Reference data from catalogs

Some values must be real: a country code your system validates, a currency with its correct minor units, the list of EU members. Catalogs hold that reference data. Use it inside a recipe, or download a list directly.

A real value per record

A catalog field picks one record per row. value chooses the column; columns takes several columns of the same record, so a country's name, code and currency always agree:

clientes.yaml
version: "1"
output: { language: spa, format: csv, quantity: 4 }
fields:
  - { name: cliente, type: text, category: fullName }
  - name: pais
    type: country
    columns: [name, iso2, currency, calling_code]
    parent: SA            # South America only
ghostbakery-output.csv
cliente,pais_name,pais_iso2,pais_currency,pais_calling_code
Elena Márquez,Islas Malvinas,FK,FKP,500
Andrea Lozano,Guyana,GY,GYD,592
Noelia Castro,Chile,CL,CLP,56
Elena Mendoza,Bolivia,BO,BOB,591

columns expands into flat fields named <field>_<column>, and names are localized to output.language. Filter with parent (a continent) or group (G20, EU, MERCOSUR, BRICS, OECD); add unique: true for no repeats.

The whole list

With all: true the field lists every matching record once, and the list size becomes the quantity:

eu.yaml
version: "1"
output: { language: fra, format: csv }
fields:
  - { name: pays, type: country, columns: [name, iso3], group: EU, all: true, order: code }
ghostbakery-output.csv
pays_name,pays_iso3
Autriche,AUT
Belgique,BEL
Bulgarie,BGR
…

Only one field per recipe can set all: true. Other fields in the recipe are generated once per listed record, so you can attach synthetic data to every country.

Direct exports

When you need the reference list itself — to seed a lookup table or to paste into a prompt — skip the recipe. In the app, Catalogs lists every dataset with its size, source and license: pick a language, columns, an optional filter and a format (json, csv, or toon, the most compact for LLM prompts), then download. Exports cost 1 credit per row.

The same through the API, with a key that has the catalog:read scope:

terminal
curl -s https://api.ghostbakery.com/api/v1/catalog/country/export \
  -H "Authorization: Bearer $GB_KEY" \
  -H "Content-Type: application/json" \
  -d '{"language":"por","columns":["name","iso2"],"format":"csv"}' \
  -o paises.csv

POST /api/v1/catalog/{type}/estimate with the same body returns the row count and cost first. limit caps the rows, and sample: true with a seed returns a reproducible random subset.

Things to know

  • Reference data is maintained live. The same recipe can produce different output after an update, for example when a country's population figure changes.
  • A language is offered only when its data is complete. The catalog page lists the output languages and each dataset's coverage.
  • Sources and licenses are listed on the catalog page and in the terms.