Skip to main content
Base URL: https://api-sim.hardlightsim.com Base path: /v0 Managed training base path: /v1 Auth: Authorization: Bearer hls_live_your_service_key

​
Core Endpoints

  • POST /v0/jobs create simulation job.
  • GET /v0/jobs/{job_id} fetch status + artifacts.
  • GET /v0/jobs list jobs for authenticated customer.
    • supports status and submission_group_id filters
  • POST /v0/jobs/{job_id}/cancel request cancellation.
  • GET /v0/jobs/{job_id}/artifacts/{artifact_name}/download-url fetch presigned URL.
  • POST /v0/inputs/presign-upload create upload URL for input files.
  • POST /v0/assets/register register an input URI and return asset_id.
  • GET /v0/assets list registered input assets.
  • GET /v0/assets/{asset_id} fetch one registered input asset.
Hardsim supports two input styles:
  • URI-based: robot.urdf_uri, robot.usd_uri, scene.usd_uri
  • Asset-ID based: robot.urdf_asset_id, robot.usd_asset_id, scene.usd_asset_id
When you send *_asset_id, the API resolves it to canonical URI server-side before job validation and execution.

​
Isaac Lab Submit Additions

POST /v0/jobs also supports an Isaac Lab rollout path with:
  • runtime_profile="isaac_lab_rollout"
  • isaac_lab.task_module
  • isaac_lab.task_class_or_name
  • isaac_lab.task_package_asset_id
  • optional isaac_lab.task_args
Use this only when you need a custom task package. For the standard customer path, submit_assets(...) with robot/scene assets is still the recommended entry point.

​
Managed Training Endpoints

  • POST /v1/training-runs
  • GET /v1/training-runs/{run_id}
  • POST /v1/training-runs/{run_id}/pause
  • POST /v1/training-runs/{run_id}/resume
  • POST /v1/training-runs/{run_id}/cancel
  • GET /v1/training-runs/{run_id}/iterations
  • GET /v1/training-runs/{run_id}/checkpoints
  • GET /v1/training-runs/{run_id}/events
  • GET /v1/training-runs/{run_id}/checkpoints/{checkpoint_id}/download-url
  • GET /v1/training-runs/{run_id}/artifacts/download-url?uri=...
  • GET /v1/training-runs supports submission_group_id filter.
Managed training payload also supports checkpoint_init_asset_id (resolved to checkpoint_init_uri).

​
Submission Group Endpoints

  • POST /v1/submission-groups
  • POST /v1/submission-groups/{group_id}/ingest
  • GET /v1/submission-groups/{group_id}
  • GET /v1/submission-groups
  • GET /v1/submission-groups/{group_id}/items
  • GET /v1/submission-groups/{group_id}/events
  • POST /v1/submission-groups/{group_id}/cancel
Submission groups are typed (simulation or training) and support partial accept with per-item machine-readable errors. Managed training scopes:
  • training:write for create/pause/resume/cancel.
  • training:read for get/list/iterations/checkpoints/events.
Managed training guardrail:
  • trainer_spec.resources.gpu must be greater than 0 (GPU-only trainer policy by default).

​
Submit Response Billing Fields

POST /v0/jobs includes:
  • estimated_credits
  • estimated_usd
  • billing_model
  • billing_price_id
  • gpu_hourly_rate_usd
These values are estimates at submit time. Final debit is settled at terminal state from runtime usage.

​
Error Types You Should Handle

  • 400 invalid payload (for example, missing required GPU trainer resources).
  • 402 insufficient credits / billing guardrail.
  • 403 plan not entitled.
  • 404 cross-tenant or unknown job.
  • 409 idempotency conflict.
  • 429 tenant guardrail exceeded.

​
Full Contract

Use the guides in this docs site for production request/response patterns: