mirror of
https://github.com/run-llama/agent-maze.git
synced 2026-06-29 10:37:45 -04:00
main
Agent Maze Playground
This repo is a playground for testing agent capabilities with complex, procedurally generated mazes.
Usage
- Install dependencies in a fresh virtual environment
uv venv
uv pip install .
- Run the attached notebook to test the agent capabilities.
In the notebook, you can change
- the types of mazes generated
- the LLMs and modeles that are being tested
- the prompts
- the tools each agent has access to
- the metrics being tracked
Details
Inside procedural_maze_generator.py you can learn more about how the mazes are generated.
Languages
Jupyter Notebook
85.5%
Python
14.5%