How can generated handover records be linked without relying on free-text matching?

A relational release could contain `cases`, `handover_requests`, `handover_items`, `dependencies` and `record_links`. Each handover request can carry `handover_id`, `case_id`, `from_team_id`, `to_team_id` and `status`. Item records can identify the individual obligations being transferred.

Typed links make the relationships operationally meaningful. A `supports` link points to evidence, a `blocks` link identifies an unresolved prerequisite, and a `supersedes` link identifies an earlier record replaced by a later one. Treating every link as a generic attachment would discard those differences.

Parquet can suit customers joining large record collections, while JSONL can package nested handover bundles for task ingestion. The formats should share identifiers and documented null semantics. An absent dependency record must not be confused with a dependency explicitly marked resolved.

A reseller can request machine-readable schemas and referential-integrity checks alongside the generated files. Documentation should define whether linked records must exist within the licensed package or may be represented by declared stubs. That distinction lets downstream customers validate completeness before building training pipelines or presenting records to an agent.