AutoGen experiments

This project is a learning exercise on learning LLM agents using AutoGen.

Examples of items to explore:

• How to use agents to solve a task.
• What are the limitations? For example, do we need GPT-4, or could a simpler (and cheaper) model do the job?
• How many tokens do we consume for the agents and the behind-the-scenes processing (the selection and coordination of the agents)?

How to use this project

If you haven't done so yet, prepare the environment. If you have already prepared the environment, activate it with source venv/bin/activate.

Each notebook is a self-contained experiment. You can run them in any order.

You may need to select a Python kernel the first time you open a notebook. Choose the kernel from the virtual environment you created.

• Sequential chat: Given a starting and ending number, select math operators (agents) to transform the starting number into the ending number.

Preparing the environment

OpenAI key

Create a file named .env in the root of the project with the following content:

OPENAI_API_KEY=your-openai-api-key

This file is in the .gitignore. It will never be committed to the repository.

Python environment

This is a one-time step. If you have already done this, just activate the virtual environment with source venv/bin/activate.

Run the following commands to create a virtual environment and install the required packages.

python3 -m venv venv
source venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt