Get startedWorkflows · Preview and tracked runs
Choose the right execution mode
Use Preview for transient inspection and a tracked workflow for saved labels, artifacts, approvals, deliveries, and receipts.
Source reviewed 2026-09-16. Availability depends on your installation, permissions, and compatible runtimes. A supported path is not a guarantee of model quality or production readiness.
Outcome
Know exactly which blocks ran, which side effects were skipped, and where to inspect retained work.
Start here
What you need to know first
Check the inputs below before starting. If you are new, begin with the first-project guide. A dataset holds media and labels; a model produces results; a deployment makes a selected model version callable. Creating one does not create the others.
Bring these inputs
- — A workflow revision and representative preview media
- — Versioned inputs and a compatible runtime for tracked execution
Expected output and limits
- — Temporary node output from Preview
- — A durable workload with typed artifacts and receipts from tracked execution
Core workflow
Preview pure processing
Run supported model and image-processing blocks against selected media. The visible result follows the terminal filter, gate, or count while technical details retain intermediate output. Overlays and downloads retain the image and graph that actually ran. If you change the input or workflow, run it again to test that setup; the previous output stays identified as a previous result.
Read skipped actions
The preview result names every durable block it intentionally did not execute, including dataset, annotation, and delivery actions.
Start tracked work
Launch a workload to bind versioned inputs, execute supported side effects once, and retain progress and receipts.
Inspect the result
Open the task or workload to review its output artifact, dataset version, evidence, approval, deployment, or delivery.
What this surface supports
Transient graph previews
Durable workload IDs
Typed artifact references
Per-block progress and receipts
Expert section
Contracts, signals, and failure modes
Use this section when you are defining acceptance criteria, automating the surface, or reviewing whether its output is safe to promote downstream.
Quality and operating signals
- — Blocks executed or deliberately skipped
- — Per-node state, percent complete, and ETA
- — Retained artifact, approval, evaluation, deployment, or delivery references
Common failure modes
- — Treating Preview as proof that labels or outputs were saved
- — A graph containing blocks without durable adapters
- — Retrying a non-idempotent side effect outside the tracked engine
Expert release checklist
□ Inputs and dependencies are pinned to immutable versions.
□ Acceptance metrics include critical classes and operating slices.
□ Failure, retry, cost, and rollback behavior are understood.
□ The resulting artifact has an owner and a downstream review path.
Engineering safeguards
- — Preview never claims skipped actions ran
- — Unavailable durable adapters block publishing and tracked execution
- — Completed side effects are not repeated blindly