Against a run that requests no deforestation at all:

• Facility runs additionally solve the supply-shed isochrone and run an Earth Engine reduce over it. The isochrone is skipped entirely when nothing consumes it, which is why a core-validation or geocode-only upload is much cheaper, and the solve is bounded by a 750 s timeout before it degrades to a 50 km buffer.
• Plot runs screen deforestation per polygon, and a plot with deforestation present takes a 0.15 penalty on its confidence score. So confidence_score and confidence_level are not invariant to these flags: the same upload scores differently with the check on. If you gate metric computation with metric_confidence_levels, that gate reads the affected level — so on a run that also requests metrics, this can change which locations are computed, not just what is displayed.
• The reprocess-only include_deforestation_screen additionally costs a derived-location pass, which include_deforestation never does. That flag puts deforestation_screen into the run's requested assessment types, so the screen also runs over each staged_location derived during the run — the sheds and plots linked from your upload. Those are new locations with no cached validation row, so each is a real upstream call and emits its own usage event, and the pass is uncapped. This is the reason the upload path does not offer the flag: on POST it would buy that fan-out and no extra data.
• An art9 row is opened per location either way, because the same check emits the isochrone/shed columns that belong to the art9 partition. The legality family is filtered out when Article 9 was not itself requested, so /assessments/{location_type}/legality can return rows whose legality fields are all null. Read legality_risk_level rather than row presence to decide whether Article 9 actually ran.

include_eudr drives both EUDR-scope groups on the DDS create path. It is silently stripped if sent to POST / PUT /ingestions/assets (see assessments-overview.md) — use the per-group include_art9 / include_art10 flags (or include_legality for both) instead.

Field	Type	Default	Group enabled	Description
include_eudr	boolean	false	art9 + art10	Shorthand that fans out to include_art9 + include_art10 only — not hazard / infrastructure.

When both include_eudr and a per-group flag are present, the union wins — include_eudr: true cannot be narrowed by setting an individual group to false.

The check_* flags are cross-cutting toggles that drive the water / port / road access checks. They sit at the top level of the assessments block — not nested under any group — because the same assessment call produces matching keys on both core and infrastructure:

• core — boolean access flags: water_access, port_proximity, road_access.
• infrastructure — numeric distances in km: nearest_water_km, nearest_port_km, nearest_road_km.

Field	Type	Default	Drives (in core)	Drives (in infrastructure)
check_water_access	boolean	false	water_access	nearest_water_km
check_port_proximity	boolean	false	port_proximity	nearest_port_km
check_road_access	boolean	false	road_access	nearest_road_km

Both rows record the check_* snapshot on their sub_options column. A subsequent request that asks for an access check not covered by the cached sub_options is treated as a cache miss and refreshes the affected rows.

---

Field	Type	Description
confidence_score	number	0–1 score combining all plausibility indicators below.
confidence_level	string	low | medium | high. Bucketing of confidence_score.
confidence_comment	string	Free-text explanation of which indicators contributed.
comment	string	Generic per-feature comment.
common_comment	string	Per-collection comment shared across features.
total_plots	number	Total plot count in the parent collection.
primary_warning	string | null	Headline warning string for UI surfacing (most severe issue).
feature_idx	number	Position of this feature in the source upload (0-based).
location_type	string	Detected location type — facility | plot | shed.
commodity_invalid	boolean	Present and true when the row's commodity cell names no supported commodity. The row was checked against the file's commodity (the declared commodity_type, or the column's most common value) instead. Absent when no commodity column was used.

Field	Type	Description
producer_confirmation_score	number	0–1 — confidence that the named producer matches points of interest / business records at the location.
producer_confirmed	boolean	True iff producer_confirmation_score exceeds the per-commodity threshold.
producer_confirmation_comment	string	Free-text explanation of which evidence supported / contradicted the producer name.
aggregation_comment	string	Comment about how multiple sources were combined into the score.
poi_data	object	Raw point-of-interest matches that informed scoring.
commodity_presence	boolean / number	Whether commodity probability maps show presence at the location.
poi_presence	boolean	Whether matching points of interest (mill, silo, warehouse, etc.) were found nearby.
high_density	boolean	Whether the surrounding area has high industrial / built-up density.
building_presence	boolean	Whether building footprints exist within the detection radius.
road_access	boolean	Whether a road network segment is reachable within plausible distance.
water_access	boolean	Whether navigable waterways are within plausible distance.
port_proximity	boolean	Whether a port facility is within plausible distance.

Each *_count is the number of plots in the upload that triggered the matching geometric check.

Field	Type	Description
max_distance_gt_min_distance_count	number	Plots whose max-vertex distance vastly exceeds their min-vertex distance (elongated shapes).
max_angle_lt_min_angle_count	number	Plots with extreme interior-angle disparity (spiked geometries).
intersects_another_polygon_count	number	Plots that overlap another plot in the same collection.
distance_nearest_neighbor_count	number	Plots that are extremely close to a neighbour without overlapping (likely boundary error).
area_too_small_count	number	Plots below the per-commodity minimum area.
triangular_geometry_count	number	Plots whose vertex layout is suspiciously triangular (placeholder geometry).
bad_precision_count	number	Plots whose coordinates have insufficient decimal precision (~10 m floor).
short_boundary_segments_count	number	Plots with sub-resolution boundary segments.
self_intersection_count	number	Plots whose polygon ring crosses itself.
duplicate_vertices_count	number	Plots with consecutive identical vertices.
nested_geometry_collection_count	number	Plots delivered as nested GeometryCollection.
simplifiable_collections_count	number	Plots whose nested collection can be flattened without loss.
commodity_absence_count	number	Plots where commodity probability maps show no presence.
protected_area_overlap_count	number	Plots that intersect a WDPA protected area.
protected_area_assessed	boolean	false when the protected-area read did not complete, so the overlap count was not verified.
plot_validity	object	Aggregate shape-validity summary { score, level, investigate[] }.
deforestation	object	Per-plot deforestation summary { value, area_ha, level } (separate from art9).
overall	object	Top-level pillar-weighted summary { score, level, investigate[] }.
protected_area_metadata	object	WDPA polygon ids and names overlapping the plot, when applicable.

---

The route geometry behind port_route_count is not returned by default. Each route is a decoded driving polyline of several hundred [lon, lat] vertices, so a location can carry tens of KB — enough to matter on a bulk upload, where a single page of results would otherwise drag the whole set through Postgres.

Request it explicitly on the infrastructure read endpoint:

GET /assessments/{location_type}/infrastructure?include=port_routes

Field	Type	Description
port_routes	array | null	Top-level attribute (not inside distances). An array of routes; each route is a list of [lon, lat] vertices, ordered from the location to the port. Route 0 is the primary. null when the location has no routed port.

The key is omitted entirely without the include, so a caller can tell "I did not ask for this" from "this location has no routes". It is an include, not a sparse field — fields[assessment]=port_routes is a 400. The two compose: ?include=port_routes&fields[assessment]=distances returns both.

port_routes is never present on the aggregate assessment relationship (GET /locations?include=assessment, DDS record reads) or in CSV artifact exports, and it is not replicated onto catchment members when one representative location is assessed for a group — the routes belong to the location they were measured from.

---