UniFI Vape Sensor & AI Speaker

[synaesthesia]

I should hope so - nothing against Verkada as it worked reasonably well, but very clearly and very vastly overinflated on the price front.

Easy to justify when so far competition has been pretty much just Halo and a slew of consumer level crud, but now not so much.

[Homer]

Got 3 of them on site for testing... will report back how we find them 

[BKGarry]

I have to say, I appreciate the Torx screw for securing to the metal mount, over the usual bit of plastic with the normal tool

[FragglePete]

I'd be interested to see if someone testing these can try this for me:  Spray some deoderant in the vacinity of these and does this cause an alarm? 

 

We tried a Vape Guardian sensor a while back and this could be triggered by an aerosol deoderant which caused all sorts of embarassement and hassle with a parent who we accused their daughter of vaping.  SMT spray a deoderant near the sensor and that caused an alert - the system was pretty much deemed useless from that point forward.  Vapeguardian claimed they could tweak the settings but by that point, the damage was done and no matter of tweaks made any difference.  If this Unifi solution is as good as it claims, I can see the likes of Vapeguardian not existing for much longer.

Pete

 

[Aprice]

21 hours ago, BKGarry said:

I have to say, I appreciate the Torx screw for securing to the metal mount, over the usual bit of plastic with the normal tool

Yeah, Still think it will need some additional protection on them. We fitted Verkada ones in a school and promptly removed them all the same day after two of them were ripped off and in a toilet.

 

 

[synaesthesia]

21 hours ago, FragglePete said:

I'd be interested to see if someone testing these can try this for me:  Spray some deoderant in the vacinity of these and does this cause an alarm? 

 

We tried a Vape Guardian sensor a while back and this could be triggered by an aerosol deoderant which caused all sorts of embarassement and hassle with a parent who we accused their daughter of vaping.  SMT spray a deoderant near the sensor and that caused an alert - the system was pretty much deemed useless from that point forward.  Vapeguardian claimed they could tweak the settings but by that point, the damage was done and no matter of tweaks made any difference.  If this Unifi solution is as good as it claims, I can see the likes of Vapeguardian not existing for much longer.

Pete

 

Yes. If you want it to. I'll pop some graphs on later if I get a chance, but this can clearly differentiate between vape and not. The vapeguardian sca^H^H^H awful things are just generic tuya nonsense which wouldn't tell the difference between a fart and a cloud of marijuana. 

 

[MYK-IT]

16 minutes ago, Aprice said:

Yeah, Still think it will need some additional protection on them. We fitted Verkada ones in a school and promptly removed them all the same day after two of them were ripped off and in a toilet.

 

 

Were you using the Verkada Reinforced SV Sensor Mounts? Which is what they have advised us to have. Admittedly, no good on suspended ceiling / tiles (as can be ripped off, as you mentioned) but we're installing to solid ceiling / wall etc.

Edited June 19 by MYK-IT

[synaesthesia]

When linked to a camera, surely it's easy to find the culprit and bill as appropriate? I think that's probably a better solution than making it into fort knox.

[synaesthesia]

Vape air quality level during testing:

 

TVOC level

 

 

Vape Confidence level

 

 

You can see a clear divide between TVOC and vape confidence.  The spikes in TVOC were from Lynx Black Body spray as well as 99.9% isoprop alcohol.

The vape confidence did not move during either, and only spiked when applying a "phat cloud" of Fanta Ice nicotine-free vape.

 

The only time I got a false positive was during soldering yesterday - I'm not sure whether it's actual solder fumes or the plastic melted onto the school DT department iron but that did flag up as vape.

It hung around a LOOOOOONG time. When I got home last night, I immediately ordered a better fume extraction system for my home workshop, as clearly fans on their own aren't enough. I don't and never have vaped outside of this testing, but I've been soldering since a kid and suspect my lungs are worse than some 20 a day folk!

 

[FragglePete]

Which suppliers in the UK have these in at the moment?  I usually use Broadband Buyer and/or 4Gon as we have accounts with them and the pricing has always been good with those guys, but they don't have them listed yet.  I can buy them directly from the EU Store with Ubiquiti but that may prove difficult with a school credit card, etc.

 

Pete

 

[gaz350b]

@FragglePete If your school allows you to prepay via invoice before goods then you can just use that option. takes a little longer as a process as they confirm payment before sending.

[Homer]

I sourced mine directly from Unifi - at the moment I've not managed to trigger the test one in the office with vaping (I tried - I hate it - I will need to get somebody else to test it) 

 

We did spray some body spray in the area of it though, no vape alarm but the other readings went waaaay high to show how crap it makes the air 

[synaesthesia]

Straight from unifi UK here, ordered friday, arrived monday

Homer - make sure to remove/adjust the vape confident level safety limit, i think it's very lenient by default

[synaesthesia]

Really struggling with the webhooks  documentation is between non-existent and dire, with the most detailed being this: https://help.ui.com/hc/en-us/articles/25478744592023-Send-UniFi-Protect-Alerts-to-Web-Services-using-Webhooks

 

All I'd like to do effectively is post to a google chat, and that doesn't seem like it should be too difficult - in the chat space, add integration, webhook, copy the URL and paste that into the URL of the webhook notifications panel in Protect.

Does nada though, and suspect it needs some sort of authentication somewhere.

 

The protect side of it clearly works, tried it with Slack and that just works, no questions asked. 

Very much out of my depth as it's software rather than something I can.... well, think Homer strangling Bart. I'll keep plugging through, notification via email will suffice at a push but I'd rather make it a bit more instant.

 

[dhicks]

1 hour ago, synaesthesia said:

All I'd like to do effectively is post to a google chat, and that doesn't seem like it should be too difficult - in the chat space, add integration, webhook, copy the URL and paste that into the URL of the webhook notifications panel in Protect.

 

Does Google Chat directly accept HTTP POST requests in the exact format detailed in that documentation page you linked to above? I'd guess it has its own API and, as you suggest, probably an authentication layer (OAuth2, being Google). My guess is that you would need some sort of server (possibly local) to accept the webhook call and then trigger a login-and-post-a-message sequence on Google Chat. If you're lucky, one of the automation platforms (IFTTT / Zapier) might have a pre-build handler for UniFi webhooks that lets you connect them up to Google Chat, otherwise a custom webhook handler is the kind of thing you ask an AI assistant to write for you easily enough these days.

[synaesthesia]

Yeah, that's pretty much what Gemini has told me - although I wouldn't trust Gemini to do anything relating to Google accurately

 

I've got it semi-working by utilising an app-script, modified from the one detailed here for door automation duties: https://github.com/DBR-it/UniFi-Emergency-Lockdown-with-Google-Chat-Email-Alerts/blob/main/Setup Guide_ UniFi Protect Lockdown Alerts.pdf

 

I relented and it actually worked...

 

function doPost(e) {
  // 1. SECURITY: Set a custom secret token to prevent unauthorized hits
  var SECRET_TOKEN = "YourSuperSecretToken123"; 
  var incomingToken = e.parameter.token;
  
  if (incomingToken !== SECRET_TOKEN) {
    return ContentService.createTextOutput(JSON.stringify({"error": "Unauthorized"}))
                         .setMimeType(ContentService.MimeType.JSON);
  }

  // 2. TARGET: Paste your Google Chat Webhook URL here
  var googleChatWebhookUrl = "YOUR_GOOGLE_CHAT_WEBHOOK_URL_HERE";
  
  try {
    // 3. PARSE: Extract data from UniFi Protect's payload
    var payloadData = JSON.parse(e.postData.contents);
    var alarmName = "Unknown Alarm";
    
    if (payloadData && payloadData.alarm && payloadData.alarm.name) {
      alarmName = payloadData.alarm.name;
    }
    
    // 4. FORMAT: Build the Google Chat message text
    var messageText = "🚨 *UniFi Protect Alert* 🚨\n\n" +
                      "*Event:* " + alarmName + "\n" +
                      "*Time:* " + new Date().toLocaleString("en-US", {timeZone: "America/New_York"}); // Change to your timezone

    var chatPayload = { "text": messageText };
    
    // 5. SEND: Forward the formatted message to Google Chat
    var options = {
      "method": "post",
      "contentType": "application/json",
      "payload": JSON.stringify(chatPayload)
    };
    
    UrlFetchApp.fetch(googleChatWebhookUrl, options);
    
    return ContentService.createTextOutput(JSON.stringify({"status": "success"}))
                         .setMimeType(ContentService.MimeType.JSON);
                         
  } catch (error) {
    Logger.log("Error: " + error.toString());
    return ContentService.createTextOutput(JSON.stringify({"status": "error", "message": error.toString()}))
                         .setMimeType(ContentService.MimeType.JSON);
  }
}

 

Edited June 22 by synaesthesia

[synaesthesia]

Think we're starting to get a couple of false positives through. Although so far it's 100% accurate for our boys WC deployment regarding vape vs Lynx/Sure/Lidl De-SmellySpray, I suspect that it's less accurate regarding girls sprays, whether it's body sprays, perfumes, mists etc I'm not yet sure. Or is that more of an indication of what is in vape these days?

 

We can at least be sure 80% of hits in the girls lavs are accurate and to me that's still much of an improvement over other sensors.

 

Webhook is working really well to link with Google Chat for instant notification and I'm going to expand it to either display something on a digital signage display right opposite those WC's to alert whoever is on duty, or rig up some basic LED lighting system. A good excuse to break out a Pi!  (Although this is likely just going to be a precursor to grabbing an "AI speaker" if SLT are happy. Signs look good for that on the technical side, however policy & procedure may be the tricky bit for them to tackle!)

[dhicks]

The new vape / air quality sensor was featured on the PC Pro Podcast last week:

 

https://podcasts.apple.com/gb/podcast/episode-796-can-eu-regulation-really-make-it-easier/id563487808?i=1000774194980

 

It's the last section, near the end, if you want to fast forward. They sound as though they'd given the device a good try-out (the person doing the review runs a hardware testing lab) and it worked well. Interestingly, although UK stocks have sold out, they had no problem ordering from France instead, which might be a handy option for some.

[Noxid_w]

Hi all,

 

I'd like to add my 2 cents in. We've been running these in 2 toilets in our secondary.

At first, very impressed. Testing with vapes ourselves, and actual cases with students, we have caught many who admit to vaping as soon as we say we know it was them from a sensor and via camera outside toilets correlating with times. 

For alerting, I didn't want to create Unifi accounts for the heads of year, so we utilised a share mailbox and power automate to forward emails to the correct contacts. That works very well. We were going to do webhooks to go into Teams, but powerautomate insists on premium license to do this. If anyone has a workaround for this, please share

 

It was all sunshine and rainbows, until we discovered this week that antiperisprant can set them off. A student who was accused was horrified and then proved that it was his antiperisprant that set it off (Rock Face). I then tested this with another brand (Right Guard) which also set it off. 

Pretty much makes the product useless as if students get wind of this fault, they will just start carrying those around and claiming they sprayed that. As per unifi's guidance, i am testing the sensitivity slider to see if I can get it to stop these positives. Although initial findings it doesn't make a difference and still sets them off. I have a ticket with them awaiting feedback from the team.

[synaesthesia]

7 minutes ago, Noxid_w said:

Hi all,

 

I'd like to add my 2 cents in. We've been running these in 2 toilets in our secondary.

At first, very impressed. Testing with vapes ourselves, and actual cases with students, we have caught many who admit to vaping as soon as we say we know it was them from a sensor and via camera outside toilets correlating with times. 

For alerting, I didn't want to create Unifi accounts for the heads of year, so we utilised a share mailbox and power automate to forward emails to the correct contacts. That works very well. We were going to do webhooks to go into Teams, but powerautomate insists on premium license to do this. If anyone has a workaround for this, please share

 

It was all sunshine and rainbows, until we discovered this week that antiperisprant can set them off. A student who was accused was horrified and then proved that it was his antiperisprant that set it off (Rock Face). I then tested this with another brand (Right Guard) which also set it off. 

Pretty much makes the product useless as if students get wind of this fault, they will just start carrying those around and claiming they sprayed that. As per unifi's guidance, i am testing the sensitivity slider to see if I can get it to stop these positives. Although initial findings it doesn't make a difference and still sets them off. I have a ticket with them awaiting feedback from the team.

We're starting to find this with a few brands - however I think there needs to be a note of realism - there is not a single product on the market, not one capable of fully distinguishing the difference and they certainly shouldn't be used on their own for evidence. Still, so far I have found these better than Halo and Verkada by a noticeable margin and unsurmountable times better than the likes of VapeGuardian.

Still it will be interesting to see what they come back with, there may be a small tweak which makes all the difference.

[pete]

^ We found the Verkada kit was very sensitive to deoderant / body spray.  Especially in the girls toilets.

 

Whether that's overlap between flavouring compounds, I don't know (less likely to have something that tastes like Lynx Africa).

[Noxid_w]

7 minutes ago, synaesthesia said:

We're starting to find this with a few brands - however I think there needs to be a note of realism - there is not a single product on the market, not one capable of fully distinguishing the difference and they certainly shouldn't be used on their own for evidence. Still, so far I have found these better than Halo and Verkada by a noticeable margin and unsurmountable times better than the likes of VapeGuardian.

Still it will be interesting to see what they come back with, there may be a small tweak which makes all the difference.

I appreciate what you're saying, but I don't see how anybody can brand something as vape detecting if it alerts to non vapes. Unless you have a camera in the area, with sound, we can almost never assume the accuracy of these products. If students get wind of this fault, they have a golden ticket to carry those around and just say they used that. Although a bag search could verify that they have a vap on them I suppose. But students are savy and will hide them in other places.

I'm giving this feedback to a member of LT tomorrow, who will ultimatley decide whether we go ahead with getting more of these.

Unifi came back with the below, which is more or less what is on there setup page.
 

What you're seeing can occur because the sensor detects airborne aerosols, and some antiperspirant or deodorant sprays produce aerosol signatures that can resemble vape emissions. As a result, certain brands may trigger a vape alert.
(For reliable vape detection, install the sensor 2–2.5 m (6.6–8.2 ft) high in smaller or semi-enclosed spaces and away from direct airflow.
In enclosed spaces, the sensor can typically detect rapid vape events within several meters of the source. Exact performance influenced by room size and ventilation levels.
To minimize false positives, avoid installing the sensor near sources of airborne aerosols or vapors, including:

* Air fresheners and fragrance diffusers
* Aerosol sprays and cleaning products
* Humidifiers and steam sources
* Areas with frequent dust or particulate disturbances

The sensor performs best in locations where vaping activity is likely to occur and where airflow does not rapidly dilute or redirect airborne particles before they reach the sensor.

 

He did say he would feed this back to the team, and then closed my ticket. So probably won't do anything about it.

[synaesthesia]

Indeed - after all, they nearly all identify as air quality sensors first and foremost, and just happens they can pick up certain compounds. When you start looking closely at the more expensive models they even go into detail about how well they detect strains of THC, CBD, CO2 etc so can effectively differentiate between smoking, vaping, weed, cigarettes - but they're all working on the same physical sensor systems and do so via software analysis.

This is certainly a trap a lot fell into when they first saw "vape detection" systems hit the market a few years ago, loads of schools ran out to buy things at silly money and were ultimately disappointed because they didn't actually read enough into it. See a headline, throw cash at it.

The difference with the UniFi stuff is that price point - they go apples for apples with the likes of the £700-£1000 per unit "big boys" (when can we start calling them scammers/opportunists?) and even make the more budget models look daft. As a whole, the product is good but expectations need managing, as does how the schools actually work with them.

This is the stage we're at - I have a workable notification system working via Google chat, something in line to flash up to whoever is on duty outside those toilets (enough to use a wand on whoever is coming out) but it's now up to SLT to decide how this is managed from a staffing perspective, how they are acted upon and ultimately is it actually worth it; for the price, assuming they stay around that mark, I think they're certainly worthwhile. Our business manager made a good point, even if they're not vaping, you can see how bad the air quality goes when they're all spraying crap as that's a viable health and safety problem on it's own.

 

I wonder if anyone would dare install one in the PE changing rooms.

 

[STK91]

I've just seen a firmware update for the vape sensor Improved vape detection accuracy by reducing false positives.

[Garacesh]

On 09/07/2026 at 15:35, synaesthesia said:

[...]

This is the stage we're at - I have a workable notification system working via Google chat, something in line to flash up to whoever is on duty outside those toilets (enough to use a wand on whoever is coming out)

[...]

I wonder if anyone would dare install one in the PE changing rooms.

"a wand"?

 

I can't imagine how many false positives you'd get in the changing rooms! My first school was constant fire alarms needing to be silenced.. Okay, sure, 50% of the time it was active vandalism of the break glass boxes, but around 75% (of the remaining 50%) was the boys changing rooms. I'd definitely be curious as to see the air quality stats ~10 minutes to end of period though!