Loading…
Four tools over the local FDC/OFF/GPC bulk copies and the curated GPC category-matching tables — read-only, computed live. Not operational health (see /status for that).
Schema, rows, and per-column coverage for the local data stores.
Loading…
Dataset freshness, per-field completeness, value distributions, and GPC category-matching coverage.
Loading…
Loading…
Two separate measurements, not one combined figure — they cover different things. See the GPC Mappings tab for the full curated-entry lists.
Loading…
Non-null % per nutrient in each local mirror — which fields are usable as ML features, and which source is the better one for a given nutrient.
Loading…
For GTINs present in both local mirrors, how often FDC and OFF agree on a nutrient value (within 15% of the larger figure). Sorted worst-agreement first — the fields most worth double-checking before trusting either source alone.
Loading…
Binned over the 1st–99th percentile range so a single garbage outlier can't collapse the histogram — clipped counts are reported below each chart, not hidden. Computed on request (a few seconds the first time for a column, then cached).
Pick a column above to load its distribution for both sources.
Browse the GS1 Global Product Classification: Segment → Family → Class → Brick → Attributes. Search reaches into attribute values too — try olive oil.
Every hand-verified entry linking an FDC branded_food_category to a GPC brick or class,
and how much of the real local FDC corpus the curated tables reach right now.
Loading…
Loading mappings…
reviewed tier
Open Food Facts category tags that have been individually hand-verified against a GPC brick or class,
and how much of the real local OFF corpus they reach. Trusted at the same level as the FDC mappings above — a
reviewed match is not a best-effort guess, unlike the fuzzy matcher's off_fuzzy
fallback.
Loading…
Loading mappings…