Overview
LLM Plays Fire Emblem is an AI gameplay system for a strategy RPG. It reads emulator state from mGBA, gives the model tactical and story context, sends controller decisions back to the emulator, and streams telemetry to a React dashboard.
How It Works
- Emulator integration through Lua/socket state capture.
- LLM decision loops for tactical games with map state, unit context, screenshots, and commentary.
- A real-time dashboard for game state, decisions, action history, and chronicle-style interpretation.
- Product instinct for AI entertainment systems: the goal is compelling clips and proof-of-work, not 24/7 infrastructure before demand exists.
Hard Parts
The key challenge is converting a game designed for human visual play into structured context a model can use. The system has to know what the emulator sees, what the game state means, which actions are valid, and how to recover when execution differs from the plan.
Results
Current scope is curated demo runs, short clips, screenshots, and an architecture writeup. It is not a production livestream service yet.
