Custom Assistant Intents

[OzGav]

Note that this discussion will be updated (including the attached files) as HA core updates occur. (Updated for HA 2024.7.0 All Volume, Pause, Resume, Previous and Next intents removed. Improvements have been made to attempt to mimic HA's capability to identify the correct player when targetting an area)

HA supports the commands of NEXT, PREVIOUS, PAUSE, UNPAUSE and VOLUME when targeted at a specific player or to an area. The PLAY command must be supported by the user a blueprint is available in a dedciated repository. For the remaining service calls (including targetting areas) you can use custom sentences. Note that targetting an area which has multiple players may result in unpredictable behaviour although for Mute, Clear Queue and Stop the player that is in the state of playing will be targeted.

Firstly add this file to \\YOUR_HA_IP\config\custom_sentences\en
media.yaml.zip

Of course you can change this file to reflect your language and change the en in the file path accordingly. Please share any translations below.

Next you need to add to your configuration.yaml file the following. Note I have the unique characters ma_ in my entity_id which is used to select the correct player when targetting an area. You could also check for players which have the mass_player_attribute. If you don't want to add those characters you can either use a different set of characters and adjust the templates accordingly or remove all of the area based intents and not use that functionality.

intent_script:
  MutePlayer:
    action:
      service: "media_player.volume_mute"
      data:
        entity_id: "{{ media_player }}"
        is_volume_muted: "{{ mute_or_unmute }}"
    speech:
      text: |
        {% if mute_or_unmute == "true" %}
          Muted
        {% else %}
          Unmuted
        {% endif %}

  MutePlayerArea:
    action:
      service: "media_player.volume_mute"
      data:
        entity_id: "{{ expand(states.media_player) | selectattr('entity_id', 'in', area_entities(area)) | select('search', 'ma_') | selectattr('state', 'eq', 'playing') | map(attribute='entity_id') | join }}"
        is_volume_muted: "{{ mute_or_unmute }}"
    speech:
      text: |
        {% if mute_or_unmute == "true" %}
          Muted
        {% else %}
          Unmuted
        {% endif %}

  ShuffleSet:
    action:
      service: "media_player.shuffle_set"
      data:
        entity_id: "{{ media_player }}"
        shuffle: "{{ shuffle_mode }}"
    speech:
      text: "Queue shuffle mode set"

  ShuffleSetArea:
    action:
      service: "media_player.shuffle_set"
      data:
        entity_id: "{{ expand(states.media_player) | selectattr('entity_id', 'in', area_entities(area)) | select('search', 'ma_') |map(attribute='entity_id') | join }}"
        shuffle: "{{ shuffle_mode }}"
    speech:
      text: "Queue shuffle mode set"

  RepeatSet:
    action:
      service: "media_player.repeat_set"
      data:
        entity_id: "{{ media_player }}"
        repeat: "{{ repeat_mode }}"
    speech:
      text: "Queue repeat mode set"

  RepeatSetArea:
    action:
      service: "media_player.repeat_set"
      data:
        entity_id: "{{ expand(states.media_player) | selectattr('entity_id', 'in', area_entities(area)) | select('search', 'ma_') |map(attribute='entity_id') | join }}"
        repeat: "{{ repeat_mode }}"
    speech:
      text: "Queue repeat mode set"

  ClearQueueArea:
    action:
      service: "media_player.clear_playlist"
      data:
        entity_id: "{{ expand(states.media_player) | selectattr('entity_id', 'in', area_entities(area)) | select('search', 'ma_') | selectattr('state', 'eq', 'playing') | map(attribute='entity_id') | join }}"
    speech:
      text: "{{ expand(states.media_player) | selectattr('entity_id', 'in', area_entities('study')) | select('search', 'ma_') |map(attribute='attributes.friendly_name') | join }} Queue cleared"

  ClearQueue:
    action:
      service: "media_player.clear_playlist"
      data:
        entity_id: "{{ media_player }}"
    speech:
      text: "Queue cleared"

  PlayerStop:
    action:
      service: "media_player.media_stop"
      data:
        entity_id: "{{ media_player }}"
    speech:
      text: "Stopped"

  PlayerStopArea:
    action:
      service: "media_player.media_stop"
      data:
        entity_id: "{{ expand(states.media_player) | selectattr('entity_id', 'in', area_entities(area)) | select('search', 'ma_') | selectattr('state', 'eq', 'playing') | map(attribute='entity_id') | join }}"
    speech:
      text: "Stopped"

  SelectSource:
    action:
      service: "media_player.select_source"
      data:
        entity_id: "{{ media_player }}"
        source: "{{ desired_source }}"
    speech:
      text: "{{ media_player }} source set to {{ desired_source }}"
      
  PlayerJoin:
    action:
      service: "media_player.join"
      target:
        entity_id: "{{ media_player_group }}"
      data:
        group_members: "{{ media_player }}"
    speech:
      text: "{{ states.media_player | selectattr('entity_id','eq',media_player) | map(attribute='attributes.friendly_name') | list | join }}  joined to {{ states.media_player | selectattr('entity_id','eq',media_player_group) | map(attribute='attributes.friendly_name') | list | join }} "

  PlayerUnjoin:
    action:
      service: "media_player.unjoin"
      target:
        entity_id: "{{ media_player }}"
    speech:
      text: "{{ states.media_player | selectattr('entity_id','eq',media_player) | map(attribute='attributes.friendly_name') | list | join }} removed from group"

[eric10k93]

so I tried adding everything as described here... if I use the assist debugging under "developer tools" I'm able to type in my sentences and see they are parsed and matched... BUT when I try to actually pass the exact same sentence to Assist, it gives me the error "Unexpected error during intent recognition" ... it seems to do this for any of the custom sentences... any thoughts? do you have this file working currently as described here?

[OzGav]

@eric10k93 You will get that if you try and target an area and there is more than one media player in that area. If that isn’t the problem look at the bottom of the Integration installation page of the docs for troubleshooting tips particularly increasing the logging so you can see what is failing . Don’t forget you can also see more information by going to Voice Assistants in the HA settings and selecting the one you are using and then debug in the top right.

edit: I am away from home so can’t try the area commands but I just tried shuffle set, repeat set and clear queue to a specific player and it worked perfectly

[eric10k93]

thank you, I removed everything targeting areas... I do think that was the issue

[ddppddpp]

Are these instructions valid as of HA 2025.1.0? I'm facing the same "Unexpected error during intent recognition" as described above.
Also, shouldn't we have a custom intent_script for the play_media functionality?

[OzGav]

@ra-kal this isn’t a LLM support thread. This is about adding custom intents to support media players actions in HA. You should ask your question in the HA forums.

[ra-kal]

I understand. I'm just trying to recover the lost function after updating to the official HA integration and removing the voice assistant (https://music-assistant.io/integration/voice/). I created a simple automation with "play {media}" sentence, but of course it does not have the power the MA had before. So I'm looking for ideas how to instruct HA to search for musing using MA and play it.

[OzGav]

I understand but this is not the correct thread.

[eric10k93]

@ddppddpp & @ra-kal , I just found there is a Blueprint for MA voice control here: https://github.com/music-assistant/blueprints/blob/main/music_assistant_blueprint.yaml

I'm sure it's a work in progress, but it's at least restored some of the functionality for me that broke with the move from HACS to the core integration. I'm able to play by voice using the Blueprint.

[ddppddpp]

Thanks, looks quite promising!

[skyw33]

As an alternative to renaming your entities, replacing
select('search', 'ma_')
with this:
selectattr('entity_id', 'in', integration_entities('music_assistant'))
in the provided configuration.yaml example above seems to work.

[RJrodrigo]

Is anyone working on a custom intent to add the currently playing song to favorites or a specific playlist?

addind songs to playlists is now possible: https://github.com/orgs/music-assistant/discussions/3147

[OzGav]

You will be able to do that yourself. Just use an intent that calls a script that calls that rest command. Note that the API calls are subject to change in the future.