Shell AI Assistant

A command-line tool that integrates AI assistance directly into your terminal. It reads your tmux pane content and provides contextually aware command suggestions using various AI models through litellm.

Features

• Reads your tmux pane content and sends it to your choice of AI model
• Automatically parses commands from AI responses and puts them into your prompt
• Optional Stack Exchange integration to provide additional context from relevant stack exchange google search.
• Terrible and poorly thought out features like auto-execution and recursive mode for automated system destruction.

Prerequisites

• Python 3
• TMux
• litellm
• BeautifulSoup4 (for Stack Exchange integration)
• requests

Installation

1. Install required Python packages:

   pip install shellai-tmux

2. Set up your API key for your chosen provider as an environment variable: The tool looks for API keys in environment variables based on the chosen model:

   # Example API key setup
   export OPENROUTER_API_KEY="your-key-here"
   export ANTHROPIC_API_KEY="your-key-here"
   export GEMINI_API_KEY="your-key-here"
   export TOGETHER_API_KEY="your-key-here"

   • And any others supported by litellm

Usage

Basic usage:

shellai [options] [input]

Options

• -A, --auto: Automatically execute the suggested command (use with caution)
• -r, --recursive: Add ;shellai to the end of suggested commands for continuous operation
• -m MODEL, --model MODEL: Choose AI model (can use number 0-5 to select from model list)
• -q, --quiet: Only output the command without explanation
• -v, --verbose: Enable verbose mode with detailed output
• --debug: Run in debug mode (skips API request)
• -t TARGET, --target TARGET: Specify target TMux pane (default: current pane)
• --log FILE: Log all output to specified file
• --log-commands FILE: Log only commands to specified file
• --file FILE: Read additional input from specified file
• -S LINES, --scrollback LINES: Number of scrollback lines to include from tmux (default: 0)
• --system-prompt FILE: Use custom system prompt from file
• --delay SECONDS: Set delay before auto-execution (default: 2.0 seconds)
• -c, --add-stackexchange: Add relevant context from Stack Exchange
• -M MODEL, --model-stackexchange MODEL: Specify model for Stack Exchange search query generation (default: gemini/gemini-1.5-flash-latest)

Examples

1. Basic command suggestion based on visible terminal content:

   shellai

2. Get a suggestion for a specific task:

   shellai how to find large files

3. Use a specific model by number or name:

   shellai -m 3 how do I automate these commands
   # or
   shellai -m anthropic/claude-3-5-sonnet-latest how do I automate these commands

4. Include Stack Exchange context with custom model:

   shellai -c -M gemini/gemini-1.5-pro-latest how to compress images in bulk

5. Auto-execute commands with auto and recursive mode(or don't):

   shellai -A -r process these files  # DO NOT DO THIS

6. Include more context from terminal history:

   shellai -S 100 why won't this compile

Security Considerations

• Review commands before execution: Always review suggested commands before running them
• Auto-execution risks: The -A flag will execute commands without confirmation
• Data privacy: Be mindful that terminal content is sent to AI providers
• API credentials: Secure your API keys and avoid exposing them in scripts or logs
• Recursive mode: Use -r flag with extreme caution as it can create command loops

Troubleshooting

• Enable verbose mode (-v) for detailed operation information
• Check API key environment variables if model requests fail

Contributing

Contributions are welcome! Please feel free to submit issues and enhancement requests.

License

GPL 3