Chatbot Widget Creator Skill

Purpose

Creates a production-ready ChatGPT-style chatbot widget with advanced features:

• Custom Portrait Layout: Compact 380px x 700px styling matching ChatGPT aesthetics
• Infinite re-render protection using useReducer and split context pattern
• Text selection "Ask AI" functionality with smart tooltips
• Streaming responses with Server-Sent Events (SSE)
• RAG backend integration ready for FastAPI/Qdrant setup
• Performance monitoring and debugging utilities
• Modern glassmorphic UI with ChatGPT-style interface

Key Improvements Based on Real Implementation

1. Refined UI/UX

• Compact Dimensions: 380px width x 700px height for optimal sidebar integration
• Clean Message UI: Removed timestamps, read receipts, and file uploads for a cleaner reading experience
• Themed Integration: Dynamically uses

  var(--ifm-color-primary)

  to match host site branding
• Bottom Alignment: Message bubbles and avatars are bottom-aligned for a modern chat look
• Minimalist Indicators: Simplified 3-dot pulsing animation for thinking state

2. State Management Architecture

• useReducer pattern instead of multiple useState hooks
• Split context (StateContext + ActionsContext) to prevent unnecessary re-renders
• Stable callbacks with proper dependencies to avoid circular references
• AbortController for proper cleanup of streaming connections

3. Performance Optimizations

• React.memo wrapping for expensive components like

  MessageBubble

• useMemo for computed values and complex operations
• useCallback with stable dependencies
• Render counter utilities for debugging
• Virtualization support for long conversations (50+ messages)

4. Text Selection Feature

• useTextSelection hook for detecting text selections
• Smart positioning tooltip with edge detection
• Context-aware prompts when asking about selected text
• Length validation with truncation warnings

Prerequisites

1. Backend Requirements:

   • FastAPI or similar with SSE support
   • Endpoint:

     POST /api/chat

     returning Server-Sent Events
   • Request format:

     { "question": string, "stream": boolean }

   • Response format: SSE with

     { "type": "chunk", "content": string }

2. Frontend Dependencies:

   npm install react framer-motion react-markdown react-syntax-highlighter remark-gfm lucide-react clsx tailwind-merge
   # Note: No ChatKit dependency - this is a custom implementation
   

Quick Start

1. Create the Widget Structure

# Create main directory structure
mkdir -p src/components/ChatWidget/{components,hooks,contexts,utils,styles}

# Copy all template files
cp -r .claude/skills/chatbot-widget-creator/templates/* src/components/ChatWidget/

2. Backend API Requirements

Your backend must implement:

@app.post("/api/chat")
async def chat_endpoint(request: ChatRequest):
    """Chat endpoint with Server-Sent Events streaming."""

    # Request format
    {
        "question": "What is physical AI?",
        "stream": true,
        "context": {  # Optional
            "selectedText": "...",
            "source": "User selection"
        }
    }

    # Response format (SSE)
    data: {"type": "start", "session_id": "..."}
    data: {"type": "chunk", "content": "Physical AI refers to..."}
    data: {"type": "chunk", "content": "embodied AI systems..."}
    data: {"type": "done", "session_id": "..."}

3. Integration

Add to your site root (e.g.,

src/theme/Root.tsx

):

import React from 'react';
import AnimatedChatWidget from '../components/ChatWidget/components/AnimatedChatWidget';

export default function Root({ children }) {
  const getChatEndpoint = () => {
    const hostname = window.location.hostname;
    if (hostname === 'localhost') {
      return 'http://localhost:7860/api/chat';
    }
    return 'https://your-domain.com/api/chat';
  };

  return (
    <>
      {children}
      <AnimatedChatWidget
        apiUrl={getChatEndpoint()}
        maxTextSelectionLength={2000}
        fallbackTextLength={5000}
      />
    </>
  );
}

Architecture Details

State Management (Critical for Performance)

// Consolidated state to prevent fragmentation
interface ChatState {
  messages: ChatMessage[];
  isOpen: boolean;
  isThinking: boolean;
  currentStreamingId?: string;
  error: Error | null;
  renderCount: number;
}

// Split context pattern
const ChatStateContext = createContext<{ state: ChatState }>();
const ChatActionsContext = createContext<{ actions: ChatActions }>();

// Components only subscribe to what they need
const messages = useChatSelector(s => s.messages);  // Re-renders on messages change
const actions = useChatActions();                   // Never re-renders

Key Anti-Patterns Avoided

1. ❌ Multiple useState hooks:

   // Bad - causes context fragmentation
   const [messages, setMessages] = useState([]);
   const [isOpen, setIsOpen] = useState(false);
   

2. ✅ Consolidated useReducer:

   // Good - single state source
   const [state, dispatch] = useReducer(chatReducer, initialState);
   

3. ❌ Circular dependencies:

   // Bad - callback depends on state that changes
   const handleChunk = useCallback((chunk) => {
     if (session.currentStreamingId) {  // Dependency on state
       updateMessage(session.currentStreamingId, chunk);
     }
   }, [session.currentStreamingId]); // Infinite re-render!
   

4. ✅ Stable references:

   // Good - no circular dependencies
   const streamingIdRef = useRef<string>();
   const handleChunk = useCallback((chunk) => {
     if (streamingIdRef.current) {
       dispatch(updateStreamingAction(streamingIdRef.current, chunk));
     }
   }, [dispatch]); // Stable dependency array
   

Streaming Response Handling

// Proper SSE parsing
const lines = chunk.split('\n');
for (const line of lines) {
  if (line.startsWith('data: ')) {
    const data = line.slice(6);
    if (data !== '[DONE]') {
      const parsed = JSON.parse(data);

      if (parsed.type === 'chunk' && parsed.content) {
        handleChunk(parsed.content);
      } else if (parsed.type === 'done') {
        handleComplete();
      }
    }
  }
}

Customization Guide

Theming

Edit

src/components/ChatWidget/styles/ChatWidget.module.css

:

/* Primary colors */
.widget {
  background: #171717; /* Dark mode background */
  width: 380px;
  height: 700px;
}

/* Message bubbles */
.userMessageBubble {
  background: var(--ifm-color-primary); /* Theme integration */
  color: white;
}

.aiMessageBubble {
  background: #21262d;
  border: 1px solid #30363d;
}

Component Reference

Core Components

1. AnimatedChatWidget: Main entry point with animations
2. ChatInterface: ChatGPT-style UI container
3. MessageBubble: Individual message display (React.memo optimized)
4. InputArea: Message input (No file upload)
5. SelectionTooltip: Text selection "Ask AI" tooltip
6. ThinkingIndicator: Minimalist 3-dot pulsing animation

Hooks

1. useChatSession: Access chat state and actions
2. useStreamingResponse: Handle SSE connections
3. useTextSelection: Detect and handle text selection
4. useErrorHandler: Centralized error management

Utilities

1. chatReducer: State transitions
2. api.ts: API request/response formatting
3. animations.ts: Framer Motion configs

Troubleshooting

Common Issues and Solutions

1. Infinite Re-renders

   • Check for circular dependencies in useCallback
   • Ensure split context pattern is properly implemented
   • Use React DevTools Profiler to identify causes

2. Memory Leaks

   • Ensure AbortController cleanup on unmount
   • Check for unclosed SSE connections
   • Monitor with

     window.performance.memory

3. SSE Not Working

   • Verify CORS headers include

     text/event-stream

   • Check that responses use correct

     data: {}

     format
   • Ensure

     Cache-Control: no-cache

     is set
   • Crucial: Ensure your API request includes

     stream: true

     in the body
   • Extract content correctly:

     handleChunk(JSON.parse(data).content)

     not the raw string

4. Text Selection Issues

   • Verify

     useTextSelection

     is enabled
   • Check for CSS

     user-select: none

     conflicts
   • Ensure z-index is high enough for tooltip

5. Markdown Rendering & Layout Issues

   • List Numbers on Separate Lines:
     • Set

       list-style-position: inside

       on

       ul

       /

       ol

     • Reset

       p

       margins in lists:

       li > p { margin: 0; display: inline; }

   • Invisible Text in Light Mode:
     • Explicitly define text colors for

       .messageRenderer

       , headings, and lists in light mode media queries or classes
     • Ensure

       .aiMessageBubble

       background is transparent or correctly colored
   • Excessive Spacing:
     • Reduce margins on

       p

       ,

       ul

       ,

       ol

       (e.g.,

       0.25rem

       )
     • Remove margins from first/last children:

       .messageRenderer > *:first-child { margin-top: 0; }

     • Use

       !important

       if global site styles (like Docusaurus) override your chat styles
   • Avatar Alignment:
     • Use CSS Grid for message layout to align avatar with the bottom of the text
     • Example:

       .aiMessage { display: grid; align-items: end; }

Production Checklist

• Remove console.log statements in production
• Add proper error tracking (Sentry, etc.)
• Implement rate limiting on backend
• Add CORS configuration
• Test on slow networks
• Verify memory leak prevention
• Test with long conversations (100+ messages)

Result

A production-ready chat widget that:

• Matches the ChatGPT visual aesthetic
• Never crashes from infinite re-renders
• Provides smooth text selection interactions
• Streams responses efficiently
• Maintains performance over time
• Handles all edge cases gracefully