In-app voice commands

Tap the microphone immediately left of the filter icon in the main left drawer to toggle voice commands. The microphone turns orange while enabled. A small strip shows what was heard and the action status. Commands navigate the app without a spoken reply.

This prototype recognizes a fixed set of English commands:

• “Open drawer” or “open left drawer”.
• “Open workspace Website” or “open Website workspace”. “Project” is also accepted when spoken.
• “Open chat Fix keyboard overlap” or “open the keyboard chat”.
• “Show workspace tabs” or “take me to workspace tabs”.
• “Go back”.
• “Type fix the keyboard overlap” or “fill prompt fix the keyboard overlap”.
• “Send message”.
• “Stop listening”.

Pause after each command and wait for Listening before speaking again. Filling replaces the active prompt; sending is a separate command and uses the same submission path as the Send button. Commands wait briefly for an available composer and are abandoned if the chat changes while waiting. A busy chat must finish before this prototype can fill or send.

Names match the account’s loaded workspaces and inbox chats. Multiple matches appear as a picker; no match does not navigate. This is not a search over unloaded chat history.

Audio recognition is required to run on-device. If the device or current speech locale cannot support it, the mode shows an unavailable message instead of using a network recognizer. The command phrases themselves are English in this prototype. Remote workspace and chat operations still need their normal connection.

Listening pauses while a command is executing or while choosing a match. It stops when the app backgrounds, the account changes, another audio session interrupts it, dictation starts, or a conversational voice-agent session starts. Tap the microphone again to resume.