​
Assets Type

  • assets: Use column_type = `assets` to specify a column that your documents will be uploaded into. Reference the sign_in_sheet example above:
Important: Every transformation requires at least one operation with transform_type = 'parse'. This operation must have column_type = 'assets'.
{
  "column_name": "sign_in_sheet",
  "column_type": "assets",
  "transform_type": "parse",
  "task_description": "N/A"
}

​
Primitive Types

These types represent fundamental data values.
  • text: A plain text value.
  • string: A synonym for text.
  • boolean: A true or false value.
  • number: A general numeric type that can represent both integers and decimals.
  • time: Represents a time value (e.g., 14:30:00).
  • date: Represents a date value (e.g., 2025-03-03).

​
List/Object Types

These types store structured or array-like data.
  • list: Use column_type = 'list' when you want to extract a list of “primitive” data from your assets or other columns. For example, if you wanted to extract a list of registered names from a sign-in sheet, you could specify the following transformation_params object:
"transform_params": {
  "model": "trellis-vertix",
  "mode": "document",
  "operations": [
    {
      "column_name": "sign_in_sheet",
      "column_type": "assets",
      "transform_type": "parse",
      "task_description": "N/A"
    },
    {
      "column_name": "members",
      "column_type": "list",
      "transform_type": "extraction",
      "task_description": "Extract the members who signed up (reference {{sign_in_sheet}}", 
      "operations": [
          {
            "column_name": "member_name",
            "column_type": "text",
            "transform_type": "extraction",
            "task_description": "The name of the member who signed up."
          }
      ]
    }
  ]
}
Important: If column_type is list, then operations should only be of length 1 since list can only be of one type.
  • object: Use column_type = 'object' when you want to extract a list of “complex” data from your assets or other columns. For example, if you wanted to extract an object for each registered member from a sign-in sheet, you could specify the following transformation_params object:
"transform_params": {
  "model": "trellis-vertix",
  "mode": "document",
  "operations": [
    {
      "column_name": "sign_in_sheet",
      "column_type": "assets",
      "transform_type": "parse",
      "task_description": "N/A"
    },
    {
      "column_name": "members",
      "column_type": "list",
      "transform_type": "extraction",
      "task_description": "Extract the members who signed up (reference {{sign_in_sheet}}",
      "operations": [
        {
          "column_name": "member",
          "column_type": "object",
          "transform_type": "extraction",
          "task_description": "A member object",
          "operations": [
            {
              "column_name": "member_name",
              "column_type": "text",
              "transform_type": "extraction",
              "task_description": "The member's name"
            },
            {
              "column_name": "member_phone_number",
              "column_type": "number",
              "transform_type": "extraction",
              "task_description": "The member's phone number"
            }
          ]
        }
      ]
    }
  ]
}
For operations with column_type = `object`, you can include as many operations as required, and even nest them!