Apppublic

korigamik/kitchen-ops-env

sourceHugging Faceupdated 5mo agoView on Hugging Face

0likes

App README

Kitchen Ops Env

OpenEnv environment for restaurant order fulfillment with inventory quantities, prep expiry, substitutions, rush restocking, and deadline-based grading.

Tasks

Task ID	Difficulty	Summary
`breakfast_omelette`	easy	Single fast breakfast ticket
`salad_opening`	easy	One cold-station salad with a short freshness window
`lunch_combo`	medium	Salad and wrap share produce and deadlines
`wrap_restock_window`	medium	Two wraps force a tortilla restock decision
`dinner_rush_stockout`	hard	Three concurrent orders with a paneer shortage
`double_shortage_service`	hard	Four-ticket rush with paneer and oil pressure

Action Space

KitchenAction supports:

—PREP_COMPONENT
—COOK_DISH
—ASSEMBLE_DISH
—SERVE_ORDER
—RESTOCK_INGREDIENT
—SUBSTITUTE_INGREDIENT
—CHECK_PROGRESS

Observation Space

KitchenObservation exposes:

—briefing
—available_actions
—metadata.last_action_error

Reward / Grading

—Step rewards stay in 0.0-1.0
—Episode scores stay in 0.0-1.0
—Grading combines completion, timeliness, progress, cost, waste, execution efficiency, and substitution quality
—Idle/no-progress play is heavily penalized
—Tasks are generated from reusable templates instead of fixed full-scenario blobs

Setup

bash

uv python install 3.11
uv sync --managed-python --python 3.11 --extra dev

Run the server:

bash

uv run --managed-python --python 3.11 python -m uvicorn server.app:app --host 0.0.0.0 --port 8000

Run tests:

bash

uv run --managed-python --python 3.11 python -m pytest

Run the baseline:

bash

export API_BASE_URL=https://models.github.ai/inference
export MODEL_NAME=openai/gpt-4o
export HF_TOKEN=...
uv run --managed-python --python 3.11 python inference.py

Environment Variables

—API_BASE_URL
—MODEL_NAME
—HF_TOKEN
—KITCHEN_ENV_URL optional, defaults to http://localhost:8000

API

—POST /reset
—POST /step
—GET /state
—GET /tasks
—GET /grade
—GET /health
—GET /metadata