Note
For a continuation of this proof of concept, see Rogue-Bench.
Make your own animations like this with VHS.
The goal of this repo is to play Rogue with LLMs.
It uses a modified version of the Rogue Collection that can be found here.
Warning
This code has only been tested on WSL2 Ubuntu 24.04.
First, clone the repo recursively to pull in the custom version of Rogue Collection.
git clone --recursive https://github.com/iwhalen/rogomatic-llm.git
cd rogomatic-llmTo get Rogue running, run the following:
make install
make build-rogue
make run-rogueThis should open a window where you can play Rogue!
First, create a .env with your API key in it:
cp .env.sample .envBy default, this works with Sonnet-4.6.
uv run rogomatic-llmOnce running, you should see output like the .gif above.
Originally made for testing, you can also play from your terminal with "human" mode.
uv run rogomatic-llm --player humanYou should see something like this, which means you're ready to play.
╭───────────────────────────────────── Rogue ──────────────────────────────────────╮
│ │
│ │
│ │
│ │
│ │
│ │
│ │
│ │
│ --------------------+---- │
│ |...............@.......+ │
│ +/......................| │
│ ------------+------------ │
│ │
│ │
│ │
│ │
│ │
│ │
│ │
│ │
│ │
│ │
│ │
│ Level: 1 Gold: 0 Hp: 12(12) Str: 16(16) Arm: 4 Exp: 1/0 │
╰──────────────────────────────────────────────────────────────────────────────────╯This is no different from regular old Rogue from the GUI (i.e., with make run-rogue).
To exit, send a CTRL+C signal.
You won't be able to save, resize, or do anything else. So, if you only want to play Rogue, run make run-rogue instead.
For more options in the CLI, run uv run rogomatic-llm --help:
Usage: rogomatic-llm [OPTIONS]
Main typer application. Starts the play session with the given options.
╭─ Options ───────────────────────────────────────────────────────────────────────╮
│ --player [human|llm] Type of player. │
│ [default: llm] │
│ --rogue-path PATH Path to the rogue │
│ executable. │
│ [default: │
│ rogue-collection/build/r… │
│ --rogue-version [unix rogue 3.6.3|unix Rogue version to play. │
│ rogue 5.2.1|unix rogue [default: Unix Rogue │
│ 5.3|unix rogue 5.4.2] 5.4.2] │
│ --model TEXT PydanticAI compatible │
│ Agent model string. │
│ [default: │
│ anthropic:claude-sonnet-… │
│ --max-history INTEGER Number of recent │
│ action/result pairs to │
│ retain in AI context. │
│ [default: 25] │
│ --action-delay FLOAT Seconds to wait between │
│ actions in LLM mode. │
│ [default: 0.5] │
│ --install-completion Install completion for │
│ the current shell. │
│ --show-completion Show completion for the │
│ current shell, to copy it │
│ or customize the │
│ installation. │
│ --help Show this message and │
│ exit. │
╰─────────────────────────────────────────────────────────────────────────────────╯
The dev setup for this repo is quite minimal. The two commands worth running are:
make lint
make testThe test suite is quite minimal. Not much is actually tested.