Run View
The run view shows a single pipeline run: the execution graph, the status of every task, and the artifacts and logs each task produced. Like the editor, it is organised as a menu bar plus a set of dockable windows around a central canvas.
This page describes the current run view, which is the default for all users. A legacy run view is still available via the version toggle — see Switching to the legacy run view. The legacy layout is documented in the Studio UI Guide.
Menu bar
The menu bar spans the top of the run view with the TangleML logo, the pipeline's name, and the menus below it.
Run
| Item | Availability |
|---|---|
| Inspect Pipeline | When the run's pipeline is available in your browser storage |
| View YAML | Always |
| Clone Pipeline | Always — creates a new pipeline from this run's definition |
| Export YAML | Always |
| Cancel Run | While the run is in progress, and only for the user who created it |
| Rerun Pipeline | Once the run has finished |
| Compare with another run… | Always — opens this run in Compare Runs as side A, ready for you to pick the run to compare it against |
Cancel Run asks for confirmation first: the run is scheduled for cancellation, and that cannot be undone.
View
Auto-layout re-arranges the execution graph using the same algorithms as the editor — Sugiyama, Sugiyama Centered, Compact, and Spread.
Windows
As in the editor, the Windows menu toggles individual windows and offers a Views submenu. The run view has two presets: Default, which puts Run Tools on the left and Run Details on the right, and Minimal, which hides every panel and leaves only the graph.
Windows
| Window | Contents |
|---|---|
| Run Tools | The run's actions as a set of buttons — see Run tools |
| Run Details | Run-level metadata, notes, arguments, and outputs — see Run details |
| Properties | Details for the selected task — see Task inspection |
Windows behave the same way as in the editor — drag to dock or float, and use the header buttons to minimize, maximize, or hide. See Windows.
Run tools
The Run Tools window collects the actions available for the run:
- View YAML — the pipeline definition this run executed.
- Share run — copy a direct link to the run.
- Inspect pipeline — open the run's pipeline in the editor. Shown only when that pipeline is available in your browser storage.
- Clone pipeline — create a new pipeline from this run.
- Cancel run — stop a run that is still in progress. Shown only to the user who created the run.
- Rerun pipeline — submit the same pipeline again. Shown once the run has finished.
When the window is floating it lays the tools out in a row; docked, it stacks them vertically.
Run details
The Run Details window opens on the pipeline's name, followed by a status bar summarising how many of the run's tasks have succeeded, failed, or are still running. Beside the status sit a View YAML button and an overflow menu carrying the same actions as the Run menu — Inspect Pipeline, Clone Pipeline, Export YAML, and, depending on the run's state, Cancel Run or Rerun Pipeline.
The rest of the window is organised into collapsible sections:
- Run Info — run id, execution id, who created the run, and when. Where the run recorded how it was started, a line below shows an icon and a sentence naming the source — see Run origin.
- Details — the pipeline's description and its tags, followed by three blocks:
- Pipeline Annotations — annotations carried by the pipeline, minus platform bookkeeping.
- Pipeline Notes and Run Notes — the pipeline's own notes, plus a notes field for this run specifically. Run notes are editable by whoever created the run and read-only for everyone else, so you can record what a run was for without touching the pipeline.
- Run Annotations — annotations attached to this run rather than to the pipeline. Runs submitted by the CLI or by an agent carry context here, and so do reruns, which record the run they came from and the failing run they are meant to fix. Keys and values are both click-to-copy. Notes, tags and source are left out because the panel already shows them elsewhere, as are the execution system's own
system/…keys; when nothing is left after that the block is not shown at all. The list is read-only — run annotations cannot be added or edited from the panel.
- Arguments — the values this run was submitted with.
- Outputs — what the run produced.
Run origin
Runs reach TangleML from more than one place, and the run view says which. Where a run recorded a source, Run Info shows a small icon and a sentence naming it. The same icon appears at the end of every row in the All Runs list, where hovering it gives the same sentence as a tooltip.
| Icon | The run was submitted |
|---|---|
| Window | From the Tangle web app |
| Bot | By an AI agent working inside the web app |
| Terminal | With the Tangle CLI |
| Sparkles | By an agent running outside the web app |
| Globe | By some other tool |
| Question mark | With no source recorded |
A source this build does not recognise draws the globe rather than failing, so a run from a tool introduced after your deployment still reads sensibly. A run that recorded no source at all keeps the question mark, and Run Info omits the origin line for it entirely.
Task inspection
Click a task in the graph to fill the Properties window with that task's execution detail. The panel header shows the task name and a status icon, followed by three tabs:
- Artifacts — the task's inputs and output artifacts. For the preview formats supported here, see Understanding Artifacts.
- Details — the component behind the task, its description and interface, and its annotations. Annotations you or the pipeline author set appear under Task Annotations; the ones the execution system adds are kept in a separate, collapsed System annotations section, so a long list of platform metadata does not bury your own.
- Logs — the task's container logs, alongside its execution events. Three controls sit in the logs panel's own header: Pop out logs moves the logs into a window you can dock, float, or resize, opened at a width that fits the longest line in the log rather than a fixed size; Open logs in a new tab opens them on their own page, which you can share or keep open beside the run; and View fullscreen expands the panel to fill the screen, with
Escto leave it again. When the backend is unavailable the new-tab control is disabled and says so on hover. Subgraph tasks have no logs of their own, so the tab is not shown for them.
Output nodes in the graph show a preview of their artifact directly on the node once the run has produced one, so you can read a result without opening a panel.
Selecting several nodes at once shows a small toolbar with Copy, which copies the selection so you can paste it into a pipeline in the editor — including an editor open in a different browser tab or window. Cmd/Alt + C does the same from the keyboard. See Copying and pasting nodes.
Subgraphs
When a pipeline contains subgraphs, the run view lets you drill into a subgraph's execution to see the tasks inside it and their individual statuses. The URL tracks which subgraph execution you are viewing, so a link you copy takes a collaborator to the same place in the run.
Comparing two runs
To see what differs between two executions rather than inspect one, use Run → Compare with another run…, or select two runs on the All Runs page and click Compare. See Compare Runs.
Switching to the legacy run view
A version toggle switches between the current and legacy run views for the same run, in the same way as the editor toggle. Both views read the same run data.