TypeGPT

June 29, 2024 update: Added support for Claude and Llama3 models, along with improved image handling capabilities!

TypeGPT is a Python application that allows you to invoke various AI's and LLM's from any text field in your operating system. Whether you're in a chat app, document, or code editor, you can seamlessly interact with ChatGPT, Google Gemini, Claude, or Llama3 with Ollama, using keyboard shortcuts.

video.mp4

Features

Global Accessibility: Invoke AI models from any text input field across your system.
Multiple AI Models: Support for ChatGPT, Google Gemini, Claude, and Llama3.
Keyboard Shortcuts: Use simple keyboard shortcuts to communicate with AI models.
Clipboard Integration: Utilize the clipboard for larger text inputs and image pasting.
Screenshot Capability: Capture and include screenshots in your queries.

Prerequisites

Before you can run the application, ensure you have the following installed:

Python 3.x
pynput package
requests package
pyperclip package
google.generativeai package
anthropic package
Pillow (PIL) package

You also need to have API keys for the AI services you plan to use. You can get yours at:

ChatGPT: https://openai.com/api/
Google Gemini: https://ai.google.dev/aistudio
Claude: https://www.anthropic.com/
Llama3: Ensure you have Ollama installed and running locally (http://localhost:11434)

Installation

Open your terminal/shell, and enter the following commands.

Clone the repository: git clone https://github.com/olyaiy/TypeGPT.git
Navigate into the project directory: cd TypeGPT
Give execute permission, and run the install.sh file:

on mac chmod +x install.sh

chmod +x run.sh

./install.sh

Running the program:

Run using the provided executable (will allow the program to run in the background)

./run.sh

To manage the program, use: ps aux | grep TypeGPT.py

and in order to stop the program from running, grab the PID you're shown when you run the command above (a 5 digit number associated with the process), and type the following: kill -9 12812

here our PID is 12812 for example.

Usage

Use the following keyboard shortcuts in ANY TEXT FIELD ( search bars, google docs, code editors, etc) to interact with the application:

/ask: Follow this command with your prompt and press Cmd+Shift+Enter to send it. You can also paste the clipboard content using Cmmd+V. Anything you paste will be included in your prompt.
/see: Captures a screenshot to include with your next query. Press Cmd+Shift+Enter to send the query with the screenshot. You can also type on the same line
/stop: Stop listening. Everytime you type /ask or /see the program starts listening for Cmmd+shift+Enter to send a prompt to the selected AI model. When you type /stop, the program will stop listening. This is incase you start a prompt but change your mind.
/chatgpt: Switch to ChatGPT model.
/gemini: Switch to Google Gemini model.
/claude: Switch to Claude model.
/llama3: Switch to Llama3 model.
/check: Check which model is currently active.
/quit: To quit the program.
Shift + Cmd + Enter: Send the text to the selected AI model when in listening mode.

Ensure you set your API keys by modifying the keys.txt file in the project directory. Update this file with your actual API keys:

OPENAI_API_KEY=your-openai-key-here
GEMINI_API_KEY=your-gemini-key-here
ANTHROPIC_API_KEY=your-anthropic-key-here

This file is critical for the application to function correctly, so be sure to update it before running the program.

Configuration

Modify the system_prompt.txt file to customize the behavior and responses of your AI based on your needs. You can also change the versions of the AI models in the api_calls.py file. Currently, the defaults are:

ChatGPT: gpt-4-turbo
Gemini: gemini-1.0-pro-vision-latest
Claude: claude-3-5-sonnet-20240620
Llama3: Uses the local Ollama instance

Contributing

Contributions are very welcome! Please fork the repository and submit pull requests with your proposed changes.

Future Plans

We plan on adding support for more AI models and improving the user interface. If you have any further ideas, we'd love to hear them!

License

Distributed under the Apache 2.0 License. See LICENSE for more information.

olyaiy / TypeGPT