/P.A.L.

A toolkit to build AI Agents that proactively interact w/ the world

Primary LanguagePythonMIT LicenseMIT

P.A.L - Proactive Agent Library

P.A.L is an easily customizable toolkit to build AI Agents that proactively interact with the world.

🚀 Motivation

According to Sam Altman, co-founder and CEO of OpenAI, proactive AI systems are the future, pushing useful information to us, instead of today's reactive systems from which information is pulled. We agree with this belief, and have taken a significant step forward in this direction. We have built P.A.L, a powerful open-source toolkit to build Proactive AI Systems with today's best Generative AI models.

🛠️ Key Features

  • Easy-to-use - With P.A.L, developers can flexibly build proactive agents that can track various signals by implementing custom plugins. We built an agent with 5 widely varying capabilities leveraging P.A.L in just 36 hours!
  • Multimodal and model agnostic - Developers can plug and play with any SOTA Generative AI model(s) across all modalities.
  • Task agnostic - Developers can build for any use-case ranging from a personal agent to an enterprise-grade virtual shopping assistant.

💻 Quickstart

To start using P.A.L, clone this repository, and install the required environment libraries

git clone https://github.com/harshsikka/ProactiveAgent.git
cd ProactiveAgent

After creating and activating a virtual environment

pip install -r requirements.txt

To use the agent built using P.A.L

python example.py

🤔 What does P.A.L do?

Here is a video showcasing the capabilities of a personal agent for an AI startup founder, that we built within 36 hours:

'Check'