autogpt-package

Introduction to the AutoGPT Package

AutoGPT, brought to life by the Kurtosis platform, ensures a smooth, installation-free experience with a single command execution reminiscent of a 'brew install' for AutoGPT. This package is a user-friendly avenue for running AutoGPT, either in-browser or locally on your machine, making it accessible for everyone from beginners to experts.

Running AutoGPT in the Browser

You can easily run AutoGPT directly in your browser without any installation steps:

1. Obtain an OpenAI API key here.
2. Click on the provided Gitpod link and select "Continue" to initiate the default setup.
3. Wait for the system to set up Gitpod and install Kurtosis, which takes about 30 seconds.
4. Replace YOUR_API_KEY_HERE with your OpenAI API key and run the command:

   kurtosis run github.com/kurtosis-tech/autogpt-package --enclave autogpt '{"OPENAI_API_KEY": "YOUR_API_KEY_HERE"}'
   

5. Once the installation completes, open the AutoGPT prompt with:

   kurtosis service shell autogpt autogpt --exec "python -m autogpt"
   

6. Enjoy exploring AutoGPT functionalities!

Running AutoGPT Locally

For users preferring to run AutoGPT on their own system:

1. Obtain an OpenAI API key.
2. Follow these instructions to install Kurtosis.
3. Use the terminal to run this command by substituting YOUR_API_KEY_HERE with your actual API key:

   kurtosis run github.com/kurtosis-tech/autogpt-package --enclave autogpt '{"OPENAI_API_KEY": "YOUR_API_KEY_HERE"}'
   

4. Once AutoGPT is ready, launch its prompt via:

   kurtosis service shell autogpt autogpt
   

   And within the prompt:

   > python -m autogpt
   

5. Experiment with AutoGPT and close the instance with:

   kurtosis enclave rm -f autogpt
   

Configuring AutoGPT

The AutoGPT Package supports extensive configurations. Simply include configuration values as properties in the JSON object when invoking Kurtosis. For example, configure AutoGPT to not restrict to a workspace by executing:

kurtosis run github.com/kurtosis-tech/autogpt-package --enclave autogpt '{"OPENAI_API_KEY": "YOUR_API_KEY_HERE", "RESTRICT_TO_WORKSPACE": "False"}'

The default 'local' backend can be changed by setting the MEMORY_BACKEND parameter. However, note that Redis is not functional with version 0.4.0 at the moment.

Leveraging AutoGPT Plugins

With Kurtosis, you can enable a variety of plugins to enhance AutoGPT's capabilities. For instance, to activate the AutoGPTTwitter plugin, use:

kurtosis run github.com/kurtosis-tech/autogpt-package --enclave autogpt '{"OPENAI_API_KEY": "YOUR_API_KEY_HERE", "ALLOWLISTED_PLUGINS": "AutoGPTTwitter"}'

To activate multiple plugins simultaneously, separate plugin names with commas:

kurtosis run github.com/kurtosis-tech/autogpt-package --enclave autogpt '{"OPENAI_API_KEY": "YOUR_API_KEY_HERE", "ALLOWLISTED_PLUGINS": "AutoGPTTwitter,AutoGPTEmailPlugin"}'

A variety of first-party and third-party plugins is supported, ranging from social media integrations to advanced analytics tools.

Running AutoGPT without OpenAI

Understandably, OpenAI can be costly or impractical for some; thus, the AutoGPT Package provides an option to operate using the GPT4All model locally powered by LocalAI. To initiate:

kurtosis run github.com/kurtosis-tech/autogpt-package '{"GPT_4ALL": true}'

Modify the model with the MODEL_URL parameter to use alternate models.

Development and Feedback

For development purposes, clone the repository and deploy directly on your local setup by following:

kurtosis run . --enclave autogpt '{"OPENAI_API_KEY": "YOUR_API_KEY_HERE"}'

To facilitate development, Kurtosis offers a Visual Studio Code extension with features like syntax highlighting and auto-completion for its associated Starlark scripts.

Feedback and queries can be directed to their Discord or Twitter @KurtosisTech. For bugs or feature suggestions, opening a GitHub issue is encouraged.