R-Pilot

Your AI-powered R pair programmer, based on IncognitoPilot.

Prerequisites

Before you begin, ensure you have the following installed:

• Git for cloning the repository
• Python (version 3.9-3.12)
  • Windows: Download from https://www.python.org/downloads/
  • Mac: brew install python@3.12
  • Linux: sudo apt-get install python3.12
• R
  • Windows: Download from https://cran.r-project.org/bin/windows/base/
  • Mac: brew install r or download from https://cran.r-project.org/bin/macosx/
  • Linux: sudo apt-get install r-base
• Bun
  • Linux & macOS:

    curl -fsSL https://bun.sh/install | bash

  • Windows:

    powershell -c "irm bun.sh/install.ps1 | iex"

Quick Start

1. Clone and setup:

   git clone https://github.com/markuryy/R-Pilot.git
   cd R-Pilot
   bun install
   bun run setup

2. Start R-Pilot:

   bun run dev

3. Open the authentication link shown in the terminal (starts with http://localhost:3000).

   • The link includes a token that's valid for your session
   • You can share this link with others to give them access
   • The token persists in your browser for future sessions

Manual Installation

If the setup script fails, follow these steps:

1. Create and activate a virtual environment:

   pushd apps/api/services
   python -m venv venv
   # On Unix/macOS:
   source venv/bin/activate
   # On Windows:
   .\venv\Scripts\activate

2. Install dependencies:

   pip install poetry
   poetry install
   popd  # Back to root
   bun install

3. Set up environment files:

   Create apps/api/services/.env:

   OPENAI_API_KEY=your_api_key_here
   # Optional - For Cloudflare rate limiting proxy via AI Gateway
   # OPENAI_API_BASE=your_cloudflare_worker_url_here
   INTERPRETER_TYPE=r
   R_PATH=/path/to/R  # Get this using 'which R' on Unix or 'where R' on Windows
   WORKING_DIRECTORY=/path/to/workspace
   ALLOWED_HOSTS=localhost:3000
   FRONTEND_URL=localhost:3000
   ENABLE_CORS=TRUE
   # Optional: Set a specific auth token
   AUTH_TOKEN=your_chosen_token
   # Optional: Use HTTPS for auth links
   USE_HTTPS=false

   Create apps/web/.env.local:

   NEXT_PUBLIC_SERVICES_URL=http://localhost:8000

4. Start the development servers:

   bun run dev

   This command starts both the backend and frontend services. If you prefer to run them separately:

   Terminal 1 (Backend):

   cd apps/api/services
   source venv/bin/activate  # or .\venv\Scripts\activate on Windows
   poetry shell
   uvicorn main:app --reload

   Terminal 2 (Frontend):

   cd apps/web
   bun run dev

Features

• AI-powered R programming assistance
• Interactive R code execution
• Real-time output and plotting
• Sandboxed environment for sharing files
• Shareable authentication links

Common Issues

1. Authentication Issues

   • Check terminal output for the correct authentication link
   • Make sure to use http:// in development (not https://)
   • You can set a custom token with AUTH_TOKEN=your_token in apps/api/services/.env
   • The token is saved in your browser for future sessions
   • Sharing the auth link with others will work as long as they use the same token

2. Docker Authentication

   • The authentication system works the same way in Docker
   • Use the link provided in the backend container logs
   • Container networking is automatically configured
   • Images and files are served correctly between containers

3. R Issues

   • Make sure R is installed: Run R --version in terminal
   • If R isn't found, reinstall from https://www.r-project.org/
   • If you must specify your own path, find it using which R

4. Python Issues

   • Make sure you have Python 3.9-3.12 installed (3.13+ not supported yet)
   • Windows: Download from https://www.python.org/downloads/
   • Mac: brew install python@3.12
   • Linux: Check your package manager for Python 3.9-3.12

5. Connection Failed

   • Check if backend is running (look for messages in terminal)
   • Make sure ports 3000 and 8000 are free
   • In Docker, check container logs with docker compose logs

Deployment

This project was never meant for public deployment, so certain features like filesystem interactions will not be available in production on Vercel, for example. You may be able to containerize the whole stack and serve the frontend with nginx; open a pull request if you want to give it a try.