User defined conditions, what happened to them?

[Stephen2615]

Greetings,

I have finally updated the docker on my X64 system after some testing on it and my RPi and now, I see that the user defined conditions as in the labels.txt have disappeared.

The doco says:

Custom Models - Bring your own trained model and labels

and later:

Add Classes: Clear, Wisps, Mostly Cloudy, Overcast, Rain, Snow

My model is what I prefer to use instead of having to recreate one where I don't get snow.

Previously, changing anything in the GUI would also write the changes (as in labels) to the docker log but that doesn't happen anymore. All it says is:

INFO - Configuration saved to /config/alpaca_config.json

At the very beginning of the docker starting, it says:

root@cloudml:/opt/clouddet# docker logs simple-cloud-detect -f
Starting Unified Cloud Detection Service...
Converting model if needed...
2026-01-23 23:16:42,765 - INFO - No 'groups' attribute found in model config

Can I continue to use my own model and labels?

Stephen

[chvvkumar]

Yes you should use your own model and label files. I only included mine as an example. Did you copy your model/label files and bind mount them to the new container?

[Stephen2615]

I have had the model and label files in place (directory) for months. I just changed from dev to latest and that is when it reverted to your settings.

volumes:
- /opt/clouddet/keras_model.h5:/app/keras_model.h5
- /opt/clouddet/labels.txt:/app/labels.txt

[chvvkumar]

Stephen, when you bring up the container, does the UI show the same classes that are in your labels.txt file? The labels themselves can not be changed from the UI, only whether they are safe or not safe. Can you confirm what you were changing in this:

Previously, changing anything in the GUI would also write the changes (as in labels)

[Stephen2615]

The config web page doesn't show the classes from my labels.txt file. It just shows the six that your doco has.

The configuration web page system status shows the condition that is in my labels.txt ("allsky too hot" as an example which is when the enclosure is too hot to run the camera) but NINA thinks it is safe when it should not be.

The container log has:

2026-01-24 02:29:53,562 - INFO - Detection: {'class_name': 'allsky too hot', 'confidence_score': 100.0, 'Detection Time (Seconds)': 0.04, 'image': <PIL.Image.Image image mode=RGB size=969x729 at 0x76FBFEE6AB50>}
2026-01-24 02:29:53,565 - INFO - Published HA states: allsky too hot, 100.0%, 0.04s
2026-01-24 02:29:53,565 - INFO - Cloud detection: allsky too hot (100.0%)

Previously, all my conditions were in the labels.txt file and when I changed the condition to go from safe to unsafe, that would appear in the container log file and NINA would update pretty quickly. Changing any condition now doesn't display in the log. It just says:

Configuration saved to /config/alpaca_config.json

It only happened after going from the dev to the latest image. Nothing else has changed that I can determine.

[chvvkumar]

Ok let me do some testing and see what's going on, thanks for the information!

[chvvkumar]

Can you post your labels.txt, I will change mine to that and test.

[Stephen2615]

0 Allsky not running
1 allsky too hot
2 aurora
3 clear
4 clear with moon
5 overcast
6 overcast with bright object
7 partly cloudy
8 rain
9 wisps of cloud

[chvvkumar]

Alright, I replaced my labels with your labels (only the first six, as my model only uses six classifications) and this is what I am seeing:

I see that the actual condition is correct but the Safety Conditions section still showing my original classifications. Is that what you are seeing too?

If this is the issue, let me see what I can do, it might take me until tomorrow as we are getting a bad snow storm tomorrow and I have to make some preparations before it hits us.

[Stephen2615]

That is what I am seeing.

[Stephen2615]

[chvvkumar]

Ok thank you for your patience in explaining the issue! Now I understand it, I will see what is wrong and push a fix out.

[chvvkumar]

Ok I pushed a fix to snd let me know how that looks. It should work with your existing saved config but if it doesn't, delete alpaca_config.json and restart the container.

[Stephen2615]

I am using the snd version and it all looks good.

The only thing that I might suggest is a warning to the user that there could be a slight delay in state changes (debounce). I had some issue that might not have been an issue so I removed the alpaca_config.json and rebooted. If I had known there was a chance of a delay, I might not have done that. This is on my RPi that runs the Allsky Camera. I haven't updated the one I usually use.

I might wait until you release it in the latest. No imaging here for a week or so as we have a heatwave here with day temps climbing to about 110 F and nights about 85.

[chvvkumar]

The debounce logic is explained with the user set timings here correct? Or do you mean in the 'How timing works' section?