Ingestion Latest Assets Summary
Bucket counts over the current record for every location โ the deduped twin of Assets Summary.
โถWhy this exists rather than a flag on the plain summary
The cards have to describe the table beneath them. Latest Assets folds to one row per location; the plain summary counts every upload. Point the Records tab at one and the other and the two disagree by exactly the superseded records โ which reads as a counting bug rather than a wiring one.
So the pairing is the contract, and it is structural:
| Table | Cards |
|---|---|
GET /ingestions/assets | GET /ingestions/assets/summary |
GET /ingestions/assets/latest | GET /ingestions/assets/latest/summary |
counted equals the total the deduped list reports for the same filters โ pinned by an integration test against the list method itself, the same way the plain pair is pinned.
โถIdentical to the plain summary in every other respect
Same filter[...] grammar, same organisation scoping, same ?deforestation= view, same response shape, same 10 000 capped behaviour, and the same rejection of page / page_size / sort / include / fields[...] with a 400. Read Assets Summary for what each bucket means and how by_compliance counts the displayed verdict.
Two differences, both inherited from the list it pairs with:
rollup=false is a 400. The deduped grain is only defined under the fold, so there is no unfolded deduped set to count. rollup=true and rollup=task stay valid and inert, so a client can move from /ingestions/assets/summary?rollup=true to this path without editing its query.
rollup=unit counts supplier units. The dedupe stays on, and the fold moves to the supplier unit, matching List Latest Assets under rollup=unit. Send the same value to both so the cards describe the table.
filter[assets.task_id] bypasses the dedupe. Drilling into one upload counts its records whether or not a later upload replaced them โ a task is the historical record of what it ingested. The list does exactly the same, so the pair still agrees under that filter.
โถResponse
{
"meta": {},
"links": { "self": "/ingestions/assets/latest/summary" },
"data": {
"type": "asset_summary",
"attributes": {
"counted": 604,
"capped": false,
"by_compliance": { "non_compliant": 71, "pending": 15, "compliant": 273, "excluded": 4, "not_assessed": 241 },
"by_status": {
"pending": 0, "staged": 0, "processing": 0, "validating": 0,
"completed": 589, "completed_with_errors": 3, "skipped": 8, "failed": 4
},
"by_location_type": { "facility": 430, "shed": 28, "plot": 140 },
"unlocated": 6,
"not_assessable": 14
}
}
}
A record that a later upload replaced is in none of these buckets. counted is therefore lower than the plain summary's for the same filters by exactly the number of superseded records โ and the difference is the point, not a discrepancy.
Required scope:
ingestions:read