Batch page#

The batch page at /batch defines one or more geographic scenes and expands each into fixed or seeded-random cases with the same schema as the Python interface (Planning and generation).

Sections#

Identity and execution

Dataset name, random seed, continue-on-error, and Resume / extend for an existing dataset with a higher case count.

Scene definitions

Add scene appends a scene. Each has a label, centre coordinates, radius, receiver height, voxel pitch, building source, buildings on or off, terrain on or off with mesh spacing, material profile, and the empty scene mode.

Cases per scene

Seeded random or Fixed strategy, case count, solver (preview or Sionna RT), transmitter count range and the ranges for frequency, power, height, offset radius, azimuth and downtilt; antenna patterns; the solver candidates for resolution, reflection depth, samples per transmitter and association metric; the retained case artifacts; and the Sionna runtime and scene mask switches.

Plan and generate

Build plan calls POST /api/batches/plan and shows the exact case names, solver seeds, receiver-cell count and the initial ray budget before anything is executed. Generate dataset calls POST /api/batches; the page then polls GET /api/batches/{job_id} and offers the manifest when the job completes.

Behaviour#

  • Cases run sequentially inside one job. Only one batch job runs at a time; a second request for the same dataset name is rejected while the first is queued or running.

  • A dataset name that already exists on disk is rejected unless resume is enabled.

  • Jobs and their status live in the service’s memory; the dataset itself is written under OWRT_DATASET_ROOT and survives a restart.

  • Sionna cases are refused up front when the runtime is not installed.