Document class schema requirements

Document object

This object describes the document type and the fields to extract from a document and has these properties:

• document_type (required, string): The document type or category.
• document_description (required, string): Description of the document.
• additional_prompt_instructions (optional, string): Extra instructions for extraction that can be applied to an entire document page. These instructions complement the field-level descriptions. With this property, you can provide additional guidance to improve extraction accuracy, for example, validation rules or formatting requirements.
• fields (required, array): The list of fields to extract from the document. These fields are defined as DocumentField objects.

DocumentField objects

These objects define the fields to extract:

• name (required, string): The name of the field. That name is referenced in the target table.
• description (required, string): A short explanation of what the field represents. The description provides context that helps the model validate and focus on the correct information in the document.
• extraction_instructions (optional, string): Additional instructions for extracting this specific field such as "remove the zip code from the address".
• examples (optional, array of strings, 1 or more items): Examples of possible field content to improve accuracy.
• available_options (optional, array of strings): Valid values for the field. If the field isn't directly available in a document, but can be deduced from the context or visual cues, one of the listed values should be returned. For example, provide a list of currencies, so that Euro can be returned if various parts of an invoice contain a € sign, but nowhere is stated that Euro is currency of the invoice. Specify available_options in addition to any examples.
• affix (optional, boolean): Determines whether the field is a prefix, infix, or suffix that is attached to another element. For example, a currency sign can be an affix.
• fields (optional, array): Nested subfields that contain field sets for logically grouped fields within a document. For example, an invoice contains line items, such as multiple entries for purchased products, that are grouped in a table or a list. These subfields are also defined as DocumentField objects.

TargetTables object

This object defines the database table where the extracted data will be stored and the format in which the data is stored, and specifies the Python transformations functions to use for converting the data to the required format. This object can occur multiple times: once for the main table and additional ones for any field sets (fields objects) that are defined for the Document object.

• name (required, string): The table name.
• description (required, string): The table description.
• metadata (optional, object): Additional metadata.
• columns (required, array): The column definitions. Each column is defined as a Column object.

Column object

Each object specifies how to derive the column value and how and transform it from string to the data type that you want.

• name (required, string): The column name.
• type (required, string): The data type of the columns, which can be one of these types: binary, boolean, date, decimal, fixed, long, string, timestamp, uuid
• description (optional, string): The column description.
• source (required, Source object): The source of the column value is defined as a Source object.

Source object

This object can be defined as follows:

• variable (optional, string): As a variable reference such as document id.
• field (optional, array): As a reference field in the extracted data as defined in the Document section.
• transform (optional, SourceTransform): As a data transformation. A SourceTransform object has these properties:
  • transform_name (required, string): The name of the Python transformation function to apply to the extracted data. See Transformation functions.
  • arguments (required, array): The name of the argument and the path to the value in the Document section.
• literal (optional, string): As a literal or constant value.

document_class_schema_version

This entry defines the schema version of the document class. The schema version of a document class that you import must match the version that is currently used in the product.

Transformation functions

You can use these transformation functions:

• currency_to_numeric
• make_date_uniform
• to_number
• weight_to_numeric

The required syntax is as follows:

"source": {
  "transform": {
    "arguments": [
      {
        "name": "<transform argument>",
        "value": {
          "field": [
            "<name of the source field>"
          ]
        }
      }
    ],
    "transform_name": "<name of the transform>"
  }
},

"source": {
  "transform": {
    "arguments": [
      {
        "name": "amount",
        "value": {
          "field": [
            "<name of the source field>"
          ]
        }
      },
      {
        "name": "locale",
        "value": "id_ID"
      }
    ],
    "transform_name": "currency_to_numeric"
  }
},

"source": {
  "transform": {
    "arguments": [
      {
        "name": "date_str",
        "value": {
          "field": [
            "<name of the source field>"
          ]
        }
      }
    ],
    "transform_name": "make_date_uniform"
  }
},

"source": {
  "transform": {
    "arguments": [
      {
        "name": "number_str",
        "value": {
          "field": [
            "<name of the source field>"
          ]
        }
      }
    ],
    "transform_name": "to_number"
  }
},

"source": {
  "transform": {
    "arguments": [
      {
        "name": "weight",
        "value": {
          "field": [
            "<name of the source field>"
          ]
        }
      }
    ],
    "transform_name": "weight_to_numeric"
  }
},