Which fields and keys link schedule records together in a delivered release?

A release can be laid out as linked tables so a loader can join them without guesswork. The exact fields are agreed in scoping; this is a typical shape.

**Tables and keys.**

- `agreements`: `agreement_id`, `party_id`, `product`, `start_date`, `total_amount`, `currency` - `schedules`: `schedule_id`, `agreement_id`, `version`, `status`, `supersedes_schedule_id`, `effective_from` - `instalments`: `instalment_id`, `schedule_id`, `due_date`, `amount`, `status`, `paid_amount`, `paid_date` - `adjustments`: `adjustment_id`, `schedule_id`, `type`, `reason_code`, `approval_id`, `policy_ref` - `approvals`: `approval_id`, `approver_role`, `decision`, `decided_at` - `mandates` and `notices`, each carrying `schedule_id` and `schedule_version`

**Rules for links.** Every foreign key resolves, a superseded schedule keeps its rows, and an adjustment always names the policy version it relied on.

**Formats.** JSONL for episodes and tasks, with nested state; Parquet for flat tables; a JSON Schema file per table so validation is automatic.

**Checks to run on receipt.** Orphan keys, instalment sums against `total_amount`, date ordering, and version chains with no gaps.

All values are generated, so run your own disclosure and grounding review on the sample before you ingest it.