agentsclimarketplace

Ai elements chatbot

Skill ComeOnOliver/skillshub/skills/secondsky/claude-skills/ai-elements-chatbot

shadcn/ui AI chat components for conversational interfaces. Use for streaming chat, tool/function displays, reasoning visualization, or encountering Next.js App Router setup, Tailwind v4 integration, AI SDK v5 migration errors.From its SKILL.md

Install
npx -y skills add ComeOnOliver/skillshub --skill ai-elements-chatbot

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its file declares

Copied from the file, not written here

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

7.7 KB, ~1.8k tokens by cl100k_base, as published. Nobody here has run it

AI Elements Chatbot Components

Status: Production Ready ✅ | Last Verified: 2025-11-18


What Is AI Elements?

Production-ready chat UI components for AI applications:

  • Built on shadcn/ui
  • 30+ components (Message, Conversation, Response, etc.)
  • Works with Vercel AI SDK v5
  • Streaming support
  • Tool/function call displays
  • Reasoning visualization

Quick Start (15 Minutes)

Prerequisites

  • Next.js 15+ (App Router)
  • shadcn/ui initialized
  • Tailwind v4
  • AI SDK v5+

1. Initialize

pnpm dlx ai-elements@latest init

2. Add Components

pnpm dlx ai-elements@latest add message conversation response prompt-input

3. Create Chat Interface

'use client';

import { useChat } from 'ai/react';
import { Conversation } from '@/components/ui/ai/conversation';
import { Message } from '@/components/ui/ai/message';
import { Response } from '@/components/ui/ai/response';
import { PromptInput } from '@/components/ui/ai/prompt-input';

export default function ChatPage() {
  const { messages, input, handleInputChange, handleSubmit } = useChat({
    api: '/api/chat'
  });

  return (
    <div className="flex h-screen flex-col">
      <Conversation>
        {messages.map((msg) => (
          <Message key={msg.id} role={msg.role}>
            <Response markdown={msg.content} />
          </Message>
        ))}
      </Conversation>

      <PromptInput
        value={input}
        onChange={handleInputChange}
        onSubmit={handleSubmit}
      />
    </div>
  );
}

Load references/setup-guide.md for complete setup.


Core Components

Message & Conversation

import { Conversation } from '@/components/ui/ai/conversation';
import { Message } from '@/components/ui/ai/message';

<Conversation>
  {messages.map((msg) => (
    <Message key={msg.id} role={msg.role}>
      {msg.content}
    </Message>
  ))}
</Conversation>

Response (Markdown)

import { Response } from '@/components/ui/ai/response';

<Response markdown={content} />

PromptInput

import { PromptInput } from '@/components/ui/ai/prompt-input';

<PromptInput
  value={input}
  onChange={handleInputChange}
  onSubmit={handleSubmit}
/>

CodeBlock

import { CodeBlock } from '@/components/ui/ai/code-block';

<CodeBlock code={code} language="typescript" />

Reasoning (Thinking)

import { Reasoning } from '@/components/ui/ai/reasoning';

<Reasoning content={thinking} />

Tool (Function Calls)

import { Tool } from '@/components/ui/ai/tool';

<Tool name="search" args={{ query: "..." }} result={result} />

Critical Rules

Always Do ✅

  1. Install shadcn/ui first (AI Elements requires it)
  2. Use Next.js App Router (Pages Router not supported)
  3. Use AI SDK v5 (breaking changes from v4)
  4. Install via CLI (pnpm dlx ai-elements@latest)
  5. Update components.json with registry
  6. Use client components ('use client' directive)
  7. Stream responses for better UX
  8. Handle loading states
  9. Add error boundaries
  10. Test on mobile

Never Do ❌

  1. Never install as npm package (components are copied)
  2. Never use Pages Router (only App Router)
  3. Never use AI SDK v4 (breaking changes)
  4. Never skip prerequisites (shadcn/ui, Tailwind)
  5. Never modify core types (extends shadcn types)
  6. Never use without streaming (defeats purpose)
  7. Never skip accessibility (ARIA labels)
  8. Never hardcode styles (use Tailwind)
  9. Never skip error handling (API failures)
  10. Never ignore mobile (responsive required)

Available Components (30+)

Core:

  • Message
  • Conversation
  • Response
  • PromptInput

Content:

  • CodeBlock
  • Markdown
  • Tool
  • Reasoning
  • Sources

Actions:

  • Actions
  • CopyButton
  • ShareButton
  • RegenerateButton

Advanced:

  • BranchNavigation
  • ThinkingDisplay
  • WebPreview

Common Use Cases

Use Case 1: Basic Chat

const { messages, input, handleInputChange, handleSubmit } = useChat();

return (
  <>
    <Conversation>
      {messages.map(m => (
        <Message key={m.id} role={m.role}>
          <Response markdown={m.content} />
        </Message>
      ))}
    </Conversation>
    <PromptInput value={input} onChange={handleInputChange} onSubmit={handleSubmit} />
  </>
);

Use Case 2: With Tool Calls

{messages.map(m => (
  <Message key={m.id} role={m.role}>
    {m.toolInvocations?.map(tool => (
      <Tool key={tool.toolCallId} name={tool.toolName} args={tool.args} result={tool.result} />
    ))}
    <Response markdown={m.content} />
  </Message>
))}

Use Case 3: With Reasoning

<Message role="assistant">
  {reasoning && <Reasoning content={reasoning} />}
  <Response markdown={content} />
</Message>

Use Case 4: With Code Blocks

<Response markdown={content}>
  {(node) => node.type === 'code' ? (
    <CodeBlock code={node.value} language={node.lang} />
  ) : null}
</Response>

Use Case 5: With Sources

<Message role="assistant">
  <Response markdown={content} />
  <Sources sources={sources} />
</Message>

API Routes

Basic Streaming

// app/api/chat/route.ts
import { streamText } from 'ai';
import { openai } from '@ai-sdk/openai';

export async function POST(req: Request) {
  const { messages } = await req.json();

  const result = streamText({
    model: openai('gpt-4'),
    messages
  });

  return result.toDataStreamResponse();
}

With Tools

const result = streamText({
  model: openai('gpt-4'),
  messages,
  tools: {
    search: {
      description: 'Search the web',
      parameters: z.object({ query: z.string() }),
      execute: async ({ query }) => {
        return await search(query);
      }
    }
  }
});

When to Use AI Elements

Use when:

  • Building ChatGPT-style interface
  • Need production-ready components
  • Using Vercel AI SDK
  • Want streaming responses
  • Need tool/function displays
  • Want reasoning visualization

Don't use when:

  • Not using Next.js App Router
  • Don't have shadcn/ui
  • Need Pages Router
  • Building custom design system

Resources

References (references/):

  • component-catalog.md - All 8 AI Elements components with examples
  • example-reference.md - Complete integration examples and patterns
  • setup-guide.md - Step-by-step setup with Next.js 15 and shadcn/ui

Templates (templates/):

  • Component examples available in reference files

Official Documentation


Questions? Issues?

  1. Check references/setup-guide.md for complete setup
  2. Verify prerequisites (Next.js 15+, shadcn/ui, AI SDK v5)
  3. See official examples

What ships with it

30.3 KB alongside SKILL.md, 3 of them executable

GitHub clipped this repository’s file list, so this is at least 8 files and may be more.

scripts/

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.