Pantry

Pantry finds and stores food-product nutrition. It searches the local Coles and AFCD shards plus a user's own records, and can acquire one product from a Coles or Woolworths URL, USDA FoodData Central, or Open Food Facts.

pantry --json search "greek yoghurt"
pantry --json search "greek yoghurt" --remote
pantry --json lookup coles 1047
pantry --json lookup coles 1047 --grams 42 | recipes edit "Snack" --input -
pantry add https://www.coles.com.au/product/example-1047
pantry add usda:2476857
pantry add off:0123456789012
pantry delete manual sourdough

Nutrients describe the record's grams, which every record and result carries and which is 100 unless --grams N asks for another weight. There is no pack size or serving size. A per-100 mL panel is read as per 100 g and says so in basis_note. An item is a flat record in the shared item format, so it pipes unchanged into Recipes, or into whatever tool records what you ate.

Energy is kcal and every other nutrient is grams. A panel printing kilojoules is converted where it is read, so no record holds a kj.

Manual input is flat JSON, restated to 100 g before it is stored:

printf '%s\n' '{"grams":90,"kcal":335,"protein":45.6,"fat":7.9,"carbs":4.9}' |
  pantry add --input - --id sourdough --name Sourdough

Unknown standard nutrients are null; zero is returned only when the source explicitly reported zero.

pantry delete SOURCE ID removes a record from that store, and only from there: a shipped shard row is refused, and deleting a correction leaves the row it shadowed visible again.

The ignored data/coles.jsonl scrape is irreplaceable. No command writes to package data. Acquired products go to $XDG_CONFIG_HOME/pantry or ~/.config/pantry.

Retailer requests are deliberately limited and paced. A block stops the run; there is no retry, proxy rotation, or CAPTCHA handling.

Development

uv run pytest -q
uv run --project importers/afcd pytest -q importers/afcd/tests
uv run ruff check src
uv run ruff format --check src