Hi HN,

We’re Michael and Scott and we’re releasing an open-source, language-agnostic framework to create, debug, and serve Generative Agents.

This project has two main goals:

  - to abstract away the complexities of prompt-engineering detailed Agents and

elaborate Storylines with an easy to use no-code dashboard - using dashboard output, expose powerful functionalities - Agent Actions, Emotion Queries, Player Guardrails, etc. - in a simple small API & server

We have a demo game that we made (https://github.com/game-kings/detective), and a demo video (https://www.youtube.com/watch?v=VSqIzjOk5p4).

The demo game is a murder mystery set in Gold Rush era San Francisco. It includes a somewhat novel scoring mechanism at the end where your ability as a detective is scored based on cosine similarity between the embeddings of your explanation and ours.

We showed this to some friends of ours and, despite it frankly being an extremely hacked together game, they thought it surprisingly fun. This makes us really excited for what kinds of games are possible with tools like this.

We know there’s some similar prior art in this category but we think this still has a lot to offer.

The open-source work like GPTeam (https://github.com/101dotxyz/GPTeam) or ai-town (https://github.com/a16z-infra/ai-town) is mostly geared towards recreations of the Sims, and are not as general purpose as our framework.

There are commercial offerings like InworldAI, but I think our framework exists in a different space - we support Local AI out of the box, and hope this is the direction that gaming goes in. I don’t really want to start paying a subscription for every single-player game in the future. While we wait for chips to get cheaper and open-source models to get better, using GPT-3.5 or GPT-4 produces pretty similar if not better results than commercial offerings.

Of course, our project isn't perfect, isn’t feature-complete and definitely isn’t polish-complete. After showing it to our friends, we realized we need release early to get early feedback from others. We want to make this great and empower open-source LLM gaming in the future. We would love to hear your feedback.

We know for sure that some of you are much more creative storytellers than we are, and we can’t wait to see what you come up with!

Have fun!