Dungeon Scribe

Home Page (Roleplay Notes)

AI Assistants & Automation Track

Inspiration

We're about to start a DnD campaign! One of us has played before, but the other is totally new to character creation and DnD in general. Building a character can already be a challenge, even if you have played before. Once character creation is done, there's still a whole campaign to play and roleplaying as your character is what makes the experience fun! It’s even harder when you’re not sure where to start or what kind of role you want to play. We want to help our friends and any other campaigners in creating their characters while also providing key points for their characters so they could roleplay with confidence and have more fun!

What it does

Dungeon Scribe is an AI-assistant that suggests key character traits and roleplay-driven tips to bring your DnD character to life. It’s built to spark ideas and give players, especially new ones, a strong foundation for storytelling. Dungeon Scribe will even generate sample dialogue to give you better reference to base your character's personality off of! Just type in what character you want to roleplay as!

How we built it

Our project was built with Python and uses AWS Bedrock to interact with Mistral Large 2 to act as an AI-assistant. We used specified prompts that would tailor the interactions with the bot to be related to DnD. We also configured endpoints on API Gateway that would run Lambda functions to handle user input to be sent to Mistral Large 2. The model returns easy to understand responses that break down personality traits, values, and behaviors, making it easier to get into roleplaying a character. Our frontend website was made with React and makes API requests using axios.

Challenges we ran into

Getting started with Bedrock had a bit of a learning curve, especially figuring out how to structure prompts, handle the chatbot responses, and connect everything smoothly to the frontend. Configuring endpoints and testing to make sure we set them up correctly was difficult too, especially having to constantly check logs while testing. Frontend isn't exactly our strong suit either, so there were a few bumps along the way, but we’re proud of what we pulled together!

Accomplishments that we're proud of

We got everything working by the end of it all! It was our first time using so many AWS services like Lambda and they were exciting to learn about. It was also our first time implementing a chatbot and we had fun experimenting with the available LLMs Bedrock provides. Neither of us had prior experience with these technologies but it was very rewarding learning how to implement them!

What we learned

We gained more experience with handling chatbots and making API request, as well as expanding our knowledge of using React. Learning how to use AI chatbots in our own projects always felt like a more difficult task than it actually was, so it was very cool being able to actually use them ourselves. It was also a new experience for us to use Lambda instead of hosting a server for our backend.

What's next for Dungeon Scribe

The idea of Dungeon Scribe was personally really fun for both of us, but the implementation needs a bit more polishing before we might be ready to deploy it. A simple step would be to transform the CLI of our Lore Builder into a website UI just like our Character Roleplay Notes Summarizer, but with chat bubbles like a messaging app. We would also like to implement RAG to retrieve and store relevant DnD information for our AI-assistant to improve and tailor information. Definitely something we'll revisit in the future and improve on.