⚠️ DO NOT USE THIS FRAMEWORK TO CHEAT ⚠️

This project is for educational and experimental purposes only. Using this framework to gain an unfair advantage in GeoGuessr or any other game violates its terms of service and goes against the spirit of fair play.

Autonomous framework using browser automation to allow large language models to play GeoGuessr.

1. pip install -r requirements.txt
2. Add GEMINI_API_KEY to .env
3. Change EDGE_USER_DATA_DIR and MSEDGEDRIVER_PATH as necessary in config.py (replace USERNAME)
4. python bot.py

1. Once a duel has been started, the loop begins by detecting relevant UI elements and either storing their value or removing them from display (e.g. HP bars).
2. A screenshot is taken and immediately given as a prompt, alongside the text of the system prompt, to the configured LLM.
3. The guess given back by the LLM is then parsed.

This suite currently supports No Moving, Panning or Zooming only, with eventual support for No Moving. Gemini 2.0 Flash, Gemini 2.5 Flash, and Gemini 2.5 Pro are supported. The nature of duels makes Gemini 2.5 Pro a poor choice despite its ability, due to its latency.