Get startedWorkflows · Durable execution
Run and control workloads
Follow annotation, model analysis, evaluation, and workflow jobs with saved progress and results.
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 what is running, where it is running, what it costs, and what artifact it produced.
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 or supported operation
- — Versioned inputs and a selected runtime
Expected output and limits
- — Durable task records
- — Artifacts, logs, usage, and reviewable evidence
Core workflow
Choose completed work
Start an annotation, inference, evaluation, or reusable workflow outcome.
Select inputs and runtime
Bind immutable versions and choose Score-managed credits, a provider, or your own device.
Control execution
Open Build & evaluate → Runs for the full run history, or Runs in the bottom bar for quick progress. Open a run to inspect its status and use pause, resume, or cancel only when supported.
Inspect the result
Follow the task back to its source page and retain outputs, logs, spend, and review evidence.
What this surface supports
Active and recent runs
Progress and task deep links
Multi-model annotation
Managed-credit execution
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
- — Phase and percent complete
- — Runtime utilization
- — Credits consumed
- — Success or actionable failure
Common failure modes
- — Insufficient credits or capacity
- — Revoked provider credentials
- — Retrying non-idempotent work blindly
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
- — Durable state survives navigation
- — Cancellation is explicit
- — Every result records inputs and runtime