pokemon-player

Automate Pokemon gameplay by running a headless emulator and extracting RAM reads to infer game state. The AI observes position, HP, and inventory, then decides and acts without a visible UI. This enables fully automated battles and exploration across supported ROM variants.

AI-driven Pokemon play runs entirely headless, meaning no visible emulator UI is required. A live server and vision module interpret screen state from RAM reads, allowing the AI to observe, decide, and act autonomously while monitoring position, HP, and dialog.

RAM-state extraction supports multiple Pokemon generations and ROM variants, specifically Red, Blue, Yellow, and FireRed. The AI agent monitors position, HP, inventory, and dialog across these variants to enable informed, automated gameplay decisions.

Set up Pokemon AI agent gameplay automation by cloning the pokemon-agent repository, creating a Python 3.10+ virtual environment, and installing the package in editable mode. Run the server with the ROM directory specified to begin the headless gameplay loop.

Automating Pokemon battles with RAM reads does not require OCR. The vision module interprets screen state directly from memory extraction, monitoring HP, inventory, and position. This RAM-read approach provides precise game state inference for AI-driven play.

Headless Pokemon emulation for AI training requires a local ROM directory, Python 3.10+, and the pokemon-agent package. It relies on a live server and headless emulator, meaning any disruption to the server process or unsupported ROM variants will halt the automation loop.