JSON Extraction Prompt

Extract structured data from the document below.

Extract exactly this JSON structure:
{
  "[field]": "string. [WHAT IT IS + FORMAT + EXAMPLE, e.g. 'invoice number as printed, e.g. INV-2024-0031']",
  "[date_field]": "string (YYYY-MM-DD) or null. [WHICH date this is; null if not stated]",
  "[enum_field]": "one of: [\"option_a\", \"option_b\", \"unknown\"]. [WHEN EACH APPLIES; use \"unknown\" when the document doesn't say]",
  "[amount_field]": "number or null. [UNIT AND SIGN CONVENTION, e.g. 'total in USD, negative for credits']",
  "[items]": [
    {
      "evidence": "string, the EXACT sentence or cell from the document this item comes from, quoted verbatim",
      "[item_field]": "string. [WHAT]"
    }
  ]
}

Extraction rules:
- Extract, don't infer: a value must be stated in the document, not deduced from vibes. If the document says "due in 30 days" and no invoice date exists, due_date is null, not today plus 30.
- Empty is a valid answer: return [] for lists with no items and null for absent fields. Never invent an entry to fill an array.
- The evidence field comes FIRST in each item and must be a verbatim quote. If you can't quote it, the item doesn't exist.
- Normalize formats exactly as the schema states (dates to YYYY-MM-DD, amounts to plain numbers without currency symbols); do not normalize anything else.
- Ambiguity: if two document values could fill one field, pick the more specific and append the alternative to an "extraction_notes" array at the root.
- Output raw JSON only, no markdown fences, no commentary.

Document:
"""
[PASTE THE TEXT / EMAIL / INVOICE / TRANSCRIPT]
"""