Buildings, Python SDK reference

BuildingsAcquisitionError

Bases: Exception

Direct building acquisition failed in a way the caller must see.

BuildingsServiceClient

Bases: ScrubbedSessionState

Client for 3D building meshes, read from the public sources.

api_key, base_url, gateway_base_url and telemetry are accepted for construction compatibility with the other service clients and are unused: this client sends no request of its own. It implements the context-manager protocol.

Parameters:

Name	Type	Description	Default
`api_key`	`str`	API key, accepted for compatibility with the other service clients.	required
`logger`	`Logger`	Logger used for progress and warning messages.	required
`base_url`	`str`	Base URL, accepted for compatibility with the other service clients.	required
`gateway_base_url`	`str`	Accepted for compatibility and unused.	`None`
`telemetry`	`Telemetry`	Accepted for compatibility and unused.	`None`
`acquisition`	`optional`	Removed option. Passing any value raises `TypeError`.	`REMOVED`

Attributes:

Name	Type	Description
`logger`	`Logger`	Logger used for progress and warning messages.
`base_url`	`str`	Base URL the client was constructed with.

close

Python

close() -> None

Release the client's resources. Kept for API compatibility.

get_area

Python

get_area(
polygon: dict,
*,
config: Optional[BuildingsConfig] = None,
on_progress: Optional[Callable[[TileProgress], None]] = None,
max_tiles_override: Optional[int] = None,
timeout: int = 60,
total_timeout: int = 600,
max_workers: int = DEFAULT_TILE_ACQUISITION_WORKERS,
overture_release: Optional[str] = None,
analysis_type: Optional[str] = None,
acquisition: Any = REMOVED,
) -> AreaBuildings

Fetch and deduplicate buildings for a polygon area.

The whole site is read at once: a single rectangle up to 4 km2, and a grid of roughly 2 x 2 km read chunks above it, at most two in flight. The meshes come back in one site frame, anchored at the polygon bounding-box south-west corner, and run_area re-anchors them per tile. A polygon straddling a registered city's outline is read tile by tile instead, because a site-level read would give the city's bodies to the tiles outside it.

There is no partial answer on the site path. A chunk that fails raises TiledRunError, whose failed_tiles names the failed chunks and whose __cause__ is the first underlying error.

Parameters:

Name	Type	Description	Default
`polygon`	`dict`	GeoJSON Polygon.	required
`config`	`BuildingsConfig`	Accepted for compatibility and unused: it shaped the removed `POST /buildings` body.	`None`
`on_progress`	`callable`	Progress callback `(TileProgress) -> None`, once per read CHUNK on the site path. `tile_id` is the chunk id and `total_count` the chunk count.	`None`
`max_tiles_override`	`int`	Override the maximum number of non-empty tiles allowed.	`None`
`timeout`	`int`	Per-REQUEST budget in seconds (default 60), trimmed to what is left of `total_timeout`.	`60`
`total_timeout`	`int`	Wall-clock deadline over the whole site read, in seconds (default 600).	`600`
`max_workers`	`int`	Read chunks in flight on the site path. A CEILING of 2 applies, so this can only ask for FEWER. On the mixed-source per-tile path it is still the thread count.	`DEFAULT_TILE_ACQUISITION_WORKERS`
`overture_release`	`str`	Pin the Overture footprints to one immutable release. The resolved release is reported on the result.	`None`
`analysis_type`	`str`	The analysis this acquisition is for. It decides the tile grid and the READ MARGIN, the half extent of each read rectangle: 256 m for the two wind analyses and 384 m for every other one, which needs the shadow casters further out. Omit it for the widest margin, valid for every analysis.	`None`

Returns:

Type Description

AreaBuildings Deduplicated buildings with coordinates in polygon-bbox-SW meter space, and origin naming that frame — pass the whole object to run_area(buildings=…) and it re-anchors from it, so one acquisition can serve several sub-areas. building_ids is always empty: the read path has no numeric building ids. failed_tiles is empty on the site path: a read that could not cover the polygon raises instead.

Type	Description
`AreaBuildings`	Deduplicated buildings with coordinates in polygon-bbox-SW meter space, and `origin` naming that frame — pass the whole object to `run_area(buildings=…)` and it re-anchors from it, so one acquisition can serve several sub-areas. `building_ids` is always empty: the read path has no numeric building ids. `failed_tiles` is empty on the site path: a read that could not cover the polygon raises instead.

Raises:

Type	Description
`PolygonValidationError`	If `polygon` is not a valid GeoJSON Polygon, or covers more non-empty tiles than allowed (see `max_tiles_override`).
`TiledRunError`	If a read chunk fails on the site path.
`TypeError`	If the removed `acquisition` argument is passed.

get_by_tiles

Python

get_by_tiles(
tiles: TileGrid,
*,
config: Optional[BuildingsConfig] = None,
on_progress: Optional[Any] = None,
timeout: int = 60,
total_timeout: int = 600,
max_workers: int = DEFAULT_TILE_ACQUISITION_WORKERS,
) -> Dict[str, Optional[dict]]

Fetch buildings for each non-empty tile in a pre-generated grid.

Returns a per-tile mapping {tile_id: buildings_dict | None} intended for inspection or manual per-tile dispatch.

All tile IDs from the grid are present as keys in the returned dict. A None value means the fetch failed for that tile (distinguishable from "no buildings found", which would be an empty dict).

To combine get_by_tiles output into a single dict for visualisation (not for run_area) see merge_buildings.

Warnings

This shape is NOT a drop-in for InfraredClient.run_area's buildings= kwarg, which expects a flat {building_key: building_data} mapping (each value a dict with top-level coordinates, mesh_id, indices). Passing this method's per-tile output directly into run_area(buildings=...) will fail every tile with "missing or non-sequence coordinates". For the area pipeline use client.buildings.get_area(polygon).buildings, the canonical flat shape.

Parameters:

Name	Type	Description	Default
`tiles`	`TileGrid`	Tile grid produced by tile generation for a polygon.	required
`config`	`BuildingsConfig`	Accepted for compatibility and unused: it shaped the removed `POST /buildings` body.	`None`
`on_progress`	`callable`	Progress callback `(TileProgress) -> None`.	`None`
`timeout`	`int`	Per-tile timeout in seconds (default 60).	`60`
`total_timeout`	`int`	Total wall-clock timeout in seconds (default 600).	`600`
`max_workers`	`int`	Maximum parallel threads for tile execution (default 20).	`DEFAULT_TILE_ACQUISITION_WORKERS`

Returns:

Type	Description
`dict[str, dict \	None]`	Maps each non-empty tile's ID to its buildings dict, or `None` if the fetch failed for that tile.

AreaBuildings

Bases: Payload

Result of get_area().

Frozen Pydantic model — field reassignment is blocked, but mutable containers (buildings dict) can still be mutated in-place (Pydantic v2 behavior), enabling the edit-before-run workflow.

Attributes:

Name	Type	Description
`buildings`	`dict[str, [DotBimMesh](#infrared_sdk.buildings.DotBimMesh)]`	Building meshes keyed by building identifier.
`attributes`	`dict[str, dict]`	`{height, min_height, num_floors, source_id}` per building, under the same keys as `buildings`. Populated on both read paths (the site-level read and the per-tile read a mixed-source polygon falls back to). Empty when the meshes did not come from a direct read, which carries meshes and no attributes. Empty therefore means "not available on this path", never "this building has no height".
`building_ids`	`list[int]`	Numeric building IDs from the API response.
`polygon`	`dict`	The input GeoJSON polygon.
`total_buildings`	`int`	Number of buildings in `buildings`.
`execution_time`	`float`	Wall-clock seconds for the fetch.
`failed_tiles`	`list[dict]`	Per-tile failure records for tiles that did not produce buildings after all retries (empty when every tile succeeded). Each entry has `tile_id`, `row`, `col`, `error`. Callers can use `len(area.failed_tiles)` to detect partial-coverage results.
`overture_release`	`(str, optional)`	Overture Maps release the footprints were read from, when the direct acquisition path produced them (`None` on the service path, which does not report one). The public index pointer moves daily; pass the value back as `overture_release=` to pin a later run to the same snapshot.
`read_margin_m`	`(float, optional)`	Half extent, in metres, of the read rectangle every tile was fetched with: 256 m for the wind analyses, 384 m otherwise (see `analysis_type`). `run_area` refuses a run whose analysis needs more than this.
`analysis_type`	`(str, optional)`	The analysis type the read margin was taken from. `None` on an object built by hand, which makes no claim and is never refused.
`origin`	`(tuple, optional)`	`(lon, lat)` of the site frame the meshes are stored in: the `polygon` bounding-box south-west corner. `run_area(buildings=)` reads it and re-anchors from THIS frame, so acquiring once for a large polygon and running several sub-areas places the buildings correctly. A bare `{id: mesh}` map carries no origin and is read as being in the run polygon's own frame, which is the behaviour that has always applied.

BuildingsConfig

Bases: CamelCasePayload

Configuration for tiled buildings retrieval.

Accepted for compatibility and unused: it shaped the removed buildings request.

Attributes:

Name	Type	Description
`output_format`	`OutputFormat`	Desired output format (default DotBim).
`compress`	`bool`	Whether to request compressed responses.
`optimizations`	`(BuildingOptimizations, optional)`	Optimizations of the removed per-tile request; no effect.

BuildingsRequest

Bases: CamelCasePayload

Request body for a buildings query.

Kept for compatibility: the SDK no longer sends this request.

Attributes:

Name	Type	Description
`coordinates`	`BuildingCoordinates`	Centre of the query.
`size`	`BuildingSize`	Extent of the query.
`return_building_ids`	`bool`	Whether to return numeric building IDs (default `False`).
`output_format`	`OutputFormat`	Desired output format (default DotBim).
`compress`	`bool`	Whether to request a compressed response (default `False`).
`optimizations`	`(BuildingOptimizations, optional)`	Optimisations to apply.

DotBimMesh

Bases: Payload

A single DotBim mesh entry from the buildings API.

Follows the Payload base pattern (frozen, kebab-case aliases).

Attributes:

Name	Type	Description
`mesh_id`	`int`	Numeric identifier of the mesh.
`coordinates`	`list[float]`	Flat `[x, y, z, ...]` coordinate array.
`indices`	`list[int]`	Triangle index array. Required.

Notes

indices is required. A mesh without it is silently dropped by the simulation service before any computation runs: no error, no warning, and a tile that reads as if the building had never been submitted, yet is billed in full. The SDK therefore refuses to build one.

merge_buildings

Python

merge_buildings(buildings_map: Dict[str, Optional[dict]]) -> dict

Combine per-tile buildings dicts into one for visualization.

Merges all non-None values from buildings_map into a single dict. Each key in the individual buildings dicts whose value is a list is concatenated; other keys are taken from the last dict encountered.

Parameters:

Name	Type	Description	Default
`buildings_map`	`dict[str, dict \	None]`	Maps `tile_id -> buildings dict` (or `None`).	required

Returns:

Type	Description
`dict`	Combined buildings dict, or empty dict if all are `None`.