ChenChen - Multi-LLM Prompt Desktop App

Send prompts to multiple LLM providers simultaneously - A privacy-focused desktop application built with Tauri and Rust.

Demo

Features

• 🤖 Multi-Provider Support: Send prompts to ChatGPT, Gemini, and Claude simultaneously
• 🔒 Privacy-First: No credential storage, no telemetry, all data stays local
• 💾 Persistent Sessions: Provider webviews with persistent authentication sessions
• 📊 Real-Time Status: Track submission progress with retry logic and timeout handling
• 🔐 Authentication Detection: Automatic detection of login requirements
• 🎨 Split-Screen Layout: Responsive grid layout for viewing multiple providers
• ⚡ Fast & Lightweight: Built with Rust and Tauri 2.0

Quick Start

Prerequisites

• Rust (latest stable) - Install Rust
• Node.js (v18+) - Install Node.js
• npm or pnpm

Installation

# Clone the repository
git clone https://github.com/your-org/chenchen.git
cd chenchen

# Install frontend dependencies
npm install

# Install Tauri CLI
cargo install tauri-cli

# Run in development mode
npm run tauri dev

Build for Production

# Build optimized binary
npm run tauri build

# Binary will be in src-tauri/target/release/

Usage

1. Launch ChenChen

   • The app opens with all three providers (ChatGPT, Gemini, Claude) available

2. Select Providers

   • Check the providers you want to send prompts to (1-3 providers)
   • Login indicators show which providers need authentication

3. Authenticate

   • Click "Login Required" buttons to authenticate with providers
   • Sessions are persisted across app restarts

4. Send Prompts

   • Enter your prompt in the text area
   • Click "Send Prompt" or press Enter
   • View real-time submission status for each provider

Development

Project Structure

chenchen/
├── src/                    # Svelte frontend
│   ├── components/         # UI components
│   ├── routes/             # Svelte routes
│   ├── services/           # TypeScript services
│   ├── app.css             # Global styles
│   ├── app.html            # App entry point
│   └── types.ts            # TypeScript types
├── src-tauri/              # Rust backend
│   ├── src/
│   │   ├── injection/      # JavaScript injection
│   │   ├── layout/         # Layout calculation
│   │   ├── providers/      # Provider management
│   │   ├── status/         # Submission tracking
│   │   ├── commands.rs     # Tauri commands (IPC)
│   │   ├── lib.rs          # Library entry point
│   │   ├── logging.rs      # Structured logging
│   │   ├── main.rs         # Binary entry point
│   │   ├── state.rs        # App state management
│   │   └── types.rs        # Shared types
│   └── tests/              # Rust tests
└── docs/                   # Documentation
    ├── privacy-policy.md
    ├── testing-guide.md
    └── linux-webview-positioning-fix.md

Running Tests

# Backend tests (Rust)
cd src-tauri
cargo test

# Frontend tests (TypeScript/Svelte)
npm test

# All tests
npm run test:all

Code Quality

# Rust linting
cargo clippy

# Rust formatting
cargo fmt

# TypeScript checking
npm run check

Architecture

Backend (Rust + Tauri 2.0)

• Commands: Tauri IPC interface for frontend communication
• Provider Manager: Manages provider selection state (1-3 providers)
• Layout Calculator: Computes split-screen dimensions based on selected providers
• Webview Manager: Platform-specific session persistence
• Injector: JavaScript injection for prompt submission
• Status Tracker: Submission state machine with retry logic
• Structured Logging: Dual-format (JSON + human-readable) logging

Frontend (Svelte 5 + TypeScript)

• ProviderSelector: Provider selection and authentication UI
• PromptInput: Prompt text area with validation
• StatusDisplay: Real-time submission status tracking
• Provider Panels: Split-screen webview containers

Data Flow

User Input → PromptInput → submit_prompt command
                                    ↓
                          StatusTracker creates submissions
                                    ↓
                          Injector generates scripts
                                    ↓
                          Execute in provider webviews
                                    ↓
                          Status updates → StatusDisplay

Privacy Guarantees

ChenChen is committed to user privacy:

• ✅ No credential storage - Passwords never leave webview sandboxes
• ✅ No prompt history - All data in-memory only
• ✅ No telemetry - Zero analytics or tracking
• ✅ Local-only operation - No backend servers
• ✅ Provider domains only - Network requests limited to LLM providers

See docs/privacy-policy.md for full details.

Verify Privacy

# Run privacy tests
cd src-tauri
cargo test privacy_test

# Manual network monitoring (see docs/testing-guide.md)

Testing

Test Coverage

• Unit Tests: 26 tests (providers, layout, injection, status, logging)
• Contract Tests: 19 tests (public API interfaces)
• Integration Tests: 24 tests (privacy, success rate, logging formats)
• Total: 69 automated tests

Success Criteria

• ✅ >=95% success rate for prompt submissions (SC-002)
• ✅ <10 second submission time to 3 LLMs (SC-001)
• ✅ Zero data collection verified by privacy tests

Configuration

Provider Configuration

Edit src-tauri/config/providers.json:

{
  "config_version": "1.0.0",
  "providers": {
    "ChatGPT": {
      "url": "https://chat.openai.com/",
      "input_selectors": ["textarea", "#prompt-textarea"],
      "submit_selectors": ["button[data-testid='send-button']"],
      "auth_check_selectors": ["button[data-testid='login-button']"]
    }
  }
}

Troubleshooting

"Failed to load providers"

• Ensure src-tauri/config/providers.json exists
• Check JSON syntax validity

"Login Required" persists after login

• Clear webview data: delete ~/.local/share/com.chenchen.app/webviews/
• Restart the application

Slow prompt submission

• Check network connection
• Verify provider websites are accessible
• Review timeout settings (30s default)

Linux: Webviews stacking vertically instead of positioning correctly

This is a known GTK issue that affects all Tauri multiwebview applications on Linux. The project includes patches to fix this:

Symptoms:

• Multiple webviews stack vertically
• set_webview_position() and set_webview_size() are ignored
• Works fine on Windows/macOS but fails on Linux

Solution: The patches are already applied in Cargo.toml. If you cloned the repository, they should work automatically. To verify:

# Check that patches are applied
cargo tree -p tao --depth 0
cargo tree -p tauri-runtime-wry --depth 0

# Should show GitHub URLs from Benjaminlooi forks

For more details, see docs/linux-webview-positioning-fix.md

Related Issues:

• tauri-apps/tauri#10420
• tauri-apps/tauri#13071

Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Follow TDD: write tests first
4. Ensure all tests pass (cargo test && npm test)
5. Run linters (cargo clippy && npm run check)
6. Commit with descriptive messages
7. Push and create a Pull Request

Development Principles

• TDD First: Write tests before implementation
• Privacy by Design: No data collection ever
• Dual Logging: JSON + human-readable formats
• Type Safety: Rust + TypeScript with strict typing

License

MIT License

Acknowledgments

• Built with Tauri 2.0
• Frontend powered by Svelte 5
• LLM providers: OpenAI, Google, Anthropic

---

Version: 0.1.1 Status: MVP Complete Platform: Windows, macOS, Linux (with positioning patches)