Skip to main content

​
Outcome

After this flow, you can:
  • create a scoped service key
  • submit simulation jobs with pip install hardsim
  • download artifacts from successful jobs
  • inspect diagnostics artifacts from failed jobs

​
1) Create Service API Key

  1. Open Hardlight dashboard.
  2. Go to API page.
  3. Click Create key.
  4. Copy the key immediately (shown once).
Use only hls_live_* service keys for SDK usage.

​
2) Install SDK

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade hardsim

​
3) Configure Environment

export HARDSIM_API_URL=https://api-sim.hardlightsim.com
export HARDSIM_API_KEY=hls_live_your_service_key

​
4) Run A Job

import hardsim as hs

client = hs.HardsimClient.from_env()

robot_asset_id = client.upload_input_asset("./assets/franka.urdf", asset_kind="robot")
scene_asset_id = client.upload_input_asset("./assets/cabinet_scene.usd", asset_kind="scene")

job = client.submit_assets(
    robot_asset_id=robot_asset_id,
    scene_asset_id=scene_asset_id,
    robot_asset_type="urdf",
    num_envs=8,
    steps=128,
    scene_id="franka_cabinet_oige_v1",
    control={"task_mode": "fr3_pick_lift_block_v1", "fix_base": True},
)

result = client.wait(job.job_id, poll_interval_s=2.0, timeout_s=900.0, raise_on_error=False)
paths = client.download(job.job_id, "./outputs")

print(job.job_id, result["status"], paths)
This keeps raw storage URIs out of normal user flow and makes reruns easier. For advanced task-package workloads, see Isaac Lab Jobs.

​
5) Status From Terminal (Anytime)

# Manual status check
print(client.status(job.job_id))

# Live stream (full watch, configurable timeout)
client.watch_job(job.job_id, poll_interval_s=2.0, timeout_s=36000)

​
Common Errors

  • 403 Plan not entitled for simulation jobs: plan is not allowed by entitlement policy.
  • 402 Insufficient simulation credits: credits exhausted or billing guardrail hit.
  • 429 Tenant submit/concurrent limit exceeded: tenant guardrail triggered.
  • 404 Job not found: cross-tenant access attempt.
If a job fails, download the diagnostics artifacts as well. Start with:
  • runner.log
  • diagnostics.json
  • user_job.log
  • command.stderr.log
See Artifacts and Diagnostics.