Inspiration

When updates are pushed to games they are mostly pushed at once and the realease is done at the same time. What if we could release the updates later and even target certain users. What if we can have multiple different agents already trained and switch to them when you want. What if we can store your converstaion in a knowledge base and the nps can remember and use it later.

What it does

Our npcs can talk to a player based on a already provided instructions. Since it uses llms it can be more interactive with the user. Now we use launch darkly to provide the developers with a way to change the models when they want to. It should probably be used when an they want to introduce a new event and the npcs are changed to use more fine tuned for the event.

How we built it

Currently we have a chat to replicate that functionality and we are using launch darkly flags to change the prompt and add additional info.

Challenges we ran into

So we were planning to use an agent and provide specific information but we ran into issues with security plus understanding how to read from the buffer and hence we had to leave it and use a much more simple approach.

Accomplishments that we're proud of

We are proud of everything that we were able to do given the short time and since it was the first time we used these technologies

What we learned

Launch darkly and aws bedrock are amazing and both of them can be used to create and solve new use cases.

What's next for Dynamic Npc.

Use it in a real game

Share this project:

Updates