TARS - Terminal Assistant with Gemini

This Python script, tars.py, is a terminal assistant that uses the Gemini API to interpret user commands and execute them in a Linux environment. It provides a limited set of allowed commands for safety and control.

How to Use

1. Set the API Key: Before running the script, you need to set the GEMINI_API_KEY environment variable with your Gemini API key. You can do this by running the following command in your terminal:

   export GEMINI_API_KEY="your_gemini_api_key"

   Replace "your_gemini_api_key" with your actual Gemini API key.

2. Run the Script: Execute the script using the following command:

   ./tars.py

3. Interact with the Assistant: The script will prompt you with >. You can then enter Linux commands, and the assistant will execute them using the execute_command function.

4. Exit the Assistant: Type exit to close the assistant.

Allowed Commands

The following Linux commands are allowed:

• ls: List directory contents
• cd: Change directory (only to /home/rp/Desktop or /home/rp/Documents)
• pwd: Print working directory
• cp: Copy files or directories
• mv: Move files or directories
• rm: Remove files or directories
• mkdir: Create directories
• find: Search for files in a directory
• grep: Search for patterns in files
• df: Display disk space usage
• du: Display file space usage

Important Notes

• The assistant uses the gemini-2.0-flash-exp model.
• The assistant will always use the full path when navigating to Desktop or Documents.
• The assistant will not explain the commands, it will execute them directly.
• The ls command will use -la when you want to show details.
• The assistant uses the current directory (.) if no path is specified.

Libraries Used

• os: Used for interacting with the operating system, specifically for accessing environment variables.
• google.generativeai: The core library for interacting with the Gemini API.
• subprocess: Used for executing Linux commands.
• json: Used for handling JSON data when interacting with the Gemini API's tool calling feature.

Tool Calling

The script uses the Gemini API's tool calling feature to execute Linux commands. The process_input function defines a function declaration for execute_command which specifies the parameters that the Gemini model can use. When the model determines that a command needs to be executed, it will return a function call with the appropriate parameters. The process_input function then extracts these parameters and calls the execute_command function to execute the command.

Code Structure

• execute_command(command, flags=None, args=None, requires_sudo=False): Executes a given Linux command with optional flags and arguments.
• process_input(user_input): Processes the user input using the Gemini API and executes the command if a valid command is detected.
• main(): The main function that starts the assistant and handles user input.