Using HA Voice Preview Edition with node-red?

I’m using HA Voice Preview Edition

I added a conversation section to my HA config, added an intent script, then added a Todolist card to Lovelace UI

So far so good, I can now speak to HA Voice and tell it to “add whisky” and it’ll insert it into my shopping list.

But I want to use node-red for all of this, and not be tweaking YAML config.

My end goal is to have nodered validate or correct the item being added so it resolves to something sensible (ie, ‘Lucozade’ and not ‘Lukes 8’ - which is what it adds right now)

Ive tried to integrate HA Voice with node-red and failed miserably.
Maybe I’m missing some crucial third-party nodes?

Is it even possible right now?

No other nodes are required. Use the sentence node for conversation. You’ll find examples towards the bottom of the page.

1 Like

I’ve used sentence node for some canned responses, but its still unclear how to grab the ‘[item]’ into node-red when I tell HA Voice to ‘add [item]’

Is there some template syntax I should be using to extract the [item] from the sentence node?

The exampes on that link show how to play a random MASH episode, but it still requires setting up specific nodes to handle MASH, rather than it just being generic

You would have to do the translation yourself. This is really a problem with speech to text. You would need to use a sentence trigger like add to the shopping list.

Then have it respond with “what would you like to add” from that response use a switch node to find “Lukes 8’” then a change node to change it to lucozade.

You could also try a sentence trigger like “Add Lukes 8”. You have to work around the limitations for now.

You can use sentence wildcards. In the Sentence node, set an output property (e.g., msg.results) to "results". Then connect a debug node to inspect the structure of the returned data. That’s where you’ll find the wildcard values extracted from the sentence.

Thanks a heap. This looks like the piece of the puzzle I was missing in nodered