Combine

Use combine when two or more branches describe the same entities. Connect the branches directly to this node and set a shared record key such as website or _carry.name.

Input and output

The node reads records from directly connected producers, waits for all branches, and groups records by the string form of the configured key. Empty keys and object/list keys are excluded and counted as unkeyed. Dotted key paths are supported; matching is not fuzzy and does not automatically normalize domains or names.

Output records appear in extraction_result.data and llm_result.data. Each carries _branches, the ids of contributing producers. combine_result contains keys, unkeyed, and record counts in by_branch.

Configuration

combine configuration

{
  "key": "website",
  "fields": {"description": "last", "tags": "union", "notes": "join"},
  "separator": "\n\n"
}

Rule	Behavior
first	Default. Keep the first nonempty value
last	Keep the last nonempty value
union	Collect unique values into a list
join	Join nonempty values as text with separator

key is required. fields maps top-level output field names to rules. separator defaults to two newlines and is limited to 20 characters. Null, empty strings, empty arrays, and empty objects do not overwrite a useful value.

Verify the merge

For inputs {"website":"acme.test","tags":["solar"]} and {"website":"acme.test","tags":["storage"]}, tags: union produces one record with both tags. A record with website: null increases unkeyed instead of silently creating an entity.

Check branch counts and sample merged records, then connect Table or MongoDB Store. Use For Each when a branch needs a separate process for each record.