Contio

A desktop app that orchestrates multi-model LLM discussions. Multiple AI models discuss a user prompt in rounds, like a Roman senate - challenging claims, building on points, and synthesizing positions.

Port of Curia from C#/.NET to Tauri + SvelteKit.

Features

• Multi-model discussions - Select 2+ models to debate any topic
• Provider support - OpenAI, Anthropic, Google, and OpenRouter
• Streaming responses - See responses as they're generated
• Auto-conclusion - Optionally end discussions when consensus is reached
• Discussion summaries - AI-generated summaries of completed discussions
• Conversation history - Persist and revisit past discussions
• Local storage - API keys and conversations stored in browser localStorage

Getting Started

Prerequisites

• Node.js 18+
• Rust (for Tauri)

Installation

# Install dependencies
npm install

# Run in development mode
npm run tauri dev

# Build for production
npm run tauri build

Configuration

1. Launch the app
2. Go to Settings (gear icon)
3. Enter API keys for the providers you want to use
4. Select default models (optional)

Tech Stack

Project Structure

src/
├── lib/
│   ├── models/          # Type definitions
│   ├── stores/          # Svelte 5 state (runes)
│   ├── providers/       # LLM SDK wrappers
│   ├── orchestration/   # Discussion logic
│   └── persistence/     # localStorage helpers
├── components/          # UI components
└── routes/              # SvelteKit pages

Development

# Type check
npm run check

# Build frontend only
npm run build

# Run Tauri dev server
npm run tauri dev

License

MIT