A GPT-based system that generates and executes high-level plans, using specialized "agents" that augment LLMs with better reasoning and external information sources
-
Sign up / get your OpenAI API key: https://platform.openai.com/signup
-
Create the env file:
cp .env.example .env
, put your API key where specified -
Install deps:
pip3 install -r requirements.txt
-
Run:
python3 src/main.py
-
"Techniques to improve reliability" (https://github.com/openai/openai-cookbook/blob/main/techniques_to_improve_reliability.md) from the openai-cookbook and many of its citations, especially CoT prompting (https://ai.googleblog.com/2022/05/language-models-perform-reasoning-via.html)
-
The concept of "agents" that iteratively prompt an llm was taken from Langchain (https://github.com/hwchase17/langchain)
-
The executive agent is an implementation of the system described in HuggingGPT (https://arxiv.org/pdf/2303.17580.pdf) and its planning step prompt was forked from JARVIS (https://github.com/microsoft/JARVIS)
-
General inspiration from Andrew Mayne (https://andrewmayneblog.wordpress.com/)
-
Project named by gpt3.5