Missing docs: how to modify warning and failure thresholds

[andreipopovici]

Hi there,

I'm trying to fine-tune and make the detection less sensitive. I see there's a THRESH constant of 0.08 in the Obico ML server.py, then there are a couple threshold-related number inputs in the HA integration, but those get set to something by the blueprint. I was under the mistaken assumption that we could set those inputs to our preferred values.

Is there a combination of changes I could make to achieve my goal?

Could you elaborate on how the server's THRESH constant relates to the return values from the ML server? Thank you!

[nberktumer]

Unfortunately I've hardcoded the failure thresholds in the automation calculations. It is possible to replace the hardcoded values with an input but need to go over the whole calculation functions.

I ported the following calculation functions to the HA blueprint. You can go over these functions and update the blueprint:
https://github.com/TheSpaghettiDetective/obico-server/blob/496605a62fcb790097c510c151994e9b2bf020c1/backend/lib/prediction.py

https://github.com/TheSpaghettiDetective/obico-server/blob/496605a62fcb790097c510c151994e9b2bf020c1/backend/app/models.py#L410

[andreipopovici]

Thank you! I'll observe the normalized p values for a while to get a sense of why it's so sensitive.

To that end, I've added a warn mode that bypasses the pause/stop button presses: https://github.com/andreipopovici/ha-bambu-lab-p1-spaghetti-detection/blob/main/blueprints/spaghetti_detection.yaml#L50

[ForceConstant]

@andreipopovici thanks, the warn feature is useful for me.