** One could imagine adding a new tool, for example sensor_set_state, which would allow the AI to write a value to a virtual sensor. **
Example:
Reading a license plate to compare it with a list of authorized plates to open the gate.
Reading the pH of my pool dispenser, for information and alarm if the pH is too low or too high, and if the indication goes to « AL », being able to automatically restart the device.
Did you plan for it not to write to real devices?
As you mentioned electronics, I was thinking of Linky values, for example, for devices that can have set operations and not just get operations.
Hello @pierre-gilles,
Thanks for your quick response.
How does the « asked the AI » action work? Does it take a picture at time t and analyze it, or does it visualize for a few seconds and analyze the video? I lean towards the first hypothesis, because I have a refresh problem on the 7-segment displays. They are powered by multiplexing and, as the value is displayed on 2 digits, they are not lit at the same time.
To fix this problem, the AI would need to analyze over a few images. Is this possible, or is it already the case? I hope I’ve been clear.
Good point! I had limited it to devices with « read_only = true », but we should actually restrict it to devices in the « Virtual MQTT Devices » integration only.
Yes, it takes a picture at the moment the « Ask the AI » action is executed!
Ah, that’s more complex. Even if we added a dynamic option to inject multiple camera images, how can we ensure that the two images are taken at different times?
