Skip to content

Commit f0e7950

Browse files
camielvsclaude
andauthored
docs: automated update - 2026-W39 (#49)
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
1 parent 0a25da3 commit f0e7950

3 files changed

Lines changed: 28 additions & 5 deletions

File tree

‎docs/core-concepts/artifacts.mdx‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -143,7 +143,7 @@ Parquet previews read the file in pages using range requests rather than downloa
143143

144144
This has a few consequences worth knowing:
145145

146-
- **Large Parquet files open quickly, at any size.** Because the viewer never downloads the whole file, Parquet previews are not subject to the 50 MB inline preview limit that applies to other artifact types. If the storage backend serving the artifact does not support range requests, the viewer falls back to a whole-file read and reports **Too large to preview** for files above that limit.
146+
- **Large Parquet files open quickly, at any size.** Because the viewer never downloads the whole file, Parquet previews are not subject to the 50 MB inline preview limit that applies to other artifact types. To read a file in pages the viewer first needs its total size: it uses the size already recorded against the artifact where that is known, and otherwise asks storage for the file's first byte and takes the total from the response. Only where neither is available does it fall back to downloading the whole file, which is subject to the 50 MB limit and reports **This parquet file is too large to preview** above it.
147147
- **Row and column counts are exact from the start**, because they come from the file's metadata rather than from the rows that have loaded so far.
148148
- **Download schema**: a **Download schema** button in the viewer header saves the file's column layout — names, types, and nullability — as a `schema.json` file.
149149

‎docs/user-guide/pipeline-editor.mdx‎

Lines changed: 25 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -144,7 +144,7 @@ There are several ways to get something onto the canvas:
144144

145145
- Drag a component from the **Component Search** window.
146146
- Drop a component YAML file straight onto the canvas. The component is added to the canvas and to your user component library at the same time.
147-
- Paste a copied selection with `Cmd/Alt` + `V`; it lands at the pointer.
147+
- Paste a copied selection with `Cmd/Alt` + `V`; it lands at the pointer. See [Copying and pasting nodes](#copying-and-pasting-nodes).
148148
- Drag **Sticky Note**, **Input**, or **Output** out of the **Component Search** window's browse view — see [Component Search](/docs/user-guide/component-search#the-editor-window).
149149

150150
### Replacing a task's component
@@ -171,6 +171,16 @@ With a multiple selection active, the **Properties** window switches to a batch
171171
- **Auto Layout Selection** — re-arrange just the selected nodes, leaving the rest of the graph alone.
172172
- **Create Subgraph** — the same action as the floating toolbar.
173173

174+
### Copying and pasting nodes
175+
176+
**Copy** — from the floating toolbar or `Cmd/Alt` + `C` — writes the selection to your system clipboard, so it is not confined to the tab you copied from. Paste it with `Cmd/Alt` + `V` into the same pipeline, into a different pipeline, or into an editor open in another browser tab or window; the nodes land at the pointer. Copying from a run view works the same way, which is how you lift part of an executed graph back into an editable pipeline — see [Task inspection](/docs/user-guide/run-view#task-inspection).
177+
178+
Where the clipboard holds no copied selection, the editor falls back to the last selection you copied in this tab, so paste still does something sensible after you have copied unrelated text from elsewhere.
179+
180+
Clipboard access depends on the browser, and some browsers withhold it. Rather than failing silently, the editor reports what happened: **Couldn't copy to the clipboard. Check browser permissions and try again.** when a copy is refused, and the matching message when a paste cannot be read.
181+
182+
Copy and paste inside a dialog, a text field, or the arguments code editor behave as they normally would — the canvas does not intercept them.
183+
174184
### Sticky notes
175185

176186
Sticky notes are annotation nodes that carry no execution meaning — use them to label regions of a graph or leave a note for a collaborator. Drag **Sticky Note** from the **Canvas Tools** folder in the **Component Search** window's browse view.
@@ -228,7 +238,20 @@ Execution settings for the selected task:
228238
- **Conditional execution** — gate whether this task runs at all. See [Conditional execution](#conditional-execution).
229239
- **Disable cache** — opt this task out of caching. Not shown for subgraph tasks. See [Caching](/docs/core-concepts/caching).
230240
- **Collapse node** — draw the node in a condensed form on the canvas.
231-
- **Compute resources** — cloud provider and per-provider resource settings. Choosing a provider swaps in that provider's fields; switching provider clears the fields belonging to the old one.
241+
- **Compute resources** — cloud provider and per-provider resource settings. See [Compute resources](#compute-resources).
242+
243+
#### Compute resources
244+
245+
**Compute resources** decides where a task runs and on what hardware. Choosing a provider swaps in that provider's fields; switching provider clears the fields belonging to the old one.
246+
247+
Where your deployment publishes a launcher configuration, the provider list is built from it rather than from a fixed list bundled with the app, so the entries you see are the targets your backend actually offers. Each one names the platform and the cluster together — `GKE: us-abc-123`, `Nebius: B300 cluster` — and a cluster belonging to a particular project shows it underneath as `Project: <project-id>`.
248+
249+
Two kinds of entry are worth recognising:
250+
251+
- **Aliases.** A deployment can publish a stable name that stands for whichever cluster currently sits behind it. An alias is listed under its own name — an alias `prod` appears as **Prod default** — with a caption reading **Currently** followed by the cluster it resolves to today. Selecting the alias pins the task to the name rather than to that cluster, so the task follows the alias if the deployment later points it somewhere else. A pipeline that already names an alias resolves to the right cluster when it is opened.
252+
- **Withdrawn clusters.** A cluster past its availability date is left out of the list so it cannot be chosen afresh. A task already pinned to one still shows it, marked as no longer available, with the date it was withdrawn and — where the deployment names one — the cluster that replaced it.
253+
254+
A cluster your deployment lists but has not enabled is shown as unavailable, noting that it is not enabled for this deployment.
232255

233256
#### Conditional execution
234257

‎docs/user-guide/run-view.mdx‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -79,11 +79,11 @@ Click a task in the graph to fill the **Properties** window with that task's exe
7979

8080
- **Artifacts** — the task's inputs and output artifacts. For the preview formats supported here, see [Understanding Artifacts](/docs/core-concepts/artifacts).
8181
- **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.
82-
- **Logs** — the task's container logs, alongside its execution events. **Pop out** moves the logs into their own window you can dock or resize, and a second control opens them in a new browser tab. Subgraph tasks have no logs of their own, so the tab is not shown for them.
82+
- **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 `Esc` to 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.
8383

8484
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.
8585

86-
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. `Cmd/Alt` + `C` does the same from the keyboard.
86+
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](/docs/user-guide/pipeline-editor#copying-and-pasting-nodes).
8787

8888
## Subgraphs
8989

0 commit comments

Comments
 (0)