Provenance manifest schema
One JSON record per delivered item: where it came from, under which licence, and which checks it passed.
Every delivery includes manifest.jsonl: one JSON record per item. It ties each record to the supplier agreement it came from and the licence it's delivered under, so chain of title holds per record.
"record_id": "qr_01HZX4T9K2M7",Quarry's ID for this item"dataset_id": "QD-7K2P",The listing it belongs to"owner_ref": "own_5TQ9",Pseudonymous owner; the real name is in your licence"licence_id": "lic_2026_0142",The signed licence covering it"rights_basis": "owner_held_rights",Why the owner may license it"period": ["2023-04-01", "2024-03-31"],When the source data was created"deid_method": "quarry-deid/audio 1.4.0",Pipeline and version, as in the report"checks": { "pii_scan": "pass", … },Each check, pass or fail"sha256": "9f2c…e41a",Hash of the delivered file"delivered_at": "2026-11-02T09:14:00Z"Delivery timestampFields
| Field | Type | Meaning |
|---|---|---|
record_id | string | Quarry's ID for the item |
dataset_id | string | The listing it belongs to (e.g. QD-7K2P) |
owner_ref | string | Pseudonymous owner ID; real identity in the licence |
licence_id | string | The signed licence covering this record |
rights_basis | enum | Owner-held rights, consent, contract, other (documented) |
period | date range | When the source data was created |
deid_method | string | Pipeline name and version |
checks | object | PII scan, burned-in text, deduplication, quality score, each pass / fail |
sha256 | string | Hash of the delivered file |
delivered_at | datetime | Delivery timestamp |
rights_basis values: owner_held_rights, consent, contract, other. When it's other, the basis is documented in the datasheet.
Example
{"record_id":"qr_01HZX4T9K2M7","dataset_id":"QD-7K2P","owner_ref":"own_5TQ9","licence_id":"lic_2026_0142","rights_basis":"owner_held_rights","period":["2023-04-01","2024-03-31"],"deid_method":"quarry-deid/audio 1.4.0","checks":{"pii_scan":"pass","dedup_web":"pass","dedup_internal":"pass","benchmark_contamination":"pass","quality_score":"pass"},"sha256":"9f2c5d0e7a41b8c3f6d29e04a7b1c58e3d0f6a92b4c7e81d5a3f9b06c2e7e41a","delivered_at":"2026-11-02T09:14:00Z"}
{"record_id":"qr_01HZX4T9K2M8","dataset_id":"QD-7K2P","owner_ref":"own_5TQ9","licence_id":"lic_2026_0142","rights_basis":"owner_held_rights","period":["2023-04-01","2024-03-31"],"deid_method":"quarry-deid/audio 1.4.0","checks":{"pii_scan":"pass","dedup_web":"pass","dedup_internal":"pass","benchmark_contamination":"pass","quality_score":"pass"},"sha256":"1b7e0c4f9a2d63e85c0f17a4b9d2e6c3a8f05d71e4b29c6a0d3f8e5b7c1a29d4","delivered_at":"2026-11-02T09:14:00Z"}
Values are a specimen. To check a file against its record, see checksums and verification.