Cheeseboard ML

A tiny Python framework for arranging reusable LLM prompts into "boards."

This is a demo project. It is not a real product, and the code is intentionally small.

The idea

A cheese board is a curated selection of cheeses, arranged so you can pick what fits the moment. Cheeseboard ML applies the same idea to prompts.

Each prompt template is a Wedge, named after a cheese and given a flavor and a set of tags. A Board holds a collection of wedges and lets you serve() or compose() a subset of them by tag.

Install

pip install -e .

Quickstart

from cheeseboard import Board, Wedge

board = Board("weeknight")
board.add(Wedge(
    name="brie",
    flavor="mild",
    template="You are a warm, patient explainer.",
    tags=["explain"],
))
board.add(Wedge(
    name="cheddar",
    flavor="sharp",
    template="You are a direct, no-nonsense critic.",
    tags=["review"],
))

print(board.compose(tag="explain"))

See examples/quickstart.py for a runnable version.

Starter board

Cheeseboard ships with four default wedges in cheeseboard.presets.default_board().

Wedge	Flavor	Tags	Personality
brie	mild	explain, teaching	Warm, patient, step by step
cheddar	sharp	review, critique	Direct, points out flaws first
gouda	mild	summarize	Calm, condenses to three bullets
blue	bold	debate, critique	Contrarian, argues the other side

CLI

cheeseboard --list
cheeseboard --tag explain
cheeseboard --list --pairing

The --pairing flag suggests a beverage for each wedge. It has no effect on prompt output and exists purely for fun.

Testing

pip install -e .
pip install pytest
pytest

Project status

Cheeseboard ML is a small demo built to illustrate a simple pattern for organizing prompts. It has no external dependencies and no network calls. Contributions and forks are welcome, but treat it as a toy rather than production infrastructure.

License

MIT. See LICENSE.