SimpleCloudDetect - I made an ASCOM Alpaca based safety monitor and cloud detection using Allsky camera images

sebaromano · Dec 30, 2025, 07:49 PM

It looks quite nice and interesting. The GUI looks super nice and very intuitive. I would love to try it out for sure.

I have been using my own code (written in Python). Very simple solution that return a text file with either SAFE or UNSAFE wording which are used by ASCOM Generic FIle Driver (in NINA) to trigger the SAFE/UNSAFE condition during my aquisition session. Thats text file is an output of my APP upon of analysisng the image produced by AllSkeye software from my all sky camera and then based on AI traned model (Teachable Machine) assignd the SAFE/UNSAFE logic.

So I hope you don’t mind if I ask a few questions:

• Is your App compatible with Allskeye software? I needed to use this one since I use WIndows.

• Does your app return any text file that can be used in NINA in the Safety Monitor? Otherwise how does triger the Safety Monitoring condition?

• Is there a simple PDF user guide to read more in deep the functionalities?

Best,

Sebastiano

Thank you for taking a look! I am actually writing an architecture doc for this. It is in draft stage but can be found below:

https://github.com/chvvkumar/simpleCloudDetect/blob/snd/architecture.md

I actually started this project with the filesafetymonitor in mind but switched over to Alpaca as it seemed to be a better fit (and native TCP) for talking to NINA rather than trying to manage SMB or NFS mounts required to update the file where NINA can read from it.

As of right now the container does not really output a file per se. But That can easily added and if not, since this already does REST API for ALPACA, it would be trivial to write a script to query the API end point and write it to a local file.

API end point:

/api/v1/safetymonitor/{device_number}/issafe

Response with GET method:

{"ClientTransactionID": 0,"ErrorMessage": "","ErrorNumber": 0,"ServerTransactionID": 117,"Value": false}

sebaromano · Dec 30, 2025, 08:25 PM

Thanks for the additional details. So I guess this is not suitable for Windows users like me, correct?

I have not tried it under windows as I run it on the same Pi as my AllSky software or in Debian while testing. This requires network_mode: host for UDP functionality for now and host networking support in Windows was in beta stage last I checked it a while ago.

I know you can specify EXPOSE PORT/udp but I have not had the time to test this yet.

EDIT: To answer your question, I have not tried to use this in Windows yet.

french

je trouve ça super. je voulais faire de meme avec un script généré par IA qui compterai le nombre d’étoiles visible et analyserai si le nombre augmente ou diminue, maiis ton script est bien intelligent et élaboré que le mon idée je te félicite, c’est vraiment super (et tu m’économise beaucoup de travail car j’y connais pas grand-chose en développement logiciel)

si j’ai bien compris ce script tourne en docker sur linux ?

Anglais

I think it's great. I wanted to do the same thing with an AI-generated script that would count the number of visible stars and analyze whether the number increases or decreases, but your script is much smarter and more sophisticated than my idea. Congratulations, it's really great (and you've saved me a lot of work because I don't know much about software development).

If I understand correctly, this script runs in Docker on Linux?

chvvkumar · Jan 1, 2026, 01:44 AM

Frédéric Tapissier · Jan 1, 2026, 01:38 AM

je trouve ça super. je voulais faire de meme avec un script généré par IA qui compterai le nombre d’étoiles visible et analyserai si le nombre augmente ou diminue, maiis ton script est bien intelligent et élaboré que le mon idée je te félicite, c’est vraiment super (et tu m’économise beaucoup de travail car j’y connais pas grand-chose en développement logiciel)

si j’ai bien compris ce script tourne en docker sur linux ?

—-Google translation—-

Merci pour ton commentaire enthousiaste!

Pour répondre à ta question : Oui, tu as tout à fait raison, ce script est conçu pour tourner sous Docker sur Linux.

—-English—-

Thank you for your enthusiastic comment!

To answer your question: Yes, you are absolutely right, this script is designed to run under Docker on Linux.

Hello. Do you think I can run your Docker in Docker Desktop for Windows (I already have dockerized applications on it)

Added multiple client support. Now multiple clients can connect to and read from a single container. Theoretically many many clients can connect but I don't have a way to practically test the limits. I have tested with:

Rostokko · Jan 8, 2026, 01:45 PM

I just wanted to thanks @chvvkumar for this simple, yet powerful tool he shared.

I have tied it to my all sky camera (Deep Sky Dad), trained it as suggested through Google’s Teachable Machines (it doesn’t take long; the all sky camera feeds you plenty of images you can easily classify over a few days of varied sky conditions), and hooked it to my NINA sequence.

It works great and reliably; and if you see some specific “sky” getting a dubious classification, it’s easy enough to add that/those image/s to the correct bucket and update the model.

This past night it was a good mix of clouds-clear-clouds-clear; my sequence automatically waited for clear conditions before opening the backyard observatory; then shot targets for a few hours; then, when clouds rolled in, shot flats and wrapped it up (my choice is to just call it a night when the clouds roll in after a certain point - even at the risk of losing a change of clearing later - but the beauty of this is that you can design whatever logic fits you best!).

Thanks again for this work; the tool is quite polished and well thought out.

Thank you Rostokko for your feedback. Yes, the training data makes all the difference. The model can be trained to be very sensitive to rain, so much so that it can detect a single drop of rain on the dome and set unsafe conditions based off of that. I find it quicker and more sensitive than any other physical rain gauge I have tried it against.

📷 image.pngI have Home Assistant trigger a rain alert on my home Nest speakers whenever there is rain detected and while the scope is out and powered ON. I ended up using images from an year’s worth of data to train my model. About ~7000 images per class.