Package Manager for AI Agent tools
Project description
AutoPack Tools
AutoPack is a repository for tools that are specifically designed for AI Agents.
The autopack
Python package is designed to facilitate the installation and usage of tools hosted in the AutoPack
repository. Tools in AutoPack are called Packs.
Note
This is still in the alpha stage. It's roughly at MVP level, and things will not work, features aren't complete, and things will change. Be forewarned.
Installation
Install the autopack
package from PyPI using pip:
pip install autopack
Or Poetry:
poetry add autopack
Usage
Pack IDs
Each pack in the AutoPack repository is identified by a fully qualified path based on its GitHub repository. This format ensures uniqueness, prevents namespace collisions, and allows for easy identification of the source code location. Importantly, it enables us to uniquely refer to a pack while keeping pack names intuitive and understandable for an LLM.
For example, the ID of a pack named web_search
hosted in the GitHub repository erik-megarad/my_packs
would be:
erik-megarad/my_packs/web_search
This format allows us to use the pack's name, web_search
, within an Agent, making it convenient and straightforward
to reference the desired tool.
Manual Tool Installation
You can manually install a pack using the following command:
autopack install author/repo_name/tool_name
Using with LangChain
To use a tool with LangChain, you can retrieve it using the get_pack
function from the autopack
module:
from autopack import get_pack
tool = get_pack("author/repo_name/pack_name")
# Add the tool to the 'packs' argument when instantiating your AgentExecutor
agent_executor = AgentExecutor.from_agent_and_tools(agent=agent, tools=[tool()])
Using with Auto-GPT
We are actively working on improving the integration with Auto-GPT. Stay tuned for updates!
To use autopack
with Auto-GPT, you can edit the file autogpt/main.py
and add autopack
to the COMMAND_CATEGORIES
list.
TODOs
This project is still in its early stages, and there are several features and enhancements that need to be implemented. If you are interested and willing to contribute, the following list provides a good starting point:
- Tool search functionality within the CLI
- Tool search capability from Python
- Optional automatic pack installation in the
get_pack
function - Tools for Agents to independently search for, install, and utilize other Tools
- Tool for utilizing the pack selection API of the AutoPack repository
- Optional contribution of feedback back to the AutoPack repository
Development
For information on how to contribute to AutoPack, please refer to the CONTRIBUTING.md file.
Feel free to modify this README to provide more specific details about your project and its functionalities.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for autopack_tools-0.1.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 9437f7ff08bd4244a9f4539141211486efa578be5425a37d57f706de1f868167 |
|
MD5 | 5ac039ef3d283e30aae0a5d98037582d |
|
BLAKE2b-256 | 3fffe53bfc6990381bf9765da40484054414028e8654b4e7754fbf19471a60eb |