Shinzo-UI

A robust, single standalone html file AI interface for Ollama and OpenAI. Features local RAG with vector search, real-time voice/video calls, document analysis (PDF/DOCX), and project workspaces. Runs entirely in the browser using IndexedDB for privacy and speed—no external backend required.

How to use

• Just download the html file
• Configure the API, if u have OpenAI's API or Setup Base URL if u use Ollama or Llama.cpp
• Then you're good to go.
• Have fun prompting.

✨ Key Features

🎯 Core Functionality

• Chat Interface: Conversational UI with markdown rendering, code syntax highlighting, and LaTeX support
• Knowledge Base: Document upload & vector storage (PDF, DOCX, TXT, etc.)
• Workspace Management: Containerized project spaces with custom instructions
• Model Builder: Create custom models using Modelfile syntax
• Live Web Search: Real-time internet search integration

🎨 User Experience

• Dark mode UI with modern design
• Responsive layout (mobile & desktop)
• Voice/Video call capabilities
• Real-time speech recognition (STT) and synthesis (TTS)
• Concurrent model conversations
• Citation and source tracking

🛠️ Technical Features

• Browser-based Vector Database: IndexedDB storage for documents and conversations
• Multiple Embedding Providers: Ollama, Local (Transformers.js), OpenAI
• File Processing: PDF, DOCX, images, audio, and text files
• Export/Import: Chat history backup and restore
• Real-time Search: Live search across chats and knowledge base

🚀 Quick Start

Prerequisites

• Modern web browser (Chrome/Firefox/Edge)
• Ollama server (optional, for local AI models)
• Internet connection (for CDN dependencies)

Installation

1. Download the Shinzo UI.html file
2. Open it directly in your browser
3. Configure your API settings (Settings tab)

Basic Configuration

1. Set your API URL (default: http://localhost:11434 for Ollama)
2. Configure embedding provider (Ollama recommended for local use)
3. Set up web search API key (optional, for live search)

📁 File Structure

The application is a single HTML file containing:

• HTML5: Structure and layout
• CSS3: Styling with CSS variables for theming
• JavaScript: Complete application logic
• Embedded Libraries:
  • Font Awesome (icons)
  • KaTeX (mathematical rendering)
  • Marked.js (markdown parsing)
  • DOMPurify (HTML sanitization)
  • PDF.js (PDF processing)
  • jsPDF (PDF generation)
  • Mammoth (DOCX processing)

🎮 Usage Guide

Starting a Chat

1. Click "New Chat" in the sidebar
2. Select your primary model from the dropdown
3. Type your message or use voice input
4. Toggle web search for real-time information

Managing Knowledge Base

1. Navigate to "Knowledge" tab
2. Upload documents using the upload button
3. Documents are automatically chunked and embedded
4. Use # symbol in chats to trigger RAG context

Creating Workspaces

1. Go to "Spaces" tab
2. Click "Create Space"
3. Add custom instructions and files
4. Start space-specific chats

Voice/Video Calls

1. Click phone or video icon in chat header
2. Grant microphone/camera permissions
3. Speak naturally - VAD (Voice Activity Detection) automatically detects speech
4. Toggle mute/video during calls

⚙️ Configuration Options

API Settings

• API URL: Backend server address (Ollama, OpenAI, etc.)
• API Key: Authentication token
• Embedding Provider: Ollama, Local browser, or OpenAI
• Chunk Size: Document processing size (default: 1000)
• Top K Results: RAG retrieval count (default: 3)

Web Search

• Serper API Key: For enhanced search results
• Proxy Mode: CORS or custom proxy for web scraping

Voice Settings

• STT Endpoint: Speech-to-text API endpoint
• TTS Endpoint: Text-to-speech API endpoint
• VAD Enabled: Voice activity detection
• Playback Speed: Audio playback rate

🔧 Advanced Features

Model Builder

Create custom AI models using Modelfile syntax:

1. Navigate to "Model Builder" tab
2. Select base model
3. Define system prompt and parameters
4. Click "Create Model"

Concurrent Conversations

• Toggle concurrent mode to use two models simultaneously
• Each model responds independently for comparison

Export/Import

• Export chat history as JSON
• Import previous conversations
• Archive chats for organization

🗂️ Data Storage

IndexedDB Structure

superRAG_DB (v5)
├── documents (metadata)
├── chunks (vector embeddings)
├── conversations (chat history)
└── spaces (workspace configurations)

Local Storage

• Application settings
• UI preferences
• Connection configurations

🔐 Security Notes

Data Privacy

• All data stored locally in browser
• No external data transmission unless configured
• API keys stored in browser local storage

File Processing

• Documents processed client-side
• Images converted to base64 for attachments
• TXT/PDF/DOCX parsing happens in browser

🌐 CDN Dependencies

The application loads these libraries from CDN:

• Font Awesome 6.5.1 (icons)
• KaTeX 0.16.9 (math rendering)
• Marked.js (markdown)
• DOMPurify 3.0.6 (security)
• PDF.js 3.11.174 (PDF processing)
• jsPDF 2.5.1 (PDF generation)
• Mammoth 1.6.0 (DOCX processing)
• Transformers.js (for local embeddings)

🐛 Troubleshooting

Common Issues

1. Connection Failed

   • Verify Ollama server is running
   • Check API URL in settings
   • Disable browser CORS restrictions if needed

2. File Upload Issues

   • Check file size limits
   • Ensure supported file types
   • Try smaller chunks in settings

3. Voice Features Not Working

   • Grant microphone permissions
   • Check STT/TTS endpoint configuration
   • Verify browser supports WebRTC

Browser Compatibility

• Chrome 90+ (recommended)
• Firefox 88+
• Edge 90+
• Safari 14.1+

📝 Development

Customization

• Modify CSS variables in :root for theming
• Extend VectorDB class for different storage backends
• Add new file processors in extractTextFromFile()

Adding Features

1. New view sections can be added to HTML
2. Extend app object with new methods
3. Update sidebar navigation accordingly

📄 License

Created by One Man Studios

This project is provided as-is for educational and personal use. Commercial use is strictly prohibited.

🤝 Contributing

Feel free to:

1. Fork the repository
2. Give a star rating 🌟
3. Submit pull requests

Note: This is a client-side application. All processing happens in your browser. No data is sent to external servers unless explicitly configured in settings.