Pipeline settings

This page describes the items and constraints shown under Settings at the top of the pipeline editor. After you change a value, select Save at the bottom of the window to apply it.

Scope

The settings window of a saved pipeline has four tabs: General, Schedule, Options, and Sharing and Permissions.

General tab

Edit the pipeline's identifying information.

Field	Description
Name	Shows the system name. It cannot be changed.
Alias	Enter the name shown to users. AI can suggest a value.
Description	Enter what the pipeline is used for. AI can suggest a value.
Tags	Enter tags used for search and classification. AI can suggest a value.
Topic	Enter one or more event topics for the event pipeline to subscribe to. Add rows with Add topic and remove them with Remove topic on each row. A name is at most 249 characters, uses only English letters, digits, periods (.), underscores (_), and hyphens (-), and the same topic cannot be entered twice.

Storage location

The storage location (collection/folder) is determined at the time of saving, not in this window. If you enter the editor from a collection/folder, the location is displayed as locked. If you enter the editor directly, select a collection when saving. To change its location, move it in the collection tree or make a copy.

For a new event pipeline, the save window and, after saving, Settings → General edit the same multi-topic value. The list and the editor's type badge show only the first topic and summarize the rest as first-topic +N, so check the full value in the settings window. An event pipeline subscribes to events on these topics and does not use the Schedule tab. A batch pipeline takes no topics and sets up recurring runs on the Schedule tab.

Schedule tab

Set up a regular execution schedule. This tab is hidden for an event pipeline. See Scheduling for how to enter the presets (Every minute, Every hour, Daily, Weekly, Monthly).

Options tab

Item	Description
Runtime Image	The execution environment image (Docker image) to use to run the pipeline. If left blank, the default image will be used.
Python Packages	Additional Python packages, one per line. Pin each new entry to an exact version with package-name==version.
Execution timeout (seconds)	Total time limit for a batch run. Blank means no limit. After the limit is reached, the current step finishes and no later step starts.

info

Use Python Packages to add Python libraries. They are installed when the pipeline pod starts, and all steps share the same Python environment. A batch pipeline with declared packages keeps its pod running even when runs are far apart.

Write new entries as package-name==version or package-name[extras]==version. Do not enter URLs, Git addresses, local paths, or installer command options. In the API, use the pipeline's options.packages array.

Saving checks the entry format and exact version pin. It does not check whether that version can be installed or whether it is compatible with packages already in the image. Installation may change existing package versions or cause installation or execution to fail. Validate the required version combination in the same image before applying it.

Runtime Image replaces the execution environment for the entire pipeline. The image must support the D.Hub pipeliner execution contract and the pipeline's execution mode; an arbitrary Python or model image is not sufficient. To build your own image, ask your administrator for a pipeliner base image and build environment compatible with the deployed version. Updating the default image does not automatically upgrade the pipeliner in your custom image.

Sharing and Permissions tab

Grant or change the pipeline's Owner, Editor, and Viewer roles for users and groups. The Sharing and Permissions tab appears only for a saved pipeline.

Individual node write mode

Set the write mode of each output, per output, on the code node's Options tab, not in the pipeline settings.

The modes you can select depend on the output target.

Output target	Mode	Behavior
Dataset	Append (append, default)	Adds the result rows after the existing data. Sending the same rows again does not remove duplicates.
Dataset	Overwrite (overwrite)	Replaces the entire table with the result rows. Existing rows that are not in the result are deleted.
Dataset	Overwrite partitions (overwrite_partitions)	Replaces the rows of the partitions included in the result. Partitions that are not included are kept.
Entity / Relation	Upsert (upsert, default)	If the identifier key matches, updates the columns you pass; if the key is new, adds the row. Existing rows that are not in the result are kept.
Entity / Relation	Insert (insert)	Adds new rows. Fails if an identifier key overlaps with an existing row or with another row in the same result.
Entity / Relation	Overwrite (overwrite)	Replaces the entire table with the result rows. If relations reference the entity, check the dependent relation handling setting.

For Overwrite partitions, specify one or more Partition columns. Enter column names and separate them with Enter or a comma. For example, if you specify year and month and return only September 2026 data, only the rows of that month are replaced. Rows you want to keep in the same partition must also be included in the result.

If the result has 0 rows, Append / Upsert / Insert / Overwrite partitions keep the existing rows. Overwrite empties the table, so make sure an ingestion failure is not returned as an empty result. An entity overwrite follows the dependent relation handling rules even when the result is empty. If you turn on Delete linked relation rows and clear reference columns, relations stored in a link table have their rows deleted, and relations stored in foreign-key columns have their reference columns cleared.

For entity and relation outputs of earlier pipelines, a stored append is shown as Upsert and a stored append-strict as Insert. Until you change the mode yourself, the existing setting value and run behavior are kept.

Configure incremental or full read mode for inputs on the same Options tab.