@kosmas10/chat-and-verify
v0.0.22
Published
A chatbot with citations and highlights in the source document for easy verification of answers. Supports PDF and Word documents. Built with React for maximum reliability.
Downloads
1,926
Maintainers
Readme
Chat & Verify
Version: 0.0.11
Package: @kosmas10/chat-and-verify
A web-based chatbot application that provides citations and highlights in source documents for easy verification of AI-generated answers. Supports PDF and Word documents with an intelligent question-answering system powered by Claude or OpenAI.
🎯 Key Features
- 📄 Document Support: Upload and analyze PDF and Word (DOCX) documents
- 💬 AI-Powered Chat: Ask questions about your documents using Claude or OpenAI
- 🔍 Source Verification: Every answer includes exact quotes from the source document
- ✨ Smart Highlighting: Automatically highlights relevant passages in the document
- 🎨 Modern UI: Clean, responsive interface with real-time document viewing
- 🔒 Privacy-First: Your API keys are stored locally, never sent to third parties
- ⚡ Fast Loading: 2-stage bundled React architecture for reliable, efficient loading
- 🎯 Claude Artifact Support: Automatically detects and uses Claude when running as a Claude Artifact
🏗️ Architecture: 2-Stage Bundled React
This application uses a simplified 2-stage loading architecture with bundled React:
Stage 1: Bootstrap (User Installs This)
- File:
chat-and-verify-bootstrap.html - Purpose: Ultra-minimal loader (~25 lines)
- What it does: Fetches the latest bundled app from npm/CDN
Stage 2: Bundled Application (Auto-Updates)
- Package:
@kosmas10/chat-and-verify(this package) - File:
dist/chat-and-verify.html - Purpose: The full Chat & Verify application
- What it does: Provides the complete document chat experience
Why This Architecture?
- 🛡️ Future-Proof: Stage 1 never changes, so users never need to update their local file
- 🔄 Auto-Updates: The application updates automatically via npm/CDN
- 📦 Self-Contained: All dependencies are bundled - no external CDN issues
- ⚡ Fast: Single HTML file with all assets inlined
🚀 Quick Start
Option A: Using the Bootstrap (Recommended)
Download the bootstrap file (one time only):
curl -O https://cdn.jsdelivr.net/npm/@kosmas10/chat-and-verify@latest/chat-and-verify-bootstrap.htmlOpen it in your browser:
- Double-click the file, or
- Open it from your browser's File menu
That's it! The application will load automatically and stay up-to-date.
Option B: Direct CDN Link
Simply navigate to:
https://cdn.jsdelivr.net/npm/@kosmas10/chat-and-verify@latest/chat-and-verify-bootstrap.html📖 Usage
1. Configure Your LLM
If running as a Claude Artifact: The app automatically detects the Claude environment and uses Claude directly - no configuration needed!
If running standalone: On first launch, you'll be prompted to configure your AI provider:
- OpenAI: Requires an API key from platform.openai.com
- Claude: Requires an API key from console.anthropic.com
Your API key is stored locally in your browser and never sent anywhere except to your chosen LLM provider.
2. Upload a Document
Click "Select PDF or Word File" and choose:
- PDF files (.pdf)
- Word documents (.docx)
3. Ask Questions
Type your question in the chat box and press Enter or click the send button.
4. Verify Answers
- Click the 🔍 icon next to any quote to find it in the document
- Highlighted passages show you exactly where information came from
- Scroll through the document while chatting
🎨 Features in Detail
Document Viewer
- PDF Rendering: High-quality PDF display with text selection
- Word Support: Native DOCX rendering with formatting preserved
- Responsive Scaling: Automatically adjusts to your screen size
- Search & Highlight: Find and highlight text across pages
- Resizable Panel: Drag the divider to adjust viewer/chat sizes
Chat Interface
- Markdown Support: Rich text formatting in responses
- Source Citations: Every answer includes exact quotes
- Message History: Scroll through your conversation
- Context Awareness: AI remembers the conversation context
LLM Configuration
- Multiple Providers: Choose between OpenRouter (GPT, Claude, Gemini) and direct OpenAI
- Model Selection: Pick specific models (GPT-4, Claude Sonnet, etc.)
- Secure Storage: API keys stored locally in browser
- Easy Reconfiguration: Change providers anytime
🔧 Development
Prerequisites
- Node.js >= 18.0.0
- npm
Setup
cd chat-and-verify
npm installDevelopment Server
npm run devBuild
npm run buildThe build produces a single HTML file at dist/chat-and-verify.html.
📦 Package Structure
@kosmas10/chat-and-verify/
├── dist/
│ └── chat-and-verify.html # Bundled application
├── chat-and-verify-bootstrap.html # Bootstrap loader
├── README.md # This file
└── LICENSE # MIT License🌐 Browser Compatibility
- ✅ Chrome/Chromium 90+
- ✅ Firefox 88+
- ✅ Safari 14+
- ✅ Edge 90+
🔒 Privacy & Security
- Local Storage: API keys stored in browser localStorage only
- No Tracking: No analytics or tracking code
- Direct API Calls: Your data goes directly to your chosen LLM provider
- Open Source: Full source code available for inspection
🐛 Troubleshooting
Application Won't Load
- Check your internet connection
- Ensure cdn.jsdelivr.net is accessible (not blocked by firewall)
- Try clearing browser cache
- Check browser console for errors (F12)
LLM Not Responding
- Verify your API key is correct
- Check you have API credits/quota available
- Ensure the API endpoint is accessible
- Try reconfiguring with the LLM Config button
Document Won't Upload
- Ensure file is PDF or DOCX format
- Check file size (very large files may be slow)
- Try a different document to isolate the issue
📝 Version History
0.0.8 (Current)
- Complete React rewrite using V2 architecture
- Bundled React application with Tailwind CSS
- Simplified 2-stage loading (no intermediate loader needed)
- All dependencies bundled (PDF.js, mammoth, lucide-react)
- Improved component-based architecture for maintainability
- TypeScript for type safety
0.0.7
- Removed all remaining console.log statements
- Console now completely silent during normal operation
0.0.6
- Removed initial non-error console output
- Bootstrap file updated with cleaner code structure
0.0.5
- Improved Claude Artifact detection
0.0.4
- Added Claude Artifact detection and automatic configuration
- LLM Config button automatically hidden when running as Claude Artifact
0.0.3
- Bug fixes and improvements
0.0.2
- Added bootstrap loader for future-proof installations
0.0.1
- Initial release
🤝 Contributing
Contributions are welcome! Please visit: https://github.com/kosmas10/ai-chat-extensions
📄 License
MIT License - See LICENSE file for details
🆘 Support
For issues, questions, or feature requests:
- GitHub Issues: https://github.com/kosmas10/ai-chat-extensions/issues
- Package: https://www.npmjs.com/package/@kosmas10/chat-and-verify
Made with ❤️ for better AI verification
