Welcome to GPT_ALL: a cutting-edge AI assistant empowered by the capabilities of large language models (LLMs). This versatile assistant is adept at harvesting and interpreting real-time data from a plethora of sources, all while seamlessly meshing with the extensive functionalities of Python 3.12. At its heart lies OpenAI's GPT-4-1106 model, skillfully selecting functions and tools from a suite of specialized "experts" to optimize token efficiency. GPT_ALL boasts a dynamic command-line interface, embodying both power and flexibility. Engineered to harness the most recent advancements, tools, and expanded context windows available in the latest OpenAI API updates. The goal is clear and ambitious: to integrate chat GPT with an array of applications and devices, unlocking unprecedented possibilities. GPT_ALL's capabilities are further amplified by a range of built-in plugins, and you're encouraged to craft your own to broaden its diverse array of functionalities.
Latest Features:
-
AsyncOpenAI Integration:
- Asynchronous parallel function calling to allow the assistant to complete multiple tool calls in a single request.
- Leverage the power of GPT-4 for complex inquiries using tools and tool calls.
-
Conversation Memory Management:
- Can remember and reference previous inputs and responses in the same session only.
-
Conversation Flow:
- Managed flow of conversation by appending user input to memory and ensuring responses are within context limits.
-
Dynamic Function Invocation Based on Tool Responses:
- Handles dynamic invocation of functions based on tool call responses.
- Manages conversation state by appending messages from tool calls.
- Generates follow-up responses considering tool call results.
-
Environment Cleanup:
- Cleans up tools list after processing each request to help manage token usage.
-
Modular Plugin System:
- easily install new functions/tools to extend GPT_ALLs' abilities.
- Plugins load dynamically if enabled via the .env
-
Ensure you have conda installed on your system.
-
VScode Dev Container extension and Docker
is recommended for safety with system commands enabled.
Clone this repository to your local machine using:
git clone https://github.com/Eloquent-Algorithmics/GPT_ALL.gitNavigate into the project directory:
cd GPT_ALL
.\install.bat
To enable specific plugins or features, modify the .env file in your project directory according to your needs.
OpenAI API key is required all others optional.
Run GPT_ALL using:
conda activate GPT_ALL
python -m app
python -m web_app # To use the web interface.
or
python -m app --talk # To use TTS.
Please note some materials may not provide the best possible or the most optimal recommendations, solutions or source codes. Try to be open minded and take everything as a step in the learning process. If you encounter something to improve in the materials, please write your suggestions to the respected authors.
This version of GPT_ALL is totally experimental, USE IT AT YOUR OWN RISK. Even if a lot of tests have been performed on this version some things can be buggy or some lack of functionality.
This version can be driven by the community. If you want to help improve this version don't hesitate to submit a github issue or PR.
Contributions are what make the open-source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated. ## Contribution guidelines
If you want to contribute to GPT_ALL, be sure to review the contribution guidelines. This project adheres to code of conduct. By participating, you are expected to uphold this code.
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/AmazingFeature) - Commit your Changes (
git commit -m 'Add some AmazingFeature') - Push to the Branch (
git push origin feature/AmazingFeature) - Open a Pull Request