Skip to content

Latest commit

 

History

History
159 lines (107 loc) · 8.9 KB

README.md

File metadata and controls

159 lines (107 loc) · 8.9 KB
Raw

Gemini SH

Gemini SH: Your Personalized AI Agent for Enhanced Productivity

Gemini SH is a powerful Python library that leverages the capabilities of Google Gemini to create personalized AI agents capable of executing diverse functions on your computer, extending far beyond simple terminal interactions. Imagine having a versatile assistant that can understand your natural language commands and execute complex tasks, from managing files and automating workflows to generating content and even interacting with your codebase.

Watch the promotional video:

Watch the promotional video

A Dynamic, Agent-Based Approach

One of the core strengths of Gemini SH lies in its dynamic agent-based architecture. Depending on where you run the geminiSH command, you can create specialized agents tailored to specific tasks and environments. Here's how it works:

  • Agent Creation: When you execute the command within a directory, a .geminiSH folder is automatically created in that location. This folder serves as the agent's workspace, housing its configuration, history, and custom functions.
  • Agent Configuration: Each agent can be uniquely configured, allowing you to customize everything from the model used (e.g., gemini-1.5-pro-latest) to the maximum number of output tokens expected. This customization can even be done through natural language interactions with the model itself.
  • Function Expansion: Within the .geminiSH folder, you can create a functions directory to define custom Python functions that expand the agent's capabilities.
  • Specialized Agents: This dynamic structure empowers you to create specialized agents for different purposes. For instance, you can create an agent dedicated to code modification with a tailored system prompt and instructions, effectively building a coding assistant within your development environment.

Core Features

  • Interactive Chat: Engage in natural language conversations with the Gemini language model through your terminal.
  • Function Execution: Gemini SH can execute user-defined Python functions, extending its capabilities beyond basic chat interactions.
  • Cross-Platform Compatibility: Designed to run seamlessly on any terminal and operating system.
  • Turn-Based Chat System: Conversations are structured in turns, allowing for clear communication and back-and-forth interactions.
  • Customizable Functions: Extend Gemini SH's capabilities by adding Python scripts with functions defined in the functions directory.
  • Rich Function Descriptions: Use docstrings to provide detailed information about each function, including its purpose, parameters, return values, and when it should be executed.
  • Specialized Function Responses: Functions can return different types of responses, including:
    • Standard String Responses: Returned directly to the model as a text message.
    • response Object: An object containing a response field, which is passed to the model as a text message, and a response_to_agent field, which triggers specific actions within Gemini SH.
    • Actionable Responses: The response_to_agent field can contain instructions like:
      • files_to_upload: A list of file paths that will automatically trigger the upload_files function for each file.
      • require_execution_result: True: Indicates that the function's results should be immediately sent back to the model upon completion, bypassing user interaction.
  • System Instructions and Configuration:
    • prompts/system_instructions.md: Contains the initial instructions for the Gemini model, defining its role and behavior.
    • config.json: Configure various system settings, including the Gemini model to use and saving options.
  • Persistent Chat History: Conversations are saved in history.json for future reference and analysis.
  • Command-Line Function Execution: Execute functions directly from the command line by passing the function name and its arguments as arguments when running Gemini SH.
  • First-Time User Guidance: A helpful message explaining the system's functionalities and usage is displayed during initial runs.
  • Modular Managers: The codebase is structured around several managers that handle specific aspects of the system (config, state, input, output, chat, function, and model).

Use Cases

Gemini SH opens up a world of possibilities for interacting with your computer:

  • File Management: Organize and process files based on user requests.
  • Code Interaction: Get help with coding errors, generate code snippets, and even modify source files directly through diffs.
  • System Automation: Automate repetitive tasks and workflows through custom functions and Bash commands.
  • Content Creation: Generate presentations and other content based on text input.
  • Information Retrieval: Get answers to questions based on the content of local files and documentation.
  • Voice Interaction: Record audio and send it to Gemini, enabling hands-free communication.
  • Visual Context Awareness: Take screenshots to provide Gemini with visual context for more relevant responses.
  • Cross-Application Integration: Use the clipboard to exchange data with other applications.

How to Install and Use Gemini SH (Updated)

This guide will walk you through the steps to install and start using Gemini SH, your personalized AI agent for enhanced productivity.

Prerequisites:

  • Python 3.6 or higher: Gemini SH requires Python 3.6 or a later version to run. You can check your Python version by running python --version or python3 --version in your terminal.

  • PIP (Package Installer for Python): PIP is typically included with Python installations. You can verify its presence by running pip --version or pip3 --version.

Installation Steps:

You can install it directly from the cloned directory using:

pip install geminish

  1. Clone the Repository:

    You can clone the Gemini SH repository from GitHub using the following command in your terminal:

    git clone https://github.com/matias-casal/geminiSH.git

  2. Navigate to the Directory:

    After cloning, navigate to the geminiSH directory:

    cd geminiSH

  3. Install Gemini SH:

    Install Gemini SH and its dependencies using PIP:

    pip install .

  4. Set Your API Key:

    You need a Google Gemini API key to use Gemini SH. You can obtain one from the Google AI Studio. Once you have your API key, set it in the config.json file located in the .geminiSH directory:

    {
  "GOOGLE_API_KEY": "YOUR_API_KEY_HERE"
  // ... other configurations
}

    Alternatively, you can set the API key as an environment variable:

    export GOOGLE_API_KEY="YOUR_API_KEY_HERE"

Using Gemini SH:

  1. Start the Interactive Chat:

    To start the interactive chat session, run the following command in your terminal:

    geminiSH

  2. Interact with the Agent:

    You can now interact with the Gemini SH agent using natural language. You can ask questions, give commands, and even execute custom functions.

  3. Execute Functions Directly:

    You can also execute functions directly from the command line. For example, to start recording and give voice commands, execute the record function:

    geminiSH record

  4. Add Custom Functions:

    • Create Python scripts in the functions directory, defining your desired functions.
    • Use docstrings to provide clear and comprehensive descriptions of each function for the model to understand.

Contributing

Contributions to Gemini SH are welcome! You can contribute by:

  • Adding New Functions: Expand the system's capabilities by creating new Python functions for specific tasks.
  • Improving Documentation: Enhance the README, docstrings, and comments for better clarity and understanding.
  • Fixing Bugs and Issues: Help maintain and improve the codebase by addressing any reported issues.
  • Sharing Ideas and Use Cases: Contribute to the project's growth by sharing your ideas and potential applications.

Future Directions

  • Enhanced System Capabilities: Continuously improve the core functionalities to provide more robust and versatile interactions.
  • Advanced Prompt Engineering: Develop and refine prompts to elicit more accurate and contextually relevant responses from the AI.
  • User Experience Optimization: Focus on enhancing the user interface and experience, making interactions more intuitive and efficient.
  • Contextual Awareness: Implement techniques to provide the AI with better context from the user's environment, such as active applications and recent files.

Acknowledgements

This project is inspired by the power and flexibility of Google Gemini and the potential it holds for revolutionizing human-computer interaction.