Debounce Node

The debounce node implements trailing-edge debouncing: every incoming message resets a timer, and only the last message is processed after the timer expires with no new arrivals.

When to Use

Collapse rapid-fire user input (e.g. typing indicators, search-as-you-type)
Wait for a burst of updates to settle before broadcasting
Prevent redundant processing when many events arrive in quick succession

How It Works

A message arrives → it is parked (not forwarded) and the debounce timer is (re)started.
If another message arrives before the timer fires → the timer resets with the new message.
When the timer fires with no new arrivals → the last parked message is re-enqueued and runs through the pipeline again from the top.
On the second pass the debounce node detects the re-enqueued marker and passes through.

Debounce is in-memory per node (backed by ETS). State is local to the server instance and resets on restart.

Configuration

Field	Type	Required	Default	Description
`window`	number	Yes	—	Quiet period to wait
`unit`	string	No	`"ms"`	Time unit: `"ms"`, `"s"`, or `"m"`

Examples

Debounce with a 300 ms window (good for typing indicators):

{ "window": 300 }

Debounce with a 2 second window:

{ "window": 2, "unit": "s" }

Difference from Throttle

	Throttle	Debounce
Which message passes	First in window	Last after quiet period
Dropped messages	All after the first within window	All except the final one
Use case	Rate limiting	Collapse bursts

Next Steps

Throttle Node — limit message rate with leading-edge throttle
Filter Node — conditionally block messages