extended_openai_conversation
extended_openai_conversation copied to clipboard
Something went wrong: Service cover.open not found.
Here is how the conversation goes. cant seem to control garage doors. Lights and locks work fine Assistant:How can I assist? close garage doors Assistant:Garage doors are closed. open left garage door Assistant:Something went wrong: Service cover.open not found.
Thanks for reporting an issue.
I also experienced this with other entities, but I didn't figure out how to solve it yet. Could you try with named alias to garage doors?
I have cover
entities (curtains and blinds) and they work fine.
I have same issue. Seems that all my covers are working fine except the garage door. all the covers that work fine are tuya based and via cloud. The garage door is an esphome device. If I ask to open I get same response as [darkneo29] . If I ask to close, it will operate the door. So if the door is open it will close it but if the door is closed it will open it. Seems to be something to do with esphome cover component.
Also I have noticed that the tuya based ones have slight different card
This is the ESPhome Garage
Not sure if this is significant or not.
We didn't pass anything about tuya or esphome yet. The information we passed to gpt is entity_id
, name
, state
, and aliases
.
Could you change entity_id
to be something different from garage
and try again?
The gpt might have trained that uses cover.open
to open garage doors while cover.open_cover
is used for curtains and blinds.
Just tried that still no change.
changed it to be a blind. I even change the class in esphome from garage to blind.
I have also tried changing the cover template to a time based cover in esphome, still same issue. "Could not find cover.open". Odd it only happens for open but close works fine.
Found a work around I added If service cover.open is to be used, use cover.open_cover instead. to the prompt. Works great now.
Great!
If we can guess where cover.open
came from (eg. mqtt?), we can tell gpt not to use its grammar.
However, @joeljody77's prompt is the only way so far.
Is this problem still present when using gpt-4o-mini
?
It seems working for me without any work around.
Closing the issue.