AGI-Llama: Bringing Modern LLMs to 1980s Sierra Adventure Games

A project that integrates Large Language Models (LLMs) into the classic NAGI (New Adventure Game Interpreter) to transform how users interact with retro Sierra titles like Space Quest, King's Quest, and Leisure Suit Larry.

Details

AGI-Llama is an experimental research project that aims to bring the power of modern Large Language Models (LLMs) to classic 1980s Sierra adventure games. By integrating LLM technology into the NAGI (New Adventure Game Interpreter) engine, the project allows users to interact with these retro titles using natural language input, rather than the traditional 'verb noun' syntax. Additionally, the LLM layer enables playing the games in any language the model supports, expanding accessibility. The project also features a modern tech stack, including SDL3 with GPU acceleration and Unicode support. AGI-Llama supports various backend options, including local inference using llama.cpp for models like Llama 3, Qwen, and Gemma, as well as cloud APIs from providers like OpenAI, Hugging Face, and Groq. The LLM logic is encapsulated in a library, potentially allowing integration into other projects like ScummV.