I noticed in the release notes assist_satellite.start_conversation
I have My Voice Assistant PE setup with LLM.
But Cannot select my PE it with assist_satellite.start_conversation
In the release notes i see (Starting a conversation with the default conversation agent is a bit more work and did not make it this release).
Does this mean it still in development for the Voice Assistant PE?
Running latest ESPHome Dev, updated to today’s build and get an error when running the update of the PE… Will follow what they do with this update…
Having the same issue. According to the docs you need an LLM setup (Assist Satellite - Home Assistant) and they explicitly mention OpenAI and not Claude (which is what I use), so I thought maybe only OpenAI is supported atm?
Can anyone using OpenAI confirm they also don’t see their satellite available to select?
Quick update - heard JLo and Paulus mention this on the Voice Chapter 9 stream tonight. They confirmed that it’s coming to voice satellites soon, its just up to an ESPHome update.
Thanks! this would work with Okay Nabu “Yes”
At least that seems to be alternative for know. Not using Node Red but indeed you can create something similar with helpers.
Announce the question “Do you want to Turn on the computer?”
Turn ON Helper for Question “Do you want to Turn on the computer”.
Wait for Answer Okay Nabu “Yes” / buildt in time for no response Turn off Question. “Do you want to Turn on the computer”
If Yes is received in time. Check the active question and trigger Turn on computer automation.
Okay Something nice in 3 automations to start to have conversation
My Home Assistant ask question and If I respond with yes or no it give corresponding answer the only thing is that.
You need to say Okay Nabu after the announcement.
And if you Say Yes it will reply Done and then trigger the correct response.
Saying no take also longer for The PE to respond with Done and continue
Is there away to use the text response Voice PE receives?
and push this through the next automation?
Create 3
Automation 1
alias: Mbot - Question Test Yes and No
description: “”
triggers:
trigger: state
entity_id:
input_boolean.mbot_question_random_question_test
from: “off”
to: “on”
conditions:
actions:
alias: Random Chat
data:
agent_id: conversation.chatgpt_4
text: >-
So I want you ask a random yes or No question to test my knowledge. Ask
a question like is this correct? Yes or No
The notification will be spoken out loud (so no emoji or *). Please use
whole numbers and try to sound advanced. This should be max 1 minutes in
length and to the point.
Automation 2
alias: Mbot - Followup to Yes and NO
description: “”
triggers:
trigger: state
entity_id:
input_boolean.mbot_answer_yes_received
to: “on”
id: Answer_yes
trigger: state
entity_id:
input_boolean.mbot_answer_no_received
to: “on”
id: Answer_no
conditions:
actions:
choose:
conditions:
condition: trigger
id:
Answer_yes
sequence:
alias: Random Chat
data:
agent_id: conversation.chatgpt_4
text: >-
You Recieved a Yes to your question
The notification will be spoken out loud (so no emoji or *).
Please use whole numbers and try to sound advanced. This should
be max 1 minutes in length and to the point.
alias: Random Chat
data:
agent_id: conversation.chatgpt_4
text: >-
You Recieved a No to your question
The notification will be spoken out loud (so no emoji or *).
Please use whole numbers and try to sound advanced. This should
be max 1 minutes in length and to the point.
I like using node red becouse in this case, with swich node i can use only one flow and 2 input tex to make how many “ask question” from satellite and answer i want. I think that using your or this method now is the only way.
Do you think that is it possible use conversation.process and use the conversation id?
action: conversation.process
data:
text: "hi ask me: do you want ope the cover?"
agent_id: conversation.google_generative_ai
conversation_id: cover123
But i cannot figure out how to tell assist to have the same conversation_id when i call “ok nabu”.