Introducing ✨ GPTeam ✨ A completely customizable group of AI agents, with independent personalities, memories, and directives. We were inspired by the Stanford "Generative Agents" paper and decided to make an implementation that anyone could run. Github link below!

1. Run 2. Enter an openai api key 3. Customize config.json 4. And run it! Here's an example config:

Each agent runs in parallel on this loop. When it observes new events, it assigns an "importance score" and stores them in its memory.

When an agent is executing a plan, it gathers relevant memories according to this logic, and populates its prompt with them.

The agents also generate their own reflections, by thinking back on their memories and coming to higher level conclusions. It works like this:

This is just an experiment for us, but it was a lot of fun. I think this could serve as a template to folks who are making entertainment or gaming experiences with LLMs. We welcome forks and contributions!

Shout out to the rest of the team @alecvxyz, @joshsny, & @haniasnyder ❤️

UPDATE: custom logs UI!!

@fofrAI So cool! Excited to poke around, thanks for sharing.

@fofrAI Thanks Yohei! Its been fun to build 😄

wow this is great! prob costs so much to run though 😂

Thanks man, yea, and its super slow! that's how we know it's still early days for all this 👀