Data

Every step receives items and passes items on. An item is one record: a link, a product, an email. Most data problems come down to two questions: how many items reached this step, and where does the value you need sit inside them?

Reading order

1. Items: what an item is, and why a step runs once per item.
2. Mapping in the UI: drag a field from an earlier step into a setting.
3. Mapping with expressions: what {{ $input.title }} means, plus fallbacks, maths and helper functions.
4. Item linking: how values from earlier steps line up with the item being processed.
5. Code: when a mapping isn’t enough, reshape data with Python.

Four ways to shape a value

Pick the lightest one that does the job.

• Drag a field → You need a value exactly as an earlier step produced it. Example Drop title from Get Page Metadata into a message. writes {{ $input.title }} for you
• Expression → You need to combine, compare or tidy values inside one setting. Example A fallback title, or upper(trim(name)). {{ … }} in any text field
• Edit Fields → Later steps need clean, named fields to rely on. Example Add a text field filled from pageText before an AI step. one node, no code
• Code → The change takes many nodes otherwise: regroup, count, parse. Example Group orders by customer and total them. Python, in the browser

Nodes that change the number of items

Node	Items in → out
Split Out	One item holding a list → one item per list entry
Aggregate	Many items → one item holding a list
Filter	Many → only the ones that match
Limit	Many → the first or last few
Remove Duplicates	Many → no repeats

Next

Start with ItemsWhy a step runs once per item, in five minutes.

Practise itRecipes that move real data between steps.