How should generated adjustment headers and lines be linked?

A generated relational package can separate `adjustment_header` from `adjustment_line`. The header could carry `adjustment_id`, `case_id`, `status` and `approval_ref`. Each line could carry `line_id`, `adjustment_id`, `account_ref`, `debit_amount`, `credit_amount`, `currency_code` and `source_allocation_ref`.

A separate relationship table can represent `reverses_line_id` and `replaces_line_id`. This avoids assuming that every original line maps to exactly one replacement. A correction may reverse one allocation and replace it with several allocations across departments.

Specify decimal precision and the representation of absent amounts. Null, zero and an omitted field should not acquire different meanings accidentally during conversion. Where a pack supports multiple currencies, define its balancing rules and any conversion fields instead of summing unlike currencies.

For Parquet delivery, publish table keys and relationship constraints alongside the schema. For JSONL delivery, state whether lines are nested within headers or emitted separately. Resellers can check round-trip conversion by reconstructing each adjustment’s line membership and relationship graph, ensuring that repackaging preserves the correction structure customers need.