HaLLMark

About the game

HaLLMark is a story-based role-playing game in which the story is generated by Large Language Models (LLM). Take the role of a king of a fantasy kingdom and decide the fate of your subjects and the whole kingdom.

Choose your name, the name of your kingdom, and your advisor, and start ruling! Be a good king, care about your subjects, stay in power for 5 days and enjoy a happy ending. Or be a cruel king, terrorize your kingdom, get dethroned, and find out what the fate of your kingdom will be.

All the dialogues and the story of the game are generated by language models. The game needs a stable internet connection to communicate with the language models. You will also need the API keys of at least one supported language model (ChatGPT, Mistral, Claude, Command R) and the API key of the vector database used - Pinecone. Command R language model and Pinecone vector database are free to use.

What you need to do to run the game:

1. Sign up at https://www.pinecone.io/ and get your API key for your free starter plan.
2. Get the API key from your favourite language model supported by the game (ChatGPT, Mistral, Claude, Command R). If you'd like to use the Claude model, you'll also have to get the API key from their embedding model service - Voyage.
3. Copy the API keys to the HaLLMark_Data/StreamingAssets/api-keys.json file to their appropriate places. Make sure you have enough funds on accounts associated with the API keys (applicable for ChatGPT, Mistral, and Claude).
4. Launch the game and enjoy!

The art (character sprites and the throneroom background) was created by amazing Aliciel.