How can I pass secondary data through an n8n node without processing it?

Below is our current recommendation based on how data is passed through n8n nodes:

What’s Happening Today

n8n nodes work on a per-item basis. In most cases, a node (like an OCR node) won’t automatically propagate extra fields from the original trigger if those aren’t part of the node’s processed output. That’s why your secondary data (for example, an explanatory text or phone number attached at the trigger) stays behind.

Some users have mentioned a “passthroughData” approach. Although you might hear the term in discussion forums or roadmap items, there currently isn’t an officially documented or out‐of‐the‐box way to simply enable “pass data without processing” on any node.

Workarounds in the Current Setup

1. Merge Node (with Custom Restructuring):
   The common workaround is to combine the OCR output with the original secondary data via the Merge node or a similar node. A merge node can combine two streams, but—as you’ve noticed—its default output may generate a JSON structure that is too nested or unwieldy for downstream processing (for instance, by your AI agent).

   To improve this, you can follow up with a Set node or a [Code node] to reshape the JSON as needed. For example, after merging, a Code node could consolidate keys from both sources into a single, flatter object:

   JAVASCRIPTCopied
   // Assuming each merged item has .ocrData and .secondaryData properties:
   return items.map(item => {
     return {
       json: {
         // Adjust these keys to what your agent expects.
         extractedText: item.json.ocrData.text,
         photo: item.json.ocrData.photo,
         explanation: item.json.secondaryData.explanation,
         phoneNumber: item.json.secondaryData.phoneNumber,
       },
     };
   });

2. Dual Branch Processing:
   Another pattern is to split your workflow into two branches. One branch sends the photo for OCR processing, and the other retains the secondary data. Later, you can use a node (like a Function or Set node) to manually combine the results into a single item structure for the AI agent. This gives you full control over the output JSON format.

Looking Ahead: The “PassthroughData” Idea

The idea behind a native “passthroughData” feature would be to have nodes automatically carry over any data that isn’t explicitly handled by that node. Some parts of the documentation—for example, the metadata passed when using the Chat Trigger node—show how extra information can be included, but that functionality isn’t generalized across all nodes yet.

For instance, when embedding the Chat Trigger node in a website, you can pass arbitrary metadata:

JAVASCRIPTCopied
createChat({
	webhookUrl: 'YOUR_PRODUCTION_WEBHOOK_URL',
	metadata: {
		'YOUR_KEY': 'YOUR_DATA'
	}
});

Here the extra data is forwarded alongside the main payload. Unfortunately, no equivalent built‐in option exists for nodes like the OCR node.

Summary

• Current Behavior: Nodes process only their defined input; extra secondary data won’t be passed along unless you explicitly merge it.
• Workarounds:
  • Use a Merge node (and then reshape the JSON)
  • Split your workflow into branches and manually combine data
• Passthrough Feature: While the idea exists (and is referenced in discussion), there’s no official documentation or default “passthroughData” option in nodes yet.

For now, using a combination of Merge (or dual branching) plus a restructuring step (via Code or Set nodes) is the best approach to ensure that your AI agent receives a clean, predictable JSON object that contains both OCR results and your secondary data.