Items iteration

items: runs a resource once per entry in a list - like a for-each loop, but each iteration is a full resource execution with its own output.

Applies to both workflow mode and agent mode. In workflow mode it is a DAG step. In agent mode it runs inside the workflow when the LLM calls that workflow as a tool.

Basic usage

# resources/process-items.yaml

actionId: processItems
items:
  - "Item 1"
  - "Item 2"
  - "Item 3"

chat:
  prompt: "Process: {{ get('current') }}"

Item context

When processing items, special getters are available:

Getter	Description
get('current')	Current item value
get('prev')	Previous item (null if first)
get('next')	Next item (null if last)
get('index')	Current index (0-based)
get('count')	Total number of items
get('all')	Array of all items

The item object

You can also access item context through the item object with callable methods:

Method syntax

# resources/example.yaml
after:
  # Method-style access
  - set('curr', item.current())
  - set('prev', item.prev())
  - set('next', item.next())
  - set('idx', item.index())
  - set('cnt', item.count())
  - set('all', item.values())

Comparison: get() vs item.method()

get() Style	item.method() Style	Description
get('current')	item.current()	Current item
get('prev')	item.prev()	Previous item
get('next')	item.next()	Next item
get('index')	item.index()	Current index
get('count')	item.count()	Total items
get('all')	item.values()	All items array

Both syntaxes are equivalent. Use whichever is more readable for your use case.

Example: using item object

# resources/process-with-item-object.yaml
actionId: processWithItemObject
items:
  - "first"
  - "second"
  - "third"
after:
  # Using item object methods
  - set('position', "Item " + string(item.index() + 1) + " of " + string(item.count()))
  - set('hasPrevious', item.prev() != nil)
  - set('hasNext', item.next() != nil)
chat:
  prompt: |
    {{ get('position') }}
    Current: {{ item.current() }}
    {{ get('hasPrevious') ? 'After: ' + item.prev() : 'First item' }}

Accessing all item values

After processing, you can access all collected values from a resource that uses items:

Using item.values(actionID)

Pass an action ID to item.values() to get all iteration values from a specific resource:

# resources/collect-results.yaml
actionId: collectResults
requires:
  - processItems
after:
  # Get all values from processItems resource
  - set('allResults', item.values('processItems'))
  - set('resultCount', len(get('allResults')))

apiResponse:
  response:
    results: get('allResults')
    count: get('resultCount')

Note: item.values() without arguments returns all items for the current iteration context (equivalent to get('all')). With an action ID, it returns all values from that specific resource's items iteration.

Examples

Simple processing

# resources/example.yaml
items:
  - "apple"
  - "banana"
  - "cherry"

chat:
  prompt: |
    Item {{ get('index') + 1 }} of {{ get('count') }}: {{ get('current') }}
    Describe this fruit.

With context

# resources/example.yaml
items:
  - "Introduction"
  - "Main Content"
  - "Conclusion"

chat:
  prompt: |
    Write the {{ get('current') }} section.
    {{ get('prev') ? 'Previous section was: ' + get('prev') : 'This is the first section.' }}
    {{ get('next') ? 'Next section will be: ' + get('next') : 'This is the last section.' }}

Skip specific items

# resources/example.yaml
items:
  - "process"
  - "skip_this"
  - "process"

validations:
  skip:
  - get('current') == 'skip_this'

chat:
  prompt: "Processing: {{ get('current') }}"

Conditional processing

Items iterate over expressions that evaluate to strings. To filter, use skip on the value:

# resources/example.yaml
items:
  - "Task 1|high"
  - "Task 2|low"
  - "Task 3|high"

validations:
  skip:
    - "!(get('current') contains '|high')"

chat:
  prompt: "Handle high-priority task: {{ split(get('current'), '|')[0] }}"

Use cases

• Run the same LLM prompt over each row of a dataset (summarize each document, classify each ticket, translate each string).
• Generate a multi-section document one section at a time, using get('prev') and get('next') for continuity.
• Fan a fixed list of URLs, files, or IDs through an httpClient:, scraper:, or sql: resource and collect the results.

See also

• Items reference - dynamic items, performance, best practices
• Resources overview - Resource configuration
• Expressions - Expression syntax
• Expression functions reference - Complete function reference
• Python resource - For complex batch processing