Expose device_id/satellite_id in intent_script context

[RamyBuilds]

Describe the feature

When a voice command triggers a custom intent via intent_script, the device_id and satellite_id of the triggering voice satellite should be accessible within the intent script context, just like they are in automations triggered with trigger: conversation. Currently, these values exist in the voice pipeline but are not passed to intent_script, making room-aware actions difficult.

Example commands

PlayAudio:
action:
- action: script.play_audio
data:
query: "{{ query }}"
device_id: "{{ device_id }}" # The satellite's device_id
satellite_id: "{{ satellite_id }}" # The satellite entity_id

PlayAudio:
action:
- action: script.play_audio
data:
query: "{{ query }}"
device_id: "{{ device_id }}"
satellite_id: "{{ satellite_id }}”

target_area: "{{ area_name(area_id(device_id)) }}"

Use cases

Use cases
1. Room-aware voice commands:
• Multiple voice satellites (M5 Atom Echo, ReSpeaker) in different rooms.
• Saying “play the Beatles” from the bedroom should play music in the bedroom without mentioning the room explicitly.
2. Multi-room audio management:
• Music Assistant, Sonos, and other integrations require the originating device for correct playback.
3. Improved UX for children and casual users:
• Users don’t have to remember or say the room name.
4. Commercial parity:
• Alexa and Google Home already handle this seamlessly; this feature brings similar functionality to Home Assistant.
5. Significant upgrade in speed from all but high end LLM setups.

Current workarounds and limitations:

Anything else?

This is for local Assist only. Coupled with an extensive custom_intents, it beats anything that cloud can do.

Would be happy to share my amateur hour yaml if useful.

The Community Blueprint has issues with multiple speakers or protocols in room, this would allow for a complete multi-room voice assist setup.

Intents are rapid!

[Bennett-Wendorf]

Absolutely necessary to make LLM voice pipelines more useful in my opinion.

Another use case:

• I have a script that creates reminders for me (really a local todo item with json in the description). My existing LLM pipeline does this very well except that it doesn't know what device it's on, so there really isn't a way to make sure the reminder goes off on the same device I asked from.