Optionally make text input default for Assist

[Bennett-Wendorf]

Describe the feature

Right now, if the default/selected voice assistant pipeline supports TTS and STT and the companion app has access to the microphone, selecting assist will default to automatically listening for voice input and will respond with TTS output. Often my phone ends up recognizing some input, and so the voice pipeline responds with something to the effect of "how can I help?" and then triggers the microphone again. I often end up getting stuck in a cycle where I can't get in text input to cancel the loop without talking.

I very much like the ability to talk back and forth in this manner, but usually when I pull up Assist on my phone I would rather type to it than talk (I can use my voice satellites if I wanted to talk).

It would be nice to have a setting to disable the automatic listening and instead start by showing the text input box with the microphone button next to it. Clicking that mic button could still start the voice dialog.

Use cases

I often want to just type to Assist on my phone, and it's frustrating to have to fight with the voice input to get to a point where I can just type. This is especially problematic when I'm in a quiet environment or in public and don't want my speaker going off.

Examples

Assist already supports text input on mobile, with a mic button next to the input field. I just want an option to make this the default input instead of listening right away.

Anything else?

No response

[bgoncal]

Currently what dictates how the App will open Assist is the frontend implementation, so if frontend asks to start text-only the App does, like if you add a tile card to your dashboard with the action to open assist just by text.

If you wish to change the behavior of the assist icon that lives in the top bar, then replace the tag from this discussion by frontend.

[Bennett-Wendorf]

I don't see a category for "frontend". If you think this is best classified as "Core functionality" or "Other UI enhancements", I'm more than happy to do that.

That being said, I'm a bit confused how this isn't a companion app issue. If you look at my first screenshot below, that is the current behavior when I first click the "top bar" assist icon on desktop. Notably, even on a device that supports the full voice pipeline, I still get the option to type first. I can then click the microphone button to launch into a voice pipeline with all the conversational features one would expect.

However, when I click what appears to be the same icon in the companion app, I'm met immediately with the interface in the second screenshot. The only way I can get to a text box to type instead is to first figure out how to get the AI to stop talking to me. Oftentimes my STT engine picks up on some words (even if there really isn't anything there), and so the rest of my pipeline tries to interpret what it said. Most of time this leads to it asking something like "How can I help you?" and then prompting for audio input again.

To clarify my original message a bit, all I'm really looking for is an option for the mobile app to behave exactly like the desktop implementation of the feature. I want to have a text box first, and then the option to use voice.

Desktop:

Mobile:

[bgoncal]

The situation is that the app receives the command from the Home Assistant UI to open Assist in a certain way, and this way currently is using voice by default. Perhaps frontend has some logic to not apply this on desktop, I don't know.

We can leave this feature request open marked for iOS but in the end it will need to be a home-assistant/frontend change (which applies for Android as well)