# IA & micro dans Gladys plus

**URL:** <https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400>\
**Category:** Configuration\
**Tags:** gladys-plus\
**Created:** [October 24, 2026, 9:53am UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400 "2026-10-24T09:53:28Z")\
**Posts on this page:** 16\
**Page:** 1

<div class="post-metadata">

**Author:** ![Damien](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/damien/32/22156_2.png) [@Damien](https://community.gladysassistant.com/u/Damien)\
**Post date:** [October 24, 2026, 9:53am UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/1 "2026-10-24T09:53:28Z")

</div>

Hello,

I am taking the liberty of creating a topic to share my questions and suggestions after a few days of using this very practical service:

- Can the default AI be modified? (in the discussion tab and/or Telegram, the « auto » model never answers my simple questions)
- Can the microphone be manually turned off? (if there is background noise, it often doesn’t stop)
- Suggestion: Put users, Gladys Plus user management, and remote users in the same tab (or make it easier to create a new user)

Thanks!

---

<div class="post-metadata">

**Author:** ![mutmut](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/mutmut/32/14064_2.png) [@mutmut](https://community.gladysassistant.com/u/mutmut)\
**Post date:** [October 24, 2026, 4:59pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/2 "2026-10-24T16:59:37Z")

</div>

@Damien I unfortunately won’t be able to answer all of that, but thanks for the questions 🙂  
However, I’ve modified your title and category to make it clearer.

And you should make a feature request for your suggestion, that way it separates the help you might need and future functions.

---

<div class="post-metadata">

**Author:** ![Damien](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/damien/32/22156_2.png) [@Damien](https://community.gladysassistant.com/u/Damien)\
**Post date:** [October 30, 2026, 11:34am UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/3 "2026-10-30T11:34:49Z")

</div>

I have a suggestion:

- be able to change the size of the «&nbsp;voice assistant&nbsp;» widget 🙂

---

<div class="post-metadata">

**Author:** ![pierre-gilles](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/pierre-gilles/32/5886_2.png) [@pierre-gilles](https://community.gladysassistant.com/u/pierre-gilles)\
**Post date:** [October 30, 2026, 1:08pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/4 "2026-10-30T13:08:19Z")

</div>

Hi @Damien,

Thanks for your feedback! I’ll address each point one by one. I think these are mostly settings that aren’t necessarily easy to find 🙂

**1. Change the default AI**

I’d like to understand what you mean by « never respond ». Can you give me 2-3 examples of questions, and exactly what Gladys responds? Nothing at all, the message « The request to the Artificial Intelligence API failed », or an off-topic answer?

**2. Stop the microphone manually**

Yes, it’s possible: during listening, the round button of the widget displays **« Stop »** , just click on it (same with the microphone button of the Discussion tab). Be careful, this **cancels** the request: what has been recorded is not sent.

Otherwise, the recording stops automatically as soon as you stay silent for more than a second. In all cases, it stops after 30 seconds, and then what you said is sent. Gladys measures the ambient noise before you speak to adapt to it. But a strong and continuous background noise (TV, music, hood…) can prevent it from detecting the silence, hence what you observe. Speaking closer to the microphone helps a lot.

**3. The users**

I understand the confusion, but these are 3 different things:

- **Settings → Users** : the accounts of your Gladys instance, at home (name, language, role, Telegram…). This is where you create a new user.
- **Settings → Plus Users** (visible only when you go through [plus.gladysassistant.com](http://plus.gladysassistant.com)): the Gladys Plus accounts, i.e. the people who can access remotely. This is where you invite someone by email.
- **Settings → Gladys Plus → Remote Users** : the security step. You validate, from home, the encryption keys of each Gladys Plus user. This is intentionally separate: even our servers cannot give access to your home without this local validation.

To give remote access to someone, the order is therefore:

1. Create their account in **Users**.
2. Invite them in **Plus Users**.
3. Once their Gladys Plus account is created, authorize them locally in **Gladys Plus → Remote Users**.
4. At their first connection to Gladys Plus, they choose which Gladys user to link their account to.

Maybe you wanted to say that we could improve the user experience of adding a user?

**4. Size of the « Voice Assistant » widget**

The width is already adjustable with the dashboard layout. In edit mode, you can create **sections** of 1 to 6 columns, and make a column « wide » with the button next to the trash can, at the top of the column. To have a smaller widget, put it in a section with more columns, or in a normal column next to a wide column. The height, however, is not adjustable for now.

Let me know if this answers your questions!

---

<div class="post-metadata">

**Author:** ![Damien](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/damien/32/22156_2.png) [@Damien](https://community.gladysassistant.com/u/Damien)\
**Post date:** [October 30, 2026, 1:36pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/5 "2026-10-30T13:36:14Z")

</div>

Thank you for taking the time to look at these points in detail.

1. Default AI. Here is an example:

 ![image](https://gladys-assistant-community.b-cdn.net/uploads/default/original/3X/c/8/c8ccc2ef0b22619fcbe6ffc2961b04cb3fb9edc6.png)

The «&nbsp;auto&nbsp;» model most often responds to a simple temperature or humidity request, where other models always respond precisely (I have IKEA timmerflotte sensors).

1. Manually stopping the microphone

Indeed, I would like to be able to press the stop button when I finish talking to it. I will try to speak closer to the microphone to avoid ambient noise.

1. Users

With your clarifications, I understand the differences between these different pages well, but indeed, I think the experience when adding a user can be optimized.

1. Size of the «&nbsp;Voice Assistant&nbsp;» widget: you answered perfectly, thank you!

---

<div class="post-metadata">

**Author:** ![pierre-gilles](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/pierre-gilles/32/5886_2.png) [@pierre-gilles](https://community.gladysassistant.com/u/pierre-gilles)\
**Post date:** [October 30, 2026, 1:45pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/6 "2026-10-30T13:45:33Z")

</div>

Thanks for the screenshot, it’s perfect: you can see exactly what’s happening, and it’s not you, it’s a bug on our end 🙂

**1. The AI in « Auto »**

In your screenshot, the AI correctly calls the tool that reads your sensors, but with `"room": "No room"`. « No room » is a special room in Gladys: it groups devices that are not placed in any room. The model used in « Auto » understands this as « no specific room = the whole house ». Gladys therefore only looks among devices without a room. It doesn’t find any humidity sensors there, and the AI concludes that there are none in the house. The other models leave this field empty, and correctly read all your sensors.

I will fix this on Gladys’ side so that this trap no longer exists. In the meantime, you can:

- choose another model in the « AI Model » menu (be careful, it reverts to « Auto » when you reload the page, and Telegram always stays in « Auto ») ;
- specify the room in your question, for example « what is the humidity in the living room ? ».

**2. The « Stop » button**

Good point: when you press « Stop » during listening, it would make more sense to send what you said rather than cancel everything. I’ll look into changing that.

**3. Users**

Great, glad it’s clearer! For the improvement of adding a user, you can create a feature request, as suggested by @mutmut, by describing how you would have liked it to work? This allows others to discuss and vote.

**4. Widget size**

Perfect 👍

---

<div class="post-metadata">

**Author:** ![Damien](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/damien/32/22156_2.png) [@Damien](https://community.gladysassistant.com/u/Damien)\
**Post date:** [October 30, 2026, 1:48pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/7 "2026-10-30T13:48:41Z")

</div>

> [@pierre-gilles](#):
>
> - choose another model in the « AI Model » menu (be careful, it reverts to « Auto » when you reload the page, and Telegram remains always on « Auto ») ;

that was the initial suggestion: that we could choose the default AI model from those offered with Gladys in the settings

---

<div class="post-metadata">

**Author:** ![cursor-bot](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/cursor-bot/32/22252_2.png) [@cursor-bot](https://community.gladysassistant.com/u/cursor-bot)\
**Post date:** [October 30, 2026, 1:56pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/8 "2026-10-30T13:56:54Z")

</div>

Hi everyone!

This topic is now **in development**.

A PR has been opened to send the voice recording when you press the button while listening:

> <https://github.com/GladysAssistant/Gladys/pull/3292>
>
> \### Description
> 
> With background noise, the voice recording never detects the en…d of speech, so it keeps listening until the 30-second cap. Pressing the button while listening used to \*\*drop\*\* the recording. Now it \*\*ends the recording and sends what was said\*\*, as if silence had been detected.
> 
> \*\*Recorder (\`front/src/utils/speechCommandRecorder.js\`)\*\*
> \- New \`stopSignal\` option (an \`AbortSignal\`). When it fires, the recording ends right away and the audio is kept. It works on both paths: AudioWorklet (main-thread \`flush\`) and ScriptProcessor (\`finish('manual')\`). The existing \`signal\` still drops the audio.
> \- If no speech was detected yet, a stop still rejects with \`NO\_SPEECH\`, as before.
> \- Small fixes in the same code:
> - A signal that is already aborted by the time the audio graph is ready (for example aborted while \`getUserMedia\` was pending) is now handled. Before, its \`abort\` event had already fired and was missed, so the mic stayed open.
> - \`NO\_SPEECH\` from the AudioWorklet path no longer falls back to a second ScriptProcessor recording. Only an AudioWorklet failure falls back now. Measured: with \`maxWaitForSpeechMs: 1500\` and a silent mic, the old code rejected after about 3.3 s; it now rejects after 1.5 s.
> 
> \*\*Voice assistant widget (\`VoiceAssistantBox.jsx\`)\*\*
> \- While listening, the button reads \*\*"Send"\*\* ("Envoyer") and a click sends the recording (state goes to \*Processing\* right away).
> \- While processing or speaking, a click still cancels ("Stop"), as before.
> \- The listening hint now says the user can also press Send.
> 
> \*\*Chat dictation button (\`ChatVoiceInputButton.jsx\`)\*\*
> \- Same behaviour. A click while listening transcribes what was said into the input. A click while the transcription runs cancels it; its label is now "Cancel dictation".
> 
> \*\*i18n\*\*: new keys \`dashboard.boxes.voice-assistant.send\` and \`chat.voiceInput.cancelListening\`, plus an updated \`listeningHint\`, in en / fr / de / es.
> 
> \### Testing
> 
> Run in Chromium (Playwright) against the demo front (\`npm run start-demo\`), with a fake microphone fed by generated WAV files. "Babble" is non-stationary noise that the browser's noise suppression does not remove, so the end of speech is never detected:
> 
> | Scenario | Before | After |
> |---|---|---|
> | Babble, stop requested at 2.5 s | records \*\*30.00 s\*\* (cap) | records \*\*2.46 s\*\*, sent |
> | Babble, stop at 2.5 s, ScriptProcessor fallback forced | n/a | 2.41 s, sent |
> | Babble, abort at 2.5 s (both paths) | AbortError | AbortError (unchanged) |
> | Silent mic, stop at 2.5 s (both paths) | n/a | \`NO\_SPEECH\` |
> 
> UI, with \`postBinary\` stubbed on the demo HTTP client:
> \- \*\*Widget\*\*: Speak → Send (listening hint shown) → click → Stop / Processing… → transcription and answer displayed. A third click while processing cancels and shows nothing. On a silent mic, Send shows "No voice detected".
> \- \*\*Chat\*\*: mic → Stop dictation → click → spinner (Cancel dictation) → the input is filled with the transcription. A click during the spinner cancels and leaves the input empty. On a silent mic, Send shows the no-voice error.
> 
> Local checks: \`npm run prettier-check\`, \`npm run eslint\` (no warnings on the touched files), \`npm run compare-translations\` and \`npm run build\` all pass in \`front/\`. Server untouched. No Cypress spec uses the voice widget or the dictation button (the button only renders when Gladys Plus is configured). \`login/Login.cy.js\` only checks the redirect to \`/dashboard/chat\`.
> 
> \### Related request
> 
> Forum: https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400
> 
> \### Checklist
> 
> \- \[x\] If a forum topic or GitHub issue exists, the description links it (\`Forum: https://community.gladysassistant.com/t/...\` or \`Closes #...\`)
> \- \[x\] Tests pass: server untouched (no coverage change); the UI change was tested manually in a browser as described above
> \- \[x\] Linter and prettier pass on front (\`npm run eslint\`, \`npm run prettier\`)
> \- \[x\] No undocumented breaking change
> 
> \---
> \_Generated by \[Claude Code\](https://claude.ai/code/session\_01MNjB35ewWNZCTZKra9LYGC)\_

Feel free to follow the PR, test (optional, especially for small requests) and give your feedback here if needed.

---

<div class="post-metadata">

**Author:** ![cursor-bot](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/cursor-bot/32/22252_2.png) [@cursor-bot](https://community.gladysassistant.com/u/cursor-bot)\
**Post date:** [October 30, 2026, 2:00pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/9 "2026-10-30T14:00:00Z")

</div>

Hi everyone!

This topic is now **in development**.

A PR has been opened to stop «&nbsp;whole house&nbsp;» readings from filtering on «&nbsp;No room&nbsp;»:

> <https://github.com/GladysAssistant/Gladys/pull/3293>
>
> \### Description
> 
> With the Gladys Plus model selector on "Auto", "quels taux d'hu…midité dans la maison ?" was mostly answered with "Il n'y a pas de capteur d'humidité dans la maison". The model called:
> 
> \`\`\`
> device\_get\_state({"device\_type":"humidity-sensor","room":"No room"})
> \`\`\`
> 
> The \`noRoom\` sentinel (\`No room\` / \`no-room\`) was part of every \`room\` enum given to the model. A smaller model reads it as "no room filter, the whole home". The callback resolved it to the \`no-room\` selector and kept only the devices that are not assigned to any room. It found no humidity sensor among them and answered \`no device of type "humidity-sensor" is configured in room "No room"\`.
> 
> \*\*Read-only tools\*\* (\`device.get-state\`, \`device.get-battery-levels\`, \`device.get-history\`):
> \- \`No room\` is no longer offered in their \`room\` enum. In a home without any room, the parameter becomes a plain string instead of an empty enum. zod renders an empty enum as \`not: {}\`, which matches nothing.
> \- The callbacks read \`No room\`, or a name close to it (through \`findBySimilarity\`), as the whole home. The chat gateway runs these callbacks with the model's raw arguments, which the enums do not bind. Reading more than asked is harmless for a read. Every result still carries each device's room, so the devices not assigned to a room still appear with \`room: "No room"\`.
> \- The "not a room of this home" message reuses the new \`realRoomNames\` list instead of filtering the sentinel out inline.
> 
> \*\*Camera and control tools\*\* (\`camera.get-image\`, \`device.turn-on-off\`, \`device.set-shutter\`, \`device.set-light\`):
> \- They keep \`No room\`. It is how these tools reach the devices not assigned to a room by room, and a command or a camera image must not widen to the whole home on a guess. (\`getImagesInRoom\` is per room anyway.)
> \- Their \`room\` description now says: \`"No room" only targets the devices that are not assigned to a room, never the whole home.\` This addresses the same misreading without changing what they do.
> 
> \*\*Tests\*\* (\`server/test/services/mcp/lib/buildSchemas.test.js\`):
> \- The exact call \`{ device\_type: 'humidity-sensor', room: 'No room' }\` returns the humidity sensors assigned to rooms.
> \- \`no-room\` (close to the sentinel) reads the whole home, and the unassigned device is labelled \`No room\`.
> \- An empty result for the whole home no longer says \`in room "No room"\`.
> \- \`device.get-history\` with \`room: 'No room'\` reads the first matching device of the home.
> \- \`device.get-battery-levels\` with \`room: 'No room'\` returns every battery.
> \- A home without rooms gets a string \`room\` parameter.
> \- The enums of the read-only tools leave out \`No room\`, while the camera and control tools keep it with the new description.
> \- The new tests fail on the previous code, including when only the callback guard is removed.
> 
> \`npm run prettier-check\` and \`npm run eslint\` pass (0 errors). In \`npm run coverage\`, all MCP tests pass and every changed line of \`buildSchemas.js\` is covered. The only uncovered lines in that file (875–878) predate this change. 21 gateway / backup / integrationHost tests fail in my sandbox, and the 20 gateway ones fail the same way on the previous commit: the sandbox's outbound proxy answers their Gladys Plus HTTP calls with a 405. None of them touch MCP code.
> 
> \### Related request
> 
> Forum: https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400
> 
> \### Checklist
> 
> \- \[x\] If a forum topic or GitHub issue exists, the description links it (\`Forum: https://community.gladysassistant.com/t/...\` or \`Closes #...\`)
> \- \[x\] Tests pass: \`cd server && npm run coverage\` (Codecov requires 100% coverage on changed lines) and Cypress (\`npm run cypress:run\`) if the UI changed. Server only, no UI change. Failures in my sandbox are network-related and described above.
> \- \[x\] Linter and prettier pass on both front and server (\`npm run eslint\`, \`npm run prettier\`). Server only, front untouched.
> \- \[x\] No undocumented breaking change
> 
> \---
> \_Generated by \[Claude Code\](https://claude.ai/code/session\_01MNjB367DJsw9vMFdS8cY2c)\_

Don’t hesitate to follow the PR, test (optional, especially for small requests) and give your feedback here if needed.

---

<div class="post-metadata">

**Author:** ![pierre-gilles](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/pierre-gilles/32/5886_2.png) [@pierre-gilles](https://community.gladysassistant.com/u/pierre-gilles)\
**Post date:** [October 30, 2026, 2:00pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/10 "2026-10-30T14:00:09Z")

</div>

I understand the idea, but I don’t plan to add this setting, and it’s intentional: the «&nbsp;Auto&nbsp;» mode exists to optimize the cost and speed of requests.

In «&nbsp;Auto&nbsp;», Gladys Plus uses:

- `mistral-small-3.2-24b-instruct-2506` for simple queries (turning on a light, reading a sensor, asking for the weather…);
- `qwen3.5-397b-a17b` for creating scenes, which requires more reasoning.

Setting another model as default would mean using a more expensive and slower model to turn on a simple light bulb, whereas Mistral Small does this very well.

Here’s the pricing grid that I pay for inference:

 ![Capture d’écran 2026-10-30 à 14.59.16](https://gladys-assistant-community.b-cdn.net/uploads/default/original/3X/5/8/587deee31e35c7ab76f44a717b935c01d9d6bfe4.png)

Source: [Model-as-a-service Solutions Pricing | Scaleway](https://www.scaleway.com/en/pricing/model-as-a-service/)

The Plus subscription costs €8.33 per month, and it also funds remote access, backups, and voice. Routing all requests through a more expensive model is not economically sustainable: it would require increasing the price of the Plus subscription!

The other models in the menu are mainly there for testing: preparing the next «&nbsp;default&nbsp;» model if prices change, or comparing the behavior of several models when looking for a bug. If I suggested choosing a more powerful one earlier, it was only to help you out until the fix. The goal remains to have «&nbsp;Auto&nbsp;» as the sole default model.

And in your case, it’s not a lack of model power: it’s the «&nbsp;No room&nbsp;» bug I mentioned earlier. Once fixed, the «&nbsp;Auto&nbsp;» mode should correctly answer your questions about temperature and humidity.

If you still want a more powerful model, it’s possible since version 5.3 with your own API key. Install an AI provider integration (there’s already a Claude integration in the catalog, for example), then select it in **Integrations → Artificial Intelligence → AI Provider**. You then choose the model you want in the integration settings, and it is used everywhere: the Discussion, Telegram, and the «&nbsp;Ask AI&nbsp;» scenes. Requests are then billed directly by this provider.

---

<div class="post-metadata">

**Author:** ![StephaneB](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/stephaneb/32/15443_2.png) [@StephaneB](https://community.gladysassistant.com/u/StephaneB)\
**Post date:** [October 30, 2026, 5:00pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/11 "2026-10-30T17:00:03Z")

</div>

@pierre-gilles, re-reading that…

> [@pierre-gilles](#):
>
> To grant someone remote access, the order is:
> 
> 1. Create their account in **Users**.
> 2. Invite them in **Users Plus**.
> 3. Once their Gladys Plus account is created, authorize them locally in **Gladys Plus → Remote Users**.
> 4. On their first connection to Gladys Plus, they choose which Gladys user to link their account to.

… step 4 surprises me, and I had never paid attention to it: does this mean that a Gladys Plus user connecting for the first time can choose to use any local Gladys account rather than the one I created for them? Shouldn’t this be chosen by the admin in step 2?

---

<div class="post-metadata">

**Author:** ![pierre-gilles](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/pierre-gilles/32/5886_2.png) [@pierre-gilles](https://community.gladysassistant.com/u/pierre-gilles)\
**Post date:** [October 30, 2026, 5:12pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/12 "2026-10-30T17:12:34Z")

</div>

> [@StephaneB](#):
>
> … step 4 surprises me, and I had never paid attention to it: does this mean that a Gladys Plus user who logs in for the first time can choose to use any local Gladys account?

Currently yes, the reasoning was that on the instance you «&nbsp;trust&nbsp;» others in your family, but it’s true that we can do better!

> [@StephaneB](#):
>
> Shouldn’t this be chosen by the admin in step 2?

That’s a very good idea! 😃

---

<div class="post-metadata">

**Author:** ![StephaneB](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/stephaneb/32/15443_2.png) [@StephaneB](https://community.gladysassistant.com/u/StephaneB)\
**Post date:** [October 30, 2026, 11:11pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/13 "2026-10-30T23:11:09Z")

</div>

Do you need me to make a feature request for that?

---

<div class="post-metadata">

**Author:** ![pierre-gilles](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/pierre-gilles/32/5886_2.png) [@pierre-gilles](https://community.gladysassistant.com/u/pierre-gilles)\
**Post date:** [November 2, 2026, 9:21am UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/14 "2026-11-02T09:21:08Z")

</div>

No, no need, I’ll take care of it this morning 🙂

Anything that can help improve the understanding/UX of Gladys Plus is super important feedback!

---

<div class="post-metadata">

**Author:** ![StephaneB](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/stephaneb/32/15443_2.png) [@StephaneB](https://community.gladysassistant.com/u/StephaneB)\
**Post date:** [November 2, 2026, 10:08am UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/15 "2026-11-02T10:08:30Z")

</div>

So I’ll take advantage of this. In Gladys Plus, I’ve already thought that we should:

- be notified when several consecutive nightly backups fail (one night without a backup can be a one-time issue and it’s not a big deal)
- have a way to reset the 2FA without having to go through you

---

<div class="post-metadata">

**Author:** ![pierre-gilles](https://gladys-assistant-community.b-cdn.net/user_avatar/community.gladysassistant.com/pierre-gilles/32/5886_2.png) [@pierre-gilles](https://community.gladysassistant.com/u/pierre-gilles)\
**Post date:** [November 2, 2026, 5:02pm UTC](https://community.gladysassistant.com/t/ia-micro-dans-gladys-plus/11400/16 "2026-11-02T17:02:43Z")

</div>

Everything is live in v5.3.1:

> [@Gladys Assistant 5.3.1: air conditioning, pilot wire and time slots](https://community.gladysassistant.com/t/gladys-assistant-5-3-1-la-climatisation-le-fil-pilote-et-les-plages-horaires/11485):
>
> Hi everyone! Version 5.3.1 of Gladys is available tada Don’t be fooled by the number: this isn’t just a bugfix version! Here’s what’s new: thermometer Thermostat controls air conditioning: air conditioners and heat pumps, multiple temperature sensors, and a delay before turning off when a window opensfire Pilot wire radiators are managed, with a new dedicated widgetbeach_with_umbrella Vacation scheduling from a scene: a new action to change the thermostat schedulethree_o_clock T…
