A free, open source, and extensible speech-to-text application that works completely offline.
use clap::Parser;
use std::path::PathBuf;
#[derive(Parser, Debug, Clone, Default)]
#[command(name = "handy", about = "Handy - Speech to Text")]
pub struct CliArgs {
/// Start with the main window hidden
#[arg(long)]
pub start_hidden: bool,
/// Disable the system tray icon
#[arg(long)]
pub no_tray: bool,
/// Toggle transcription on/off (sent to running instance)
#[arg(long)]
pub toggle_transcription: bool,
/// Toggle transcription with post-processing on/off (sent to running instance)
#[arg(long)]
pub toggle_post_process: bool,
/// Cancel the current operation (sent to running instance)
#[arg(long)]
pub cancel: bool,
/// Enable debug mode with verbose logging
#[arg(long)]
pub debug: bool,
/// Transcribe this WAV (16 kHz mono) headlessly and exit. Runs the same
/// batch transcription path as the app — no mic, no silence removal, no
/// download (the model must already be installed). Long audio is still
/// chunked at pauses, as in the app.
#[arg(short = 'f', long, value_name = "WAV")]
pub transcribe_file: Option<PathBuf>,
/// Model id to load for --transcribe-file (default: the selected model).
#[arg(long)]
pub model: Option<String>,
/// Hard-select the compute device for --transcribe-file by its registry
/// index (see --list-devices). Omit to use the persisted accelerator
/// setting. transcribe-cpp (whisper-family) models only.
#[arg(long, value_name = "N")]
pub device_index: Option<usize>,
/// List the transcribe-cpp compute devices (with indices) and exit.
#[arg(long)]
pub list_devices: bool,
/// List the available models (with ids) and exit. Pass an id to --model.
/// Honors --json for machine-readable output.
#[arg(long)]
pub list_models: bool,
/// Repeat the transcription N times (best_ms reports the fastest run).
#[arg(long, value_name = "N")]
pub repeat: Option<usize>,
/// Emit --transcribe-file results as JSON.
#[arg(long)]
pub json: bool,
}