Button 1 single click triggering on every wake up when prevent sleep is false

[syntesys]

I have a BTN-1 battery version with firmware 26.3.2.1.

In the home assistant blueprint, I setted on button 1 click a light toggle. The problem is that when I charge the battery with prevent sleep off, the device continuousy trigger wake up and button 1 pressed, toggling my light:

How can I prevent this situation? The prevent sleep in this mode is unusable.
Thank you!

[bharvey88]

Thanks for reaching out - this was an issue on the original version of the blueprint, but was solved when we pushed an update a few months back. Please try re-importing the blueprint and it should stop doing the clicks when it wakes up!

If this does not fix your issue please follow up and we'll figure it out together.

[syntesys]

I instsalled the blueprint the day I opened the bug, I received and updated the BTN1 3 days ago.

[jeremywhelchel]

Confirming #41 on a separate installation. This is not the blueprint — the click is generated on the device.

[fwiw Claude investigated and helped compile this report]

Setup: 4× BTN-1 (battery), firmware 25.12.5.1, HA 2026.8.1. BTN-1 blueprint re-imported from main today (d5e71c0c). Ordinary deep-sleep wakes, not charging, Prevent Sleep off.

Integrations/ESPHome/Core.yaml fires a synthetic click after every API connect:

  api:
    on_client_connected:
      then:
        - delay: 5s
        - lambda: |-
            std::string wake_button = id(wakeup_button_pressed).state;
            if (wake_button != "0" && wake_button != "") {
              if (wake_button == "1") {
                id(button_1_event).trigger("click");
              } ...
              id(wakeup_button_pressed).publish_state("0");
            }

Four consecutive wakes each produced a click 4.909 / 4.949 / 5.012 / 4.949 s after the ESPHome handshake — matching delay: 5s. Two of the four had nobody near the device.

Since trigger("click") runs on the device, HA receives a genuine event update with a fresh timestamp and valid event_type, indistinguishable from a real press. That's why re-importing the blueprint doesn't help. The blueprint's old_state.state !='unavailable' guard does fix a real but different bug — spurious events on availability transitions after an HA restart, which I also hit — but it cannot see this one.

Suspected cause: on a timer wake esp_sleep_get_ext1_wakeup_status() returns 0, and the on_boot else branch only clears the latch when it's empty:

  else {
    if (id(wakeup_button_pressed).state == "") {
      id(wakeup_button_pressed).publish_state("0");
    }
  }

So either a stale wakeup_button_pressed survives the sleep, or ext1 is spuriously reporting GPIO2. Every phantom I captured was button 1. Not fixed in 26.3.2.1: diffing Core.yaml between tags 25.12.5.1 and 26.3.2.1, the wake-detection lambda and on_client_connected block are byte-identical — consistent with @syntesys still seeing it after updating.

Scope: #41 reports this while charging. I see it on ordinary deep-sleep wakes, so it isn't charging-specific.

[syntesys]

@jeremywhelchel did you resolve it?
@bharvey88 I confirm that the problem persists.

[jeremywhelchel]

@syntesys nope. i didnt actually try to fix it