Hi, great work!
I am using OpenAI integration and in assist I have activated the flag to prefer local intents before AI but it seems that when invoked by Alexa it is ignoring local custom intents (automations with voice triggers).
Hi,
Make sure you setup the home_assistant_agent_id configuration to the voice assistant agent ID you configured to prefer local intents.
Maybe you setup directly to the OpenAI integration service agent ID.
Another point is that you need to disable the home_assistant_room_recognition setting because it sends the echo device ID to the HA which does not understand this ID and starts to consider “complex” all the messages, ignoring the local agent.
Hello,
no luck.
I added the environment variable to disable areas and checked the conversation agent.
It is “conversation.chatgpt” that is openai with selected the prefer local intents…
It seems to not trigger automation that starts with the
trigger: conversation
“conversation.chatgpt” is the default agent_id from OpenAI integration.
You need the agent_id from the Voice Assistant agent configured to use OpenAI and fallback to local intent.
Go to Developer Tools > Actions > Conversation: Process and fill the values, execute and after executed, click in the “go to yaml mode” in the upper form.
This is strange…
In my home assistant I only have a local agent (not using IA) and this one:
As you can see, this is where I set the prefer local option.
These are the two only conversation entities I have… 
Ok, I removed the explicit conversation agent in environment variables and now, using the preferred set in HA, it seems to work good 
1 Like
No, unfortunately I was wrong…removing explicit conversation agent make it fall back to local assistant (so local intents works but no more OpenAI)
I believe that you need to check this step in the documentation to get the correct agent id you need:
1 Like
Thank you for you help.
I did it and it is conversation.chatgpt:
I noticed from the latest updates that now it does not put that hex code again but a better readable name…
Ok, please, open an issue in the github project.
Enable debug logs in Home Assistant and attach AWS and Home Assistant log in the ticket, I’ll take a look.
Hi Fabian, I dont know if it is only with me, but when I try to send a command, I get this erros on log:
What version are you using?
Is possible to attach all logs?
hi fabian.
this is the entire log screen, when sending the activate, it works fine.
but when I try to do some actions ( command or open dash ), I got this error.
The version is the latest from GitHub, I just imported it yesterday.
Hi everyone,
First of all, thank you for the great work – the skill works really well! I initially used the Alexa-hosted version, but kept running into the timeout limit when using LLM features. Since switching to the AWS-hosted version, everything runs much more smoothly.
I do have one suggestion for improvement:
It would be great to have a variable or setting that allows us to define whether the skill should speak a sentence upon launch. I’d love to trigger the skill via a physical button when entering a room, but in that case, a welcome phrase is a bit counterproductive.
Keep up the great work!
Hi, I’m glad you liked the skill.
Could you please access the github and describe your request there?
I can implement it as soon as I have some free time.