Skip to main content
These two commands form the OS-automation layer of the dictation pipeline. The frontend handles audio capture and transcription via the Groq API, then calls transcription_complete to paste the result. copy_selected_text is used at the start of a recording session to capture any pre-selected text for the select-to-transform feature.

​
transcription_complete

Pastes the supplied text into the previously active application by writing it to the clipboard and simulating a Ctrl+V keystroke. 
import { invoke } from '@tauri-apps/api/core';

await invoke('transcription_complete', { text: 'Fix the authentication bug' });

​
Parameters

​
text
string
required
The fully formatted text to paste. This is the final output of the dictation pipeline — raw transcription optionally post-processed by the Groq Chat LLM.

​
Returns

Promise<void>

​
Behavior

Calling transcription_complete performs the following steps in order:
  1. Hides the main overlay window and waits 90ms for the OS to restore focus to the target application.
  2. Writes text to the system clipboard using the arboard crate.
  3. Simulates a Ctrl+V keystroke using the enigo crate to trigger a paste in the focused application.
  4. Waits 500ms before resolving, giving the target application time to process the paste.
If the enigo-based paste fails (e.g., on a Wayland session where direct key injection is unavailable), the backend falls back to leaving the text in the clipboard and emits a show-warning event to the frontend so the user can paste manually.

​
Full pipeline example

import { invoke } from '@tauri-apps/api/core';

async function completeDictation(audioBlob: Blob) {
  // 1. Transcribe via Groq Whisper (frontend network call)
  const rawText = await transcribeAudio(audioBlob);

  // 2. Optionally format with LLM
  const formattedText = await formatWithLLM(rawText, systemPrompt);

  // 3. Paste into the active application
  await invoke('transcription_complete', { text: formattedText });
}

​
copy_selected_text

Simulates Ctrl+C to copy whatever text is currently selected in the focused application, then reads and returns the clipboard contents. 
import { invoke } from '@tauri-apps/api/core';

const selectedText = await invoke<string | null>('copy_selected_text');

if (selectedText) {
  console.log('User had selected:', selectedText);
  // Use as context for LLM transformation
}

​
Returns

Promise<string | null> — The text that was in the clipboard after the simulated Ctrl+C, or null if nothing was selected.

​
Behavior

  1. Simulates a Ctrl+C keystroke via enigo.
  2. Reads the clipboard contents via arboard.
  3. Returns the clipboard text, or null if the clipboard is empty or contains non-text data.

​
Usage in the select-to-transform flow

copy_selected_text is called automatically at the start of each recording session (triggered from shortcuts.rs when the global hotkey fires). The selected text is passed to the LLM alongside the spoken instruction so the model can transform the selection according to the user’s spoken command. 
import { invoke } from '@tauri-apps/api/core';

// Called when recording starts
const selectedText = await invoke<string | null>('copy_selected_text');

// Later, when building the LLM prompt
const systemPrompt = selectedText
  ? `Transform the selected text according to the user's instruction: "${selectedText}"`
  : 'Transcribe and format the user\'s speech.';