Open Food Facts

Open Food Facts is a collaborative, open database of food products, built the way Wikipedia is: volunteers photograph labels and type in what they say, and the result belongs to everyone. It has grown to cover several million products across more than a hundred countries, and it is run by a small non-profit rather than a company — the data is free because people chose to give it away, and it stays free because a handful of staff and a lot of contributors keep it that way.

Support them. If Buttery's shopping list is useful to you, some of the credit belongs upstream. Donate to Open Food Facts — they run on donations, and their data does real work in this app every time you add a recipe to your list.

Links

Database	world.openfoodfacts.org
Donate	Support the project
Data & licensing	Exports, dumps, and terms
Developer docs	openfoodfacts.github.io
Wiki	wiki.openfoodfacts.org
Taxonomies	taxonomies/ on GitHub

What they publish

Most of Open Food Facts is product data — barcodes, brands, ingredient lists, nutrition facts, allergens, packaging, Nutri-Score and NOVA classifications, and label photographs. You can reach it through a JSON API for single lookups, or as bulk exports (MongoDB dumps, JSONL, CSV, Parquet) for anything larger.

Alongside the products they maintain a set of hierarchical taxonomies: controlled vocabularies for ingredients, additives, allergens, categories, labels and more, each one a tree of canonical terms with synonyms and translations.

What Buttery uses

Just one file: taxonomies/food/ingredients.txt — roughly 5,600 ingredient entries, about 4,700 of which carry English names. None of the product data, no nutrition, no barcodes.

We use it to answer a question the recipe itself cannot: what food is this line about, and which aisle is it in? At build time we generate a lexicon from the taxonomy, hand-assign aisles to around 170 nodes, and let every other food inherit from its nearest mapped ancestor. That is how "chicken breast" ends up under Meat & seafood without anyone typing it. Their canonical ids also give us a stable identity for a food, which is what lets "1 lb chicken breast" from one recipe and "8 oz chicken breast" from another become a single line on your list.

This all happens when Buttery is built, not when you use it. The running app never calls Open Food Facts.

The shape of the data

The taxonomy is a plain text file of blocks separated by blank lines. One entry looks like this:

< en: chicken meat
en: chicken breast, chicken breast meat
fr: blanc de poulet, blancs de poulet, poitrine de poulet
de: Hähnchenbrust, Hähnchenbrustfleisch
ciqual_proxy_food_name:en: Chicken, breast, without skin, raw

That is the whole format:

• < en: chicken meat — the parent, which is where the hierarchy comes from. An entry can have more than one.
• en: chicken breast, chicken breast meat — names in a language, canonical first, synonyms after. The first language line mints the entry's id, here en:chicken-breast.
• Other language lines add translations in 15+ languages.
• ciqual_proxy_food_name:en: and friends are properties, not names — links to other food databases, carbon-footprint figures, allergen flags. The colon inside the key is what tells a property apart from a language.

Licensing

Open Food Facts data is published under the Open Database License (ODbL) 1.0, with individual contents under the Database Contents License and product images under CC BY-SA 3.0.

Buttery's generated lexicon is a derived database under the ODbL: it reuses their identifiers, English names and hierarchy, and adds our own aisle assignments on top. It is offered under the ODbL in turn, and it ships with a notice recording the exact source revision it was generated from. Open Food Facts is credited on Buttery's acknowledgements page.