vindeklaration

A public dataset of the additives declared in wines sold by Systembolaget, and a bilingual site on top of it at vindeklaration.se.

Since 8 December 2023 the EU requires wine to declare its ingredients and nutritional values. Systembolaget shows that declaration on each product page, but it cannot be searched, filtered or compared — you can look up one wine at a time and nothing more. This project reads every wine's declaration, turns the free text into structured data, and publishes the result.

What is in here

File	Contents
data/wines.json	The dataset: one record per wine, with parsed additives. Not in git — download wines.json.gz, mirrored on the latest release. Monthly snapshots live in their own releases.
data/wines.sqlite	The same rows, for querying — built by src.build from wines.json, not committed
data/additives.yaml	The substance dictionary — names, E-numbers, aliases
data/unknown.json	Text the parser could not identify, ranked by impact
data/producer-declarations.json	A second source: declarations found at the producer's own e-label where Systembolaget carries none. Kept separate and never merged into the dataset
src/site.py, templates/	The site — a static build, uv run python -m src.site
worker/index.js, cf-build.sh	What serves the dataset at /data/*, and what Cloudflare's build runs

Caveats worth reading before using this

Most wines declare nothing yet. The requirement turns on when the wine was produced: wine produced before 8 December 2023 may be sold under the old rules until stocks run out. "The 2024 harvest onwards" is the Commission's own shorthand for this, and it is not the criterion — a wine produced in, say, January 2024, or a non-vintage sparkling wine whose second fermentation happened later, is covered without carrying a 2024 vintage.

The dataset has no production date, and Systembolaget publishes none, so vintage is used as a stand-in throughout. The error runs one way: every wine of vintage 2024 or later was necessarily produced after the cutoff, so nothing is counted as covered when it is not. But 33.2% of the catalogue carries vintage 2023 or none at all, and an unknown share of that sits inside the requirement while a vintage filter cannot see it. Any figure here computed "over the wines the requirement covers" means the wines it certainly covers.

20.4% of the assortment currently carries a declaration, rising to 66% among wines of vintage 2024 or later. Coverage improves on its own as stock rotates. See docs/legal-notes.md §1f for the sources.

The text is supplier-entered. Each importer types the declaration by hand, so spelling, terminology and completeness vary. raw_ingredients is kept on every record so any claim here can be checked against the source.

Counting is a judgement call, and here is the one made. additive_count counts declared additives, sulfites included. Grapes, must and added sugar are raw materials and are not counted. Bottling gases (nitrogen, CO₂, argon) are recorded separately in gases and not counted, because they are a packaging step rather than something that stays in the wine. A wine with text the parser could not read is marked parse_status: "partial" and is excluded from any ranking — omitting a wine is better than understating its additives.

Fewer additives does not mean healthier. This is a record of what is on the label, not a health assessment. Wine is alcohol; that is the relevant risk.

The label wins. The bottle in your hand is more current than this dataset.

Record shape

{
  "product_number": "253108",
  "name": "Adobe Sauvignon Blanc",
  "vintage": "2025",
  "country": "Chile",
  "price": 249.0,
  "declaration_status": "declared",
  "parse_status": "complete",
  "additive_count": 2,
  "additives": [
    {"id": "sulfites", "e_number": "E220-E228", "category": "preservative",
     "name": {"sv": "Sulfiter", "en": "Sulfites"}, "allergen": "sulfites"}
  ],
  "gases": [{"id": "carbon_dioxide", "e_number": "E290"}],
  "base_ingredients": [{"id": "grapes"}],
  "processing_notes": [{"id": "protective_atmosphere"}],
  "allergens": ["sulfites"],
  "nutrition": {"kcal_per_100ml": 73, "sugar_g_per_100ml": 0.2},
  "raw_ingredients": "Druvor*, koldioxid, konserveringsmedel (SULFITER), ...",
  "source_url": "https://www.systembolaget.se/produkt/vin/x-253108/"
}

declaration_status is declared or not_declared; parse_status is complete, partial or not_declared. Undeclared wines are kept in the dataset — leaving them out would hide how much of the assortment says nothing.

以上为 README 的开头部分。其余内容请阅读原文。 查看完整 README ↗