What records and relationships can a schedule release contain?

A release can be built around the full life of a schedule rather than its first version. At the centre sits a schedule record: the parties, the product, the frequency, the start and end dates and the governing policy.

Around it a dataset can include:

- **Instalment lines** with due dates, amounts, splits between principal, interest and fees, and status - **Payment events** that match, partly match or miss each line - **Amendment requests** with the requester role, reason code and effective date - **Decisions** from an approver, with a reference to the policy clause applied - **Side effects** such as notices, arrears flags, fee postings and reconciliation tasks

Policies can be expressed both as readable text and as structured thresholds. That pairing matters for training, because a model can be tested on whether it applies a limit as written or only imitates past outcomes.

A task pack could add instructions, starting state and expected end state. A simulated environment could expose the same state through tool actions, so an agent can read a schedule, propose a change and see the consequences.

Everything is generated. Relationships are designed to be internally consistent, and buyers should still inspect grounding and disclosure risk before use. For a reseller, the useful point is that the same underlying world can be cut into a dataset, a task pack or an environment listing.