Skip to main content

​
Running on demand

Click the [Run] button on the canvas toolbar to open the run panel. If the refinery uses a source that pulls its own data (Schedule, Cloud Source), you can run without any sample input; expand Advanced · Override with sample input to run with a sample instead of the real source. While running, the active node is highlighted on the canvas so you can follow along.

​
Run history

The [Logs] button in the toolbar opens the run history drawer. The list of recent runs sits on the left; picking a run opens its detail on the right.
  • Search: search past runs.
  • All / OK / Errors: filter by status.
Empty until you’ve run at least once. After a run, the stored report appears.

​
Run detail

Picking a run shows its Overview on the right, with the stage tree on the left. Click a stage in the tree and the right pane swaps to Stage detail.

​
Right pane: Overview (default)

  • Header: Status pill (ok / partial / error), run_kind (dry_run · file · folder · …), trigger source via <origin>, input label, timestamp / input mode / document count (for folder runs).
  • StatGrid: latency · stages (ok / total) · errors · output size.
  • Halt banner · notice · error text: if the run hit a first error, a banner points at the offending stage. Any notice attached to the run or the run-level error text appears here too as callout cards.
  • Trigger metadata: triggered_by · idempotency_key (API triggers) · preview (dry-run flag) · input_mode · input_size.
  • Variables override: per-run flow-variable overrides. An empty-state label shows when nothing was overridden.
  • Stage summary: a table view of every stage, including block, palette, status, latency_ms, output_size, and message.
  • Folder documents: per-document rows showing the document, stages, latency_ms, and output_size. Available for folder runs only.

​
Left pane: stage tree

Each stage the run went through appears in a chronological tree. A row shows:
  • Status glyph (ok / error / skipped).
  • Node name and palette.
  • Latency.
  • Input → output size delta: +4.1MB / -820B style, showing how much the payload grew or shrunk.
  • Artifact stats chips: records / tokens / cost / cache hits when the stage produced them.
If the run attached anything to the knowledge graph, a KG routing subgroup appears under the stages. It shows one row per source with attached/examined and the ≥ threshold. Click a KG row to open its detail on the right. Click a stage row to swap the right pane from Overview to Stage detail.

​
Right pane: Stage detail (after clicking a stage)

A segmented pill row (Config · Output · New output · Artifacts · Error) sits at the top. Click a pill to smooth-scroll to that section; the pill also updates as you scroll (scroll anchors, not tabs).
  • Config: the config snapshot of this node (how the node was wired). Only shown when the node has config.
  • Output: the full output the stage handed downstream.
  • New output: this Foundry-specific view shows the stage’s net output with the upstream context stripped out (delta preview). It helps you spot exactly what a stage added when payloads get large.
  • Artifacts: runner counters (records, tokens, cost, cache hits, etc.).
  • Error: only shown when the stage errored.
The Readable / Raw toggle in the top-right switches presentation. Readable renders JSON as a table for scanning; Raw shows the JSON text as-is. Your choice is saved and picks up next time. Fields marked as secrets are automatically masked.

​
Pre-publish checks

If wiring is broken or a required field is empty, an error / warning badge appears at the top-right of the canvas. Fix it before running. For deeper per-run analysis, continue in Traces.