Purpose
Extract the raw order: customer, PO reference, lines exactly as written (customer SKU, qty, UOM, price).
When to use
Start of every order run.
Inputs
Email/PDF/portal order document; customer identification. Emits the entry hop (schema in contracts/).
Steps
- Extract order_ref, customer id, lines with line_no, customer_sku VERBATIM, qty, uom, unit_price, promo_ref if quoted.
- Nothing is corrected at ingest - a wrong-looking alias or price must reach the validator as written.
- Validate against the contract; write order.json.
Output
order.json - the entry hop.
Grounding requirements
Every line cites the source document position.
Constraints
- No resolution of aliases, prices or UOMs here (#1-#3 are the engine's).
Escalation / uncertainty
Unreadable lines -> listed, order held (#6). Never proceed on guessed lines.