Available
Data
Open pairing dataset
Anonymised, opt-in pairing decisions and their reasoning, a public corpus for training and evaluation.
We are the founding contributor to the Open Gastronomy Standard, a machine-readable vocabulary for ingredients, dishes, beverages and pairings, and the OGS-Bench evaluation suite for gastronomy-capable AI.
The internet of taste should be interoperable. We publish standards, evaluations and reference implementations so recipes, menus, tasting notes and pairing engines can be exchanged and reasoned about by humans and machines alike.
ogs:<namespace>:<type>:<id>
Gastronomy AI is moving fast. Without shared definitions and shared evaluation, every team rebuilds the same primitives in incompatible ways. Open source is how the field grows up.
Shared vocabulary for ingredients, dishes, beverages and pairings, so any system can speak with any other system about food and wine.
Researchers, retailers, restaurants and platforms can contribute, audit and extend the standard together, not behind any single vendor.
Open evaluation, open prompts and open reference implementations let anyone verify quality claims and reproduce results.
A JSON-first, schema-validated, ontology-compatible vocabulary for the building blocks of gastronomy. Designed for both humans and machines to exchange and reason about.
Identifiers use ogs:<namespace>:<type>:<id>
Sensory profile, category, provenance.
Components, roles, cooking methods, resulting sensory profile.
Structural attributes, aromatics, typology.
Scored, explainable match between a dish and a beverage.
Culinary tradition with endonym-first naming and optional hierarchy.
A reproducible evaluation suite for gastronomy-capable AI. Twelve capability categories, deterministic metrics, scored on a 0–100 scale.
Beyond the standard itself, we publish supporting projects so others can build, audit and improve gastronomy AI without starting from scratch.
Anonymised, opt-in pairing decisions and their reasoning, a public corpus for training and evaluation.
Audited prompt templates for sommelier reasoning, dish characterisation and pairing explanations.
Translated controlled vocabularies for ingredients, cuisines and wine categories, across the kitchens we ship in.
Lightweight client libraries for parsing, validating and emitting OGS-conformant payloads.
Conversion between schema.org/Recipe, OGS dish, and common cookbook formats.
Reproducible runner for OGS-Bench, bring your model, get a score and a report.
OGS is governed in the open. Contributions, critiques and reference implementations from researchers, retailers, restaurants and platforms are all welcome.
Start with OGS v0.2, the spec, examples and rationale.
Guides, examples and adoption notes on opengastronomy.org.
Run your model through the eval harness and publish your score.
Talk to us about adoption, pilots or governance participation.