Hi folks, we’re a small group of AI / AI Infra enthusiastic building AI + Gaming applications.
This tweet is trailer of our stage 1 in a very contained environment, where a simple puzzle is generated based on agent narrative on the fly, and obviously the dialog is LLM generated in real-time as well.
We’re polishing stage 2-4 featuring player <> agent interaction as well as multi-agent interactions and will post more updates on our accounts, looking forward to get more inputs from the community (feedbacks, user inputs, feature requests, etc.) 
5 Likes
Nice! Looks great.
What model are you using on the backend?
Thanks ! We tried from mpt, llama v1, llama v2 to gpt series. This particular video uses gpt-3.5
1 Like
Nice. I’ve been doing work on NPC and dynamic quest generation. I’m also trying to stick with gpt-35-turbo to keep the cost down, but it’s not as easy as GPT-4, for sure!
Welcome to the forum. Hope you stick around. We’ve got a great group of people here…
This is an interesting area with many emerging use cases for sure, glad there’re people exploring it too 
I think depends on the nature of application, in our case, multi-agents tend to be much more token intensive that plain chatbots, that calls for low cost APIs / self hosted LLM. So far using GPT-4 also yields best result but clearly not affordable beyond a cool tweet or blog post.
I will post more stuff down the road here for sure, and looking forward to discuss more with the community!
Wow, good job, Amazing, good luck .
1 Like