Experiment tracking

Record experiment objectives and log measured runs, parameters, metrics, outcomes, and artifact locations.

Open Experiments

Track a real workload

  1. Create an experiment with its type, objective, primary metric, metric direction, and dataset reference.
  2. Run the training or evaluation workload in a configured notebook, pipeline, or your own execution environment.
  3. Log its actual result using the run form or runs API. Include the primary metric for a completed run.
  4. Compare recorded runs using the same metric definition and inspect their parameters and artifact references.

Tracking and execution

The experiment registry stores run records. Selecting automatic tracking is a configuration choice; the workload or integration must actually send the run data. Creating an experiment does not launch training or connect an external tracking server automatically.

API endpoints

Use your Genorbex deployment’s origin followed by the path below. Replace IDs with resources accessible to your account. Authentication and permissions are checked by each endpoint.

GET /api/orbexdb/ai-ml/experiments/{id}/runs

Reads stored runs for an accessible experiment.

Authentication: Genorbex account token and service/resource access; API-key support follows the endpoint’s authentication guard.

POST /api/orbexdb/ai-ml/experiments/{id}/runs

Records a completed, failed, or stopped run. Completed runs must include the experiment’s primary metric.

Authentication: Genorbex account token and service/resource access; API-key support follows the endpoint’s authentication guard.

Example JSON body

{
  "name": "evaluation-1",
  "status": "COMPLETED",
  "parameters": {},
  "metrics": {
    "accuracy": 0.9
  }
}

Need help with your setup? Visit the Help Center or contact Genorbex.