Results and provenance
Interpret item wrappers, metadata counts, partial failures, and deduplication.
Rust and CLI response
The following is a schematic response; the item dictionary is abbreviated, not a complete STAC Item.
{
"items": [
{
"catalog_id": "earth-search",
"item": {"id": "example-scene", "collection": "sentinel-2-l2a"},
"seen_in": ["earth-search", "microsoft"]
}
],
"metadata": {
"catalogs_queried": 2,
"catalogs_succeeded": 2,
"catalogs_failed": 0,
"total_items": 1,
"duplicates_removed": 1,
"failures": [],
"unsupported_collections": []
}
}catalog_id is the catalog whose item body was retained. seen_in tracks catalogs returning the same item ID. Deduplication does not combine asset dictionaries or resolve which provider has the best data.
Metadata fields
| Field | Meaning |
|---|---|
catalogs_queried | Number selected for execution, after health/source filtering. |
catalogs_succeeded | Selected catalogs that completed successfully. |
catalogs_failed | Selected catalogs that failed. |
total_items | Items in the final response, after optional deduplication. |
duplicates_removed | Number of collapsed item records. |
failures | List of {catalog_id, reason} for failed searches. |
unsupported_collections | Requested canonical IDs that no candidate catalog was known to serve, reported conservatively when collection knowledge is complete. |
A catalog excluded for health or capability reasons is not necessarily represented in failures. Compare metadata with your configured registry when diagnosing missing sources.
Python representation
search.items() unwraps the item bodies into Python dictionaries. The per-item catalog_id and seen_in fields are not currently exposed by the Python Search object. search.metadata exposes the run-level fields above.
search.to_geojson() returns { "type": "FeatureCollection", "features": [...] }. It omits the Rust provenance wrappers and run metadata. Store search.metadata separately if you need a record of search failures.
Deduplication identity
Deduplication compares Item.id across the entire response, without including the collection ID. Disable it if your sources reuse item IDs for unrelated records:
from superstac import Client
client = Client(settings={"deduplicate_items": False})Repeated observations with different IDs are not recognized as duplicates. Output order and the retained provider are not a stable ranking.
