Modeling Workflow

This page describes the full flow of a modeling task from creation to downloading its outputs, the task status transitions, and what the modeling outputs contain. Complete the setup in the Access Guide before calling the APIs.

Input data requirements

Choose the input data based on your modeling needs:

Input type	Description	Recommendation
LCC2 scene package	Complete scene directory, including related resources such as `info/poses.json`, with capture trajectory	Preferred when you need fine model detail, clear textures, and legible text
LAS point cloud	LiDAR point cloud without a capture trajectory. Must contain RGB color (Point Data Record Format 2 / 3 / 5 / 7 / 8 / 10). LAS without color is not supported	Preferred when you need high dimensional accuracy
PLY point cloud	Complete 3DGS point cloud; can be processed without a capture trajectory	Use when neither an LCC2 scene package nor a LAS point cloud is available

Two upload forms are supported:

Directory form: Upload multiple files directly to the input directory, keeping their relative paths.
Archive form: Upload a .zip; the server extracts it automatically.

Minimal integration flow

Create a task: Call Create Task and Get Upload Credentials. Automatic modeling is enabled by default (autoRebuild = 1). The response contains the task modelId and the temporary credentials needed for upload.
Upload the input data: Using the temporary credentials from the previous step, upload in parts through the object storage API of the corresponding cloud provider. See Temporary Upload Credentials. The final file Key is savePath plus the file's relative path, for example savePath + "render2/info/poses.json".
Report upload completion: After all files are uploaded, call Update Upload Status with status = 3. The system then starts modeling automatically.
Poll task progress: Call Paginated Task List or Task Details and watch status and percent. status = 8 means modeling completed; status = 9 means modeling failed. Poll no more often than every 5 seconds.
Download the outputs: Call Download Modeling Results to get temporary download links.

Note: Modeling task concurrency is not limited. You can submit multiple tasks at once; each enters the queue right after submission and is scheduled by the compute center as nodes become free.

Other supported flows

Resumable upload and expired credentials

If an upload is interrupted or the temporary credentials expire, call Get Upload Credentials Again to obtain new temporary upload credentials, then continue the multipart upload.

Start modeling manually

Applies when the task was created with autoRebuild = 0 and all input data has been uploaded. Call Start Task to start modeling manually.

Task state machine

status	Meaning	Description
0	Waiting for upload	Task created, upload not started
1	Uploading	Default status after the task is created
2	Upload paused	Pause reported by the integrator
3	Upload completed	After the report, the server scans the `input` directory; with `autoRebuild = 1`, modeling starts automatically
4	Upload failed	Failure reported by the integrator
5	Waiting to queue	Reserved status; does not occur in Open API scenarios
6	Queued	Submitted to the compute center, waiting for a free node
7	Modeling	Watch `percent` for modeling progress
8	Modeling completed	Call the download API to get the outputs
9	Modeling failed	Read `errorCode` and `failedReason` to find the cause

Status transitions with automatic modeling:

0 Waiting for upload → 1 Uploading → 3 Upload completed → 6 Queued → 7 Modeling → 8 Modeling completed
                           ↓                                                 ↓
                     4 Upload failed                                 9 Modeling failed

Modeling outputs

The following outputs are produced when modeling completes. The core downloads are the three files in LCC native coordinates.

Output file	type value	Description	Units and coordinates
`model_align_scan.dxf`	`DXF`	Editable vector drawing, 3D 3DFACE mesh	Millimeters, 1:1, LCC native coordinates
`model_align_scan.glb`	`GLB`	GLB 3D model, ready for preview or engines	Meters, explicit Z-up, LCC native coordinates
`model_align_scan.blend`	`BLEND`	Editable Blender project file	LCC native coordinates
`model_align_scan.rvt`	`RVT`	Revit model	LCC native coordinates

Other auxiliary outputs include a floor plan BEV image, a point cloud overlay image, a furniture layout image, an overall rendered preview, and coordinate transform metadata coordinate_transform.json.

This chapter is from the official XGRIDS user manual. Official source (opens in a new tab) · Updated: Nov 2, 2026