Academy
5. Running, Budgeting and Reviewing Results
Execute a run with new inputs, monitor live progress, set a budget, cancel if needed, and retrieve deliverables from the run page and file space.
A Live Artifact run is a single execution of a reusable workflow with fresh inputs. When you open an artifact and fill in its declared inputs — a file from your file space, a date range, a keyword list — the platform validates each value against the types and constraints set at forge time. Hit Run, and the run page opens with live progress: you see each step as @oupi works through the workflow. Every run is tracked with its inputs, cost in credits, deliverables and a verification verdict, all accessible from the artifact's history.
Each run has a budget, expressed in credits. The budget caps how much a single execution can consume; if the work reaches the limit, the run stops and preserves what was produced so far. This is critical for advanced workflows that call multiple tools or process large files — without a budget, a complex run could drain credits unexpectedly. Set the budget before launching, especially for artifacts shared with your team where each member's runs are billed to them individually.
Keep budgets tight on first runs of a new or derived artifact. Review the cost in the history after a successful run, then adjust the budget upward only if needed. This protects your balance while you learn the artifact's real consumption profile.
You can cancel a run at any time from the run page. Cancellation stops the work immediately and keeps whatever deliverables were already produced — nothing is lost. This mirrors how missions work across the platform: cancel preserves partial output. Use it when live progress shows the run heading in the wrong direction, or when you realize an input was incorrect. You can then correct the inputs and launch a fresh run.
Deliverables from a completed run are available in two places. First, the run page itself lists every output file — open or download directly. Second, all deliverables are stored in your file space under the artifact's dedicated folder, so they persist and are accessible outside the run context. The full history view lets you compare runs side by side: inputs used, credits consumed, and the verification verdict (verified, failed, or could not be verified). Use this to audit recurring workflows or spot regressions after modifying an artifact.
When the verdict says "could not be verified," it means the checker could not read the output (e.g., a binary file). It is not a failure — but you should review the deliverable yourself before acting on it.
Open one of your Live Artifacts, fill in new inputs, set a conservative budget, and launch a run. Watch the live progress. Once it completes, check the run page for deliverables and the history tab for cost and verdict. Then navigate to your file space to confirm the outputs are stored in the artifact's folder.
A Live Artifact run takes declared inputs, validates them, and executes with live progress. Set a budget to cap credit consumption; cancel anytime to preserve partial output. Deliverables appear on the run page and in your file space folder. The history tracks every run's inputs, cost and verdict — use it to audit, compare and refine. Master these controls and each run becomes predictable, auditable and cost-efficient.