Why Does the Loop Over Items Node Emit the DONE Output Before Completing All Loop Iterations?

Below is an explanation of what’s happening and what the expected behavior is:

What’s Happening

In the current workflow the Loop Over Items node appears to process a batch and immediately send a “DONE” output along with that batch’s data before processing the next batch. In your case you can see the following sequence in the execution (as seen in the GIF):

1. The node sends the aggregated “split out” data via the DONE branch.
2. Then it runs the first loop iteration (loop branch #1) and returns its data.
3. It then sends the DONE branch output for the first batch.
4. Next the second loop iteration (loop branch #2) is executed and so on…

This interleaved behavior (alternating between loop iterations and sending DONE outputs) can be confusing when you expect a single final aggregated result.

What’s Expected

The intended design of the Loop Over Items node is to:

1. Process all batches (each batch being sent through the loop branch for further node processing).
2. After the final loop iteration has been processed, combine (or “aggregate”) all the results from the loop iterations.
3. Finally, output that combined result through the DONE branch once the looping is completely finished.

That means that if you want to work with the entire aggregated JSON from all the loop iterations, you should wait for (or exclusively use) the DONE output which will only contain the combined data after every iteration has completed.

How to Address This in Your Workflow

If you require subsequent nodes in your workflow to operate on the complete aggregated result rather than on each batch individually, you should:

• Connect your next processing steps to the DONE branch of the Loop Over Items node.
• Ensure that any node expecting the final combined output only uses data from the DONE branch.
• Keep in mind that although the node internally processes each batch sequentially, the final “DONE” data is only meant to be used after every batch has been processed.

For more details, refer to the documentation on the Loop Over Items (Split in Batches) node which explains:

• The node temporarily saves incoming data.
• It outputs a predefined number of items (as per your Batch Size) via the loop branch.
• Upon completion of all iterations, it provides the complete aggregated result via the done branch.

By restructuring your workflow to use only the DONE branch output for follow-up processing, you should achieve the expected flow: first fully completing all loops and then receiving the final aggregated JSON as output.

If you have further questions or need additional adjustments to your workflow, let me know!