Exporting and importing pipelines

Every pipeline can leave as one file and come back as a new draft, on this Trinium or another one.

Exporting

Export downloads a pipeline as a single .trinium.json file. It is in two places:

The builder's header exports what the canvas shows: your saved draft, or the version you are viewing. Unsaved changes are not in the file; save first if you want them.
A version's menu, on the Versions tab, exports that version.

Exporting needs the same right as editing the pipeline. A workspace that keeps its work in the workspace (its setting Members may copy pipelines out is off) lets only its owner or an admin export, just as it lets only them copy a pipeline out. Every export is written to the audit trail.

What the file holds

The canvas: every node with its settings and name, every wire, the Notes and Containers, and where everything sits.
The parameters the pipeline declares.
The connections, by name and kind: never a password, a key or an id. Each node names a connection by a short reference such as conn-1, listed with the connection's name and kind.
The names of uploaded files the pipeline reads, without their contents.
Removed values: a value that looks like a password or a key (an Authorization header typed into a REST node, for example) is left blank in the file and listed, so whoever imports it knows what to enter again.

It does not hold the pipeline's owner, workspace, schedules, run history, earlier versions, server tag or time limit.

Check what you typed before you share a file: a password written inside a SQL statement, an email template or a link cannot be told from ordinary text, so it is not removed.

Importing

Import, beside New Pipeline on the Pipelines page, reads a file and shows what it holds before it creates anything:

Choose the file.
Name the new pipeline and choose where it lives: private, or a workspace you own or hold an Edit seat in.
Pick one of your connections for each connection the file names. A connection with the same name and kind is picked for you when there is exactly one. Leave one empty to pick it on the node later.
Upload the files it lists to the new pipeline's file library, and enter any removed values again.

The pipeline arrives as a draft, with no versions, ready to test and deploy like any other.

Import reads files Trinium wrote. It refuses a file from a newer version of Trinium than this one, because settings that version added would be lost; upgrade first. A file from an older version imports, and any setting this version no longer has is ignored, as it is on a pipeline saved before an upgrade.

What a file is for

A copy in your own Git, as a record of the pipeline.
A support ticket: attach the file instead of describing the pipeline.
Another Trinium, such as a second instance for another region or business unit.
Python: an AI can read the file and write the pipeline as PySpark or pandas code. See Turning a pipeline into Python.

A pipeline file runs only in Trinium. Moving a pipeline between Staging and Production does not need one: deploy a version instead (see Versions, Staging and Production).