Custom Document

The starting point for any document type that doesn't fit one of our pre-built templates. Define your own fields from scratch.

When to use

Use Custom Document when:

Your document type has no standard layout (e.g. internal forms, client-specific spreadsheets).
You need to extract just a handful of specific fields and don't want every default field a pre-built template returns.
You're prototyping and want to feel out what's extractable before building a production schema.

Defining fields

In the dashboard wizard:

Click New Extraction → Custom Document.
Add fields one row at a time. For each field:
- Key — short, lowercase, snake_case (becomes the JSON key).
- Type — string / number / date / boolean / object / list.
- Description — what the field represents.
- Example (optional) — a real value the field might hold; this dramatically improves accuracy.
For object or list<object> types, click the chevron to define the nested fields.

Example schema — Real estate listing

A minimal custom schema for parsing real estate brochures:

json

{
  "name": "Real estate listing",
  "fields": [
    { "key": "address",       "type": "string",   "example": "742 Evergreen Terrace, Springfield" },
    { "key": "list_price",    "type": "number",   "example": 425000 },
    { "key": "bedrooms",      "type": "integer",  "example": 4 },
    { "key": "bathrooms",     "type": "number",   "example": 2.5 },
    { "key": "square_feet",   "type": "integer",  "example": 2400 },
    { "key": "agent_contact", "type": "object", "fields": [
      { "key": "name",  "type": "string" },
      { "key": "phone", "type": "string" },
      { "key": "email", "type": "string" }
    ]}
  ]
}

Tips

Start small. Schemas with 5-10 well-defined fields outperform schemas with 30 vaguely-defined ones.
Use the playground. Upload one document and iterate the schema until you're happy before committing to a batch upload.
Add examples for any tricky field. A single concrete example often beats 3 sentences of description.