Skip to content

control.md

Latest commit

 

History

History
74 lines (51 loc) · 5.19 KB

control.md

File metadata and controls

74 lines (51 loc) · 5.19 KB
Raw

Control App Settings

The Control app is the AV / room control interface, typically shown on an in-room touch panel. It provides controls for the room's displays, sources, lighting, microphones, cameras, calls and room joining.

Set settings in Backoffice zone metadata under control_app for the standard /control/ URL. Use the organisation, region or building zone. See settings storage and priority. The examples below show the metadata details object, without an app wrapper.

For example, the key logo_light means:

{
    "logo_light": "https://example.com/logo.svg"
}

Note: Most of the Control app's behaviour — available inputs and outputs, lighting scenes, microphones, cameras, help pages, room joining and so on — is driven by the System and its module bindings, configured on the System in Backoffice rather than through these app settings.

Application Identity

Setting Type Default Description
name string "Control" Application name sent with PlaceOS API requests. If you do not set it, the app uses short_name.
short_name string "PlaceOS" Prefix for console log messages. The app also uses it as the API application name when name is not set.
prevent_space_init boolean true Whether to prevent loading the full list of bookable spaces on application startup. Leave this on — the Control app only talks to the room's own System, and loading every space can be slow in large organisations.

The app does not set the browser page title from settings.

Branding & Appearance

Setting Type Default Description
logo_light string or icon object "assets/logo-light.svg" Logo shown in the top bar, the tabbed room view and the help modal when the light theme is active. Accepts an image URL, or an icon object with a src field.
logo_dark string or icon object "assets/logo-dark.svg" Logo shown in the same places when the dark theme is active.
allow_dark_mode boolean false Whether the app is allowed to use the dark theme. When off, the app always uses the light theme (and the light logo) regardless of the user's theme preference.
environment_bar string — CSS colour for thin identifier bars shown at the top and bottom of the app, useful for telling environments apart (e.g. "#f59e0b" for staging). Leave unset for no bars.
banner object — Message banner at the top of the app. The user can close it. See below.

Logo settings accept either a plain image URL string or an icon object:

{
    "logo_light": { "type": "img", "src": "https://example.com/logo.svg" },
    "logo_dark": { "type": "img", "src": "https://example.com/logo-dark.svg" }
}

The banner object:

Field Type Description
id string Unique ID for the message. After a user closes a banner, it stays hidden until the id changes.
type "info", "warn" or "error" Colour style for the banner. Defaults to "info".
content string Message text to show.

Voice Assistant

The System's voice_control binding turns on the voice assistant. To give a command, say "Hey Room" and then the command, or tap the microphone button. A green ring shows on the button when the panel hears speech. If you say "Hey Room" and stop, the panel plays a short tone and waits for the command. While you give a command, the words show next to the button. Speech recognition and speech synthesis run on the device:

  • Speech to text: Whistle on the Needle Wasm engine (Apache-2.0 licence). The app uses English. The engine is in apps/control/src/assets/needle.
  • Text to speech: Matcha-TTS on LiteRT (English only, MIT licence).

The app downloads approximately 110 MB of model files when the voice assistant starts for the first time. While the speech to text model loads, the microphone button shows an hourglass and the panel ignores speech. The browser keeps the files in the Cache API for later visits. When the model files change, the app deletes the old files from the cache. By default, the files come from Hugging Face at fixed commits.

Setting Type Default Description
voice_models_url string — URL of a folder that holds copies of the model files. Use this setting when the panels cannot connect to Hugging Face. A relative URL starts at the app's base URL.

The folder must hold these files, with the same names as in the Hugging Face repositories:

  • whistle.cact from Cactus-Compute/whistle
  • matcha_textenc_fp16.tflite, matcha_decoder_fp16.tflite, matcha_vocoder_fp16.tflite, dp_g2p_matcha_fp16.tflite, emb.bin, g2p_dict.txt.gz, config.json and g2p_meta.json from litert-community/Matcha-TTS

Use the commits in apps/control/src/app/ui/speech/litert.ts. If the server is on a different origin, it must send CORS headers.

The browser must support Web Workers, AudioWorklet and WebAssembly. WebGPU makes speech synthesis faster, but it is not necessary.