AI Brainstore Python

This is a proof-of-concept of a brain for an AI agent heavily inspired by Mckay Wrigley's ai-brainstore project.

This is a python port of the original project and it relies on the same high-level approach. The main difference is that this project does not use langchain or chroma. Instead I chose to learn and implement the vector database using weviate, and borrowed some of the web-scraping and text summarization code from Auto-GPT.

How It Works

Ask the agent a question.

If it knows the answer, then it will recall it from memory.

If it doesn't know the answer, then it will browse the web and learn about it.

As it learns, it will save its memories to its brain.

Running Locally

1. Clone Repo

git clone https://github.com/Chryron/ai-brainstore-python

2. Install Dependencies

I use pipenv to manage dependencies, but you can use whatever you want.

Installing pipenv

pip install pipenv

Installing packages

cd ai-brainstore-python
pipenv install
pipenv shell

3. Configure Settings

Copy the .env.template file in the root of the repo and rename it to .env after you have filled in the values.

4. Set up your Weviate Database

This project uses a Weviate instance without authorization, hosted on Weviate Cloud Services (WCS) on the free tier. The WCS quick-start guide should help you get started.

Note: The endpoint in your .env file should be the endpoint URL of your WCS instance.

5. Run Script

python main.py

6. Use It

You should now be able to interact with the agent via the terminal.