Extraction schema reference

ParseHawk accepts a focused JSON Schema Draft 2020-12 authoring dialect. The dialect is intentionally smaller than general JSON Schema so an extractor can turn the same contract into model guidance, output validation, and typed downstream data.

Download the canonical meta-schema or validate a file with parsehawk schemas validate schema.json.

Supported shapes

Shape	Supported authoring fields
Object	type, properties, required, additionalProperties, title, description
Array	type, items, title, description
String	type, pattern, minLength, maxLength, title, description
Number, integer, boolean	type, title, description
Enum	type, enum, optional x-parsehawk, title, description
Enum union	oneOf or anyOf branches containing const or enum
Nullable value	A two-item type union such as ["string", "null"]

The root must be an object with type and properties. Object schemas must set additionalProperties to false. Unknown schema keywords are rejected.

Minimal schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "invoice_number": {
      "type": ["string", "null"],
      "description": "The invoice number exactly as printed."
    },
    "total": {
      "type": ["number", "null"],
      "description": "The final amount due."
    }
  },
  "required": ["invoice_number", "total"],
  "additionalProperties": false
}

Semantic hints

Set x-parsehawk.semantic on a string field to express the value’s meaning more precisely than JSON Schema’s primitive type. These hints guide compatible models; the JSON output still uses ordinary strings, numbers, integers, or booleans.

Family	Values
Text	string, verbatim-string
Scalar	integer, number, boolean
Date and time	date, time, date-time, duration
Locale	country, currency, language, language-tag, script
Identifiers	url, email-address, phone-number, iban, bic, unit-code
Regions	region:US, region:FR, region:IE, region:GB, region:IT, region:ES, region:DE, region:PT, region:CA, region:MX, region:BR, region:AU, region:JP, region:KR

{
  "type": ["string", "null"],
  "description": "Invoice date in ISO 8601 format.",
  "x-parsehawk": { "semantic": "date" }
}

Source of truth

The canonical meta-schema is committed at docs/schemas/parsehawk-extraction-schema.schema.json. The generated page and download endpoint are checked for drift in pre-commit and CI.