Better Cursor - LLM Council IDE

A VS Code-like IDE powered by an LLM Council (inspired by Andrej Karpathy's llm-council). Instead of using a single LLM, this IDE uses a council of multiple LLMs that review and rank each other's responses, with a Chairman LLM synthesizing the final answer.

Features

1. First Opinions: All council LLMs provide initial responses
2. Review: Each LLM reviews and ranks other responses
3. Final Response: Chairman LLM synthesizes the best insights

Setup

Prerequisites

• Python 3.10+
• Node.js 18+ and npm
• OpenRouter API key

Installation

1. Clone or download this repository

2. Set up the backend:

   # Install Python dependencies (using uv or pip)
   pip install -r requirements.txt
   # OR if you have uv:
   uv sync
   

3. Set up the frontend:

   cd frontend
   npm install
   cd ..
   

4. Configure your API key:

   # Copy the example env file
   cp env.example .env
   
   # Edit .env and add your OpenRouter API key
   # OPENROUTER_API_KEY=sk-or-v1-your-actual-key-here
   
   # Optional: Adjust max tokens based on your credits
   # Lower values = less credits needed, but shorter responses
   # MAX_TOKENS=256  # Default (works with low credits)
   # MAX_TOKENS=512  # Medium
   # MAX_TOKENS=2048 # High (needs more credits)
   

5. Configure models (important):

   You must verify model names are available on OpenRouter!

   • Visit https://openrouter.ai/models to see available models
   • Check the exact model ID format (e.g., openai/gpt-4o, anthropic/claude-3.5-sonnet)
   • Edit backend/config.py to use models that exist:

   COUNCIL_MODELS = [
       "openai/gpt-4o",              # Verify this exists
       "anthropic/claude-3.5-sonnet", # Verify this exists
       "openai/gpt-3.5-turbo",       # Fallback option
   ]
   
   CHAIRMAN_MODEL = "openai/gpt-4o"  # Use a reliable model
   

   Note: Model names change frequently. If you get 404 errors, check OpenRouter's model list and update the config.

Running the Application

Option 1: Use the start script

chmod +x start.sh
./start.sh

Option 2: Run manually

Terminal 1 (Backend):

# Load environment variables and run backend
export $(cat .env | xargs)
python -m backend.main
# OR with uv:
uv run python -m backend.main

Terminal 2 (Frontend):

cd frontend
npm run dev

Then open http://localhost:5173 in your browser.

Usage

1. Chat with LLM Council:

   • Click the chat panel on the right (or the chat button if hidden)
   • Type your question and press Cmd/Ctrl + Enter to send
   • The council will discuss and provide a final answer
   • You can view individual LLM responses by clicking the tabs

2. View Individual Responses:

   • After receiving a response, click on the model tabs to see what each LLM said
   • The "Final Response" tab shows the Chairman's synthesized answer

Project Structure

better-cursor/
├── backend/
│   ├── __init__.py
│   ├── config.py          # Configuration (models, API keys)
│   └── main.py             # FastAPI backend with LLM Council logic
├── frontend/
│   ├── src/
│   │   ├── App.jsx         # Main React component
│   │   ├── App.css         # Styles
│   │   ├── main.jsx        # React entry point
│   │   └── index.css       # Global styles
│   ├── index.html
│   ├── package.json
│   └── vite.config.js
├── data/
│   └── conversations/      # Saved conversations (auto-created)
├── .env                    # Your API keys (create from env.example)
├── env.example             # Example environment file
├── pyproject.toml          # Python dependencies
├── README.md
└── start.sh                # Startup script

Tech Stack

• Backend: FastAPI (Python), OpenRouter API, async httpx
• Frontend: React + Vite, Monaco Editor, react-markdown
• Storage: JSON files in data/conversations/

Customization

Change Council Models

Edit backend/config.py:

COUNCIL_MODELS = [
    "openai/gpt-4o",
    "google/gemini-2.0-flash-exp",
    # Add your preferred models
]

Change Chairman Model

Edit backend/config.py:

CHAIRMAN_MODEL = "google/gemini-2.0-flash-exp"

Change Ports

• Backend: Edit backend/config.py (default: 8000)
• Frontend: Edit frontend/vite.config.js (default: 5173)

Notes

• Make sure you have credits on OpenRouter or have automatic top-up enabled
• Conversations are saved in data/conversations/ as JSON files
• The backend loads environment variables from .env file

License

This project is provided as-is for inspiration and learning purposes.