Parallelization Node: Editing Previous Message Doesn't Retrigger Parallel Models

[MathisVerstrepen]

When a user edits a previously sent message that triggered a Parallelization Node in the chat view, the regeneration process incorrectly re-triggers only the aggregator response. The underlying 'Parallel models' are not re-invoked, causing the aggregator to use stale or cached results instead of generating new ones based on the updated input. This leads to incorrect or outdated responses from the Parallelization Node.

Steps to Reproduce:

1. In the canvas view, create a 'Parallelization Node' setup.
2. Send the message to execute its parallel models and aggregator.
3. Inchat view, edit that previously sent message.
4. Observe the regeneration and parallel models generation output.

Expected Behavior:
Editing a previous message associated with a Parallelization Node should trigger a full regeneration, including re-invoking all underlying 'Parallel models' to ensure fresh, updated results are fed to the aggregator.

Observed Behavior:
Only the aggregator response is regenerated; the 'Parallel models' are bypassed, resulting in the aggregator using outdated information.