts to resolve a core contradiction: traditional GUI operations are efficient but have a high learning curve, while natural language interaction is flexible but expresses intent ambiguously.
HUI divides interaction modes into three types:
| Mode | Characteristics | Use Cases |
|---|---|---|
| Do-heavy | GUI-dominant, AI-assisted | Structured forms, data entry |
| Chat-heavy | Natural language-dominant | Open-ended Q&A, content generation |
| HUI Hybrid | Free switching between GUI + Chat | Intelligent customer service, Agent consoles |
This design allows the same interface to perform precise operations through traditional forms while also expressing flexibly through natural language β βhaving your cake and eating it too.β
4. Core Components at a Glance
4.1 Bubble β Conversation Bubble
Bubble is the foundational component for displaying conversation content, supporting multiple roles and content types:
import { Bubble } from '@ant-design/x';
<Bubble placement="start" avatar={{ icon: <RobotOutlined /> }} content="Hello, I'm an AI assistant"/>4.2 Sender β Input Box
Sender is an enhanced input component that supports file attachments, voice input, and quick suggestions:
import { Sender } from '@ant-design/x';
<Sender onSubmit={(value) => console.log(value)} placeholder="Type your question..."/>4.3 ThoughtChain β Chain of Thought
Displays the thinking process of an AI Agent, allowing users to understand the AIβs reasoning path:
import { ThoughtChain } from '@ant-design/x';
<ThoughtChain items={[ { title: 'Analyze the question', description: 'Identify keywords...', status: 'success' }, { title: 'Search the knowledge base', description: 'Matched 3 related documents', status: 'success' }, { title: 'Generate answer', status: 'pending' }, ]}/>4.4 Sources β Cited Sources
Displays the reference sources for AI responses, enhancing credibility:
import { Sources } from '@ant-design/x';
<Sources items={[ { title: 'React official documentation', url: 'https://react.dev' }, { title: 'Ant Design X documentation', url: 'https://x.ant.design' }, ]}/>5. XMarkdown: Streaming Markdown Rendering
@ant-design/x-markdown is a Markdown rendering engine optimized specifically for AI streaming output. It solves three core problems that traditional Markdown renderers face in streaming scenarios:
5.1 Incomplete Syntax Fault Tolerance
When AI streams output $$E = mc^ (unclosed formula), XMarkdown does not crash, but instead gracefully degrades the rendering:
import { XMarkdown } from '@ant-design/x-markdown';
<XMarkdown content={streamingText} streaming={{ tail: 'β' }} //Streaming cursor indicator/>5.2 Custom Streaming Tail Indicator
The streaming.tail configuration added in v2.6.0 supports a custom blinking cursor:
<XMarkdown content={streamingText} streaming={{ tail: 'β', tailComponent: <BlinkingCursor /> }}/>5.3 Safe Rendering
The escapeRawHtml property added in v2.3.0 prevents XSS attacks:
<XMarkdown content={untrustedContent} escapeRawHtml={true}/>6. X SDK: AI Data Stream Management
@ant-design/x-sdk encapsulates the complete data flow for interacting with AI models:
6.1 XRequest β Request Management
import { XRequest } from '@ant-design/x-sdk';
const request = XRequest({ baseURL: 'https://api.openai.com/v1', model: 'gpt-4o', apiKey: process.env.OPENAI_API_KEY,});v2.2.0 supports reconnection and custom stream delimiters:
XRequest({ baseURL: 'https://api.example.com', model: 'custom-model', streamSeparator: '\n', //Stream delimiter partSeparator: 'data: ', //Part delimiter kvSeparator: ': ', //Key-value delimiter});6.2 useXChat β Conversation Management
import { useXChat } from '@ant-design/x-sdk';
const { messages, onRequest } = useXChat({ request, queueRequest: true, //New in v2.3.0: request queue support});7. X Card: A2UI Dynamic Cards
The @ant-design/x-card released in v2.5.0 is the most eye-catching new member of the entire ecosystem. It is based on the A2UI protocol, allowing AI Agents to dynamically generate structured UI:
import { XCard } from '@ant-design/x-card';
<XCard //A2UI JSON description returned by AI a2ui={agentResponse} dataModel={userData}/>The core value of A2UI lies in this: AI doesnβt just return text β it returns interactive cards. For example, AI can return a complete interactive interface containing forms, charts, and buttons, rather than just a text description.
8. Quick Start
8.1 Installation
#Core component librarynpm install @ant-design/x
#Streaming Markdown (optional)npm install @ant-design/x-markdown
#AI SDK (optional)npm install @ant-design/x-sdk8.2 A Complete AI Chat Application
import { Bubble, Conversations, Sender, XProvider,} from '@ant-design/x';import { XMarkdown } from '@ant-design/x-markdown';import { useXChat, XRequest } from '@ant-design/x-sdk';
export default function ChatApp() { const request = XRequest({ baseURL: 'https://api.openai.com/v1', model: 'gpt-4o', apiKey: process.env.NEXT_PUBLIC_OPENAI_API_KEY, });
const { messages, onRequest } = useXChat({ request });
return ( <XProvider> <div style={{ height: '100vh', display: 'flex', flexDirection: 'column' }}> <Conversations items={messages.map((msg) => ({ key: msg.id, label: msg.content.slice(0, 30), }))} /> <div style={{ flex: 1, overflow: 'auto', padding: 16 }}> {messages.map((msg) => ( <Bubble key={msg.id} placement={msg.role === 'user' ? 'end' : 'start'} content={<XMarkdown content={msg.content} />} /> ))} </div> <Sender onSubmit={onRequest} /> </div> </XProvider> );}In under 40 lines of code, a complete chat application with streaming Markdown rendering, session management, and AI conversation is implemented.
9. 2026 Version Evolution
Ant Design X maintained a monthly minor version release cadence in 2026:
| Version | Date | Key Updates |
|---|---|---|
| v2.7.0 | April 30 | XProvider zeroRuntime, XCard path reference resolution |
| v2.6.0 | April 17 | XMarkdown tail indicator component, x-components/x-card skill |
| v2.5.0 | March 31 | π₯ X Card dynamic card module released |
| v2.4.0 | March 13 | π₯ Folder file tree component, x-markdown skill |
| v2.3.0 | February 26 | useXChat queueRequest, XMarkdown escapeRawHtml |
| v2.2.0 | January | Mermaid configuration, XRequest reconnection, XMarkdown performance testing |
As can be seen, the team continues to explore the boundaries of AI interface as codeβfrom streaming Markdown to dynamic card rendering, every step is lowering the barrier to AI application development.
10. Comparison with Competitors
| Dimension | Ant Design X | Vercel AI SDK | LangChain |
|---|---|---|---|
| Positioning | UI component library + data flow | Data flow + lightweight UI | AI orchestration framework |
| UI system | β Complete component library | β οΈ Basic components | β None |
| Streaming Markdown | β XMarkdown | β οΈ Basic support | β None |
| Hybrid UI | β Hybrid UI | β | β |
| Framework binding | React-exclusive | React/Vue/Svelte | Cross-platform |
Ant Design Xβs differentiated advantage lies in its complete UI component system + streaming rendering engine, not just data flow management.
11. Summary
Ant Design Xβs positioning is very clear: it is not yet another AI SDK, but a complete AI interface solution. Its core value lies in:
- The RICH paradigm provides a standardized design language for AI interaction
- Hybrid UI breaks down the boundary between GUI and Chat
- XMarkdown solves the core pain points of streaming rendering
- X Card opens up the possibility of AI dynamically generating UI
- Out-of-the-box: seamless integration with the Ant Design ecosystem, zero learning curve for React developers
If you are developing an AI chat product, Agent console, or intelligent assistant interface, Ant Design X is currently the most mature and complete choice in the React ecosystem.