🤖 Free Pilot for Vim/Neovim (BETA)

A fast, lightweight, and configurable AI completion plugin that works with both local and cloud models. Get GitHub Copilot-like functionality for free or at a fraction of the cost! Built as an experiment to see what was possible with local ollama and remote openrouter AI models.

FYI: If you are are looking for a coding assistant plugin, check out: https://github.com/whatever555/vim-code-checker

✨ Features

• 🚀 Real-time AI-powered code completion
• 🏠 Support for local models via [Ollama](https://ollama.com]
• ☁️ Cloud model support via OpenRouter
• ⚡ Minimal latency, maximum productivity
• 🎯 Filetype-specific enabling/disabling
• 🎨 Fully customizable behavior

🎬 Demo

rec1.mov

💡 Why Free Pilot?

• Free/Low-Cost Alternative: Use local models or pay-as-you-go cloud services instead of expensive subscriptions
• Privacy-Focused: Run everything locally with Ollama
• Flexible: Choose between local and cloud models based on your needs
• Lightweight: Minimal impact on editor performance
• Customizable: Configure exactly how and when you want AI assistance

🚀 Getting Started

Prerequisites

Option 1: Local Setup (Free!)

1. Install Ollama
2. Pull a coding-focused model:

ollama pull codellama:7b
# or for better results:
ollama pull codellama:13b

Option 2: Cloud Setup (Pay-as-you-go)

1. Create an account at OpenRouter
2. Get your API key

Installation

Using Vundle:

Plugin 'whatever555/free-pilot-vim'

Using vim-plug:

Plug 'whatever555/free-pilot-vim'

Using packer.nvim:

use 'whatever555/free-pilot-vim'

🛠️ Configuration Options

General Settings

" How long to wait before triggering completion (in ms)
let g:free_pilot_debounce_delay = 500

" Maximum number of suggestions to show
let g:free_pilot_max_suggestions = 3

" Enable debug logging
let g:free_pilot_debug = 0

" Choose backend: 'ollama' or 'openrouter'
let g:free_pilot_backend = 'ollama'

" AI temperature (0.0 - 1.0, lower = more focused)
let g:free_pilot_temperature = 0.1

" Debug log file location (empty = no logging)
let g:free_pilot_log_file = ''

" Maximum tokens to generate
let g:free_pilot_max_tokens = 120

Ollama Settings

" Model to use with Ollama
let g:free_pilot_ollama_model = 'codellama:13b'

" Ollama API endpoint
let g:free_pilot_ollama_url = 'http://localhost:11434/api/generate'

OpenRouter Settings

" Your OpenRouter API key
let g:free_pilot_openrouter_api_key = 'your-api-key'

" Preferred model
let g:free_pilot_openrouter_model = 'anthropic/claude-2:1'

" Your site URL for OpenRouter analytics
let g:free_pilot_openrouter_site_url = 'https://github.com/whatever555/free-pilot-vim'

" Your site name for OpenRouter analytics
let g:free_pilot_openrouter_site_name = 'FreePilot.vim'

Behavior Settings

" Enable on startup
let g:free_pilot_autostart = 1

" Only enable for specific filetypes (empty = all)
let g:free_pilot_include_filetypes = []

" Disable for specific filetypes
let g:free_pilot_exclude_filetypes = ['help', 'netrw', 'NvimTree', 'TelescopePrompt', 
    \ 'fugitive', 'gitcommit', 'quickfix', 'prompt']

💰 Cost Comparison

Service	Cost	Notes
GitHub Copilot	$10/month	Fixed subscription
Free Pilot (Ollama)	$0	Free, runs locally
Free Pilot (OpenRouter)	~$0.01-0.10/1000 tokens	Pay for what you use

🚦 Usage

1. Start typing code as normal
2. Watch as AI suggestions appear
3. Press Tab to accept a suggestion
4. Press Ctrl-] to skip a suggestion

Commands

• :FreePilotEnable - Enable completion
• :FreePilotDisable - Disable completion
• :FreePilotToggle - Toggle completion
• :FreePilotStatus - Check current status

📝 A Note About Local Models

While freePilot offers completely free local AI completion through Ollama, it's important to set realistic expectations:

Current Limitations

• Local models (like CodeLlama) running on consumer hardware may be:
  • Slower than cloud solutions
  • Less accurate in their suggestions
  • More memory-intensive
  • Limited in context understanding

This is not a limitation of freePilot itself, but rather the current state of running large language models locally.

The Future Looks Bright

• Smaller, more efficient models are being developed
• Local model performance is improving quickly
• Hardware acceleration is getting better

Recommended Approach

• Start with local models to test the waters
• If you need more reliable completion, consider using the OpenRouter backend
• Keep updating your Ollama models as new versions are released
• Consider this an investment in the future of local AI tools

🤝 Contributing

Contributions are welcome! Feel free to:

• Report bugs
• Suggest features
• Submit pull requests

🐛 Troubleshooting

Common Issues

1. Completion not showing up?

   • Check if the backend is running (ollama ps for local)
   • Verify API key for OpenRouter
   • Check :FreePilotStatus

2. Slow completion?

   • For Ollama: Try a smaller model
   • For OpenRouter: Check your internet connection
   • Adjust g:free_pilot_debounce_delay

3. Wrong completions?

   • Try a different model
   • Check if the correct filetype is detected

📝 License

MIT License - see LICENSE file for details

🙏 Acknowledgments

• Ollama team for making local AI accessible
• OpenRouter for providing affordable cloud AI
• The Vim/Neovim community

---

Made with ❤️ by the Free Pilot team

Note: This is not affiliated with GitHub Copilot or OpenAI