Skip to main content
Every record Splendor stores carries provenance: a record of where it came from. Provenance travels with the record into the index and comes back on every search result, so you can always trace a result to its origin and trust which tenant it belongs to.

​
What provenance contains

Each result includes a provenance object: 
{
  "tenant_id": "acme",
  "dataset_id": "app-logs",
  "source": "app-logs",
  "source_key": "2024/06/01/app-0001.jsonl.gz",
  "source_etag": "\"9b2cf5...\"",
  "line": 4821,
  "transform_version": "v3"
}
FieldMeaning
tenant_idThe tenant that owns the record.
dataset_idThe dataset it belongs to.
source / source_keyThe source and the specific object the record was ingested from.
source_etagThe version of that source object, so you can tell which revision produced the record.
lineThe line or offset within the source object.
transform_versionThe parser/transform revision that produced the record.

​
Why it matters

  • Traceability. From any result you can find the exact source object and line, which makes debugging and auditing concrete rather than approximate.
  • Reproducibility. source_etag and transform_version pin a result to a specific revision of the data and the code that parsed it.
  • Isolation. tenant_id on every record means tenant isolation holds at the data layer, not only at the query layer.
Provenance is attached at ingestion to every record, so it is present on results across all search modes.