Item linking

In 20 seconds

While a step works on one item, {{ $input }} and a reference to a directly connected step both point at the matching item. A reference to a step further back returns that step’s whole output, not the matching item. To keep a title next to its URL, carry the field forward on the item, join the lists again with Merge on a shared field, or process one item at a time in a lambda workflow.

See it happen

Three product links go through Get All Text from Link, then Summarization summarises each page. Get All Text from Link keeps url because Include input fields is on. Summarization outputs only summary, so the link is gone after it. Click each step to see where url disappears.

Example run: Split Out, 3 items → Get All Text from Link, ran 3 times, 3 items → Summarization, ran 3 times, 3 items. Illustration of a run. The ways to keep url are below.

1. Split Out 3 items →
2. Get All Text from Link ran 3× 3 items Include input fields: on →
3. Summarization ran 3× 3 items outputs summary only

Output of Summarization Table JSON

#	summary
0	A light trail running shoe.
1	A waterproof jacket.
2	A bright head torch.

Illustration of a run. The ways to keep url are below.

Which item does a reference return?

You write…	While the step works on item 2, you get…
{{ $input.url }}	url of item 2, the item this run received.
{{ $('Get All Text from Link').url }} in a step directly connected to it	url of the item from that step used in this run.
{{ $('Split Out').url }} in a step further down	Empty: that step’s whole output is a list, and a list has no url field.
{{ $('Split Out').1.url }}	url of the second item of that list (counting starts at 0), whatever item is being processed.

In a step’s settings, the input panel shows the same split: steps under Direct inputs give you the matching item, steps under Relatives give you their whole output.

Three ways to keep values together

• Carry it forward Copy the field onto the item before a step that would drop it, or turn on a node's option to keep incoming fields. Example Include Other Fields on Split Out, Include input fields on Get All Text from Link. simplest, when the nodes allow it
• Join on a key → Two branches produce lists you need to line up again. Example Merge products and prices where url equals url. Merge · Inner / Left Join
• One item at a time → A chain of steps drops your fields and has no key to join on. Example A lambda workflow takes url, summarises the page and returns url + summary. Lambda Workflow

The third option works because a lambda workflow runs once per item: inside it, every earlier step holds a single item, so {{ $('Lambda Input').url }} reads the right value anywhere.

Add a stable field such as id or url before a workflow splits into branches. It gives Merge something reliable to match on later.

You’ll notice this when…

Every row shows the same title

The title comes from a step that isn’t directly connected, or from a step that output a single item, so every run reads the same value. Carry the title forward on each item and read it with $input.

A reference to an earlier step comes out empty

That step is not directly connected, so the reference reads its whole output. If the output is a list, use an index (.0.title) or carry the field forward instead. Also check the name: $('…') must match the node label exactly, including capitals and spaces.

After a Merge, fields are called left-title and right-title

Both inputs had a title field with different values, so Merge kept both and renamed them. Rename one side before the Merge, or use the prefixed names.

Next

Next concept: CodeReshape items with Python when a mapping isn't enough.

Practise itRecipes that read several pages and keep each result with its source.