Implement core speech-to-text pipeline

All major components: hotkey listener (rdev), audio capture (cpal), resampling (rubato), VAD (Silero ONNX), Parakeet v3 TDT transcription (ort), overlay window (winit+softbuffer), paste simulation (enigo+arboard), audio feedback (rodio), YAML config, CLI with clap, HuggingFace model download. ~2400 lines of Rust across 16 source files. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-10 16:47:46 +01:00
parent 6b737f92fe
commit 9b0bf7d9e3
22 changed files with 7750 additions and 0 deletions
@@ -0,0 +1,127 @@
+use anyhow::Result;
+use dialoguer::{Input, Select};
+
+use crate::config::{Accelerator, Config, OverlayPosition, PasteMethod, RecordingMode};
+
+pub fn show() -> Result<()> {
+    let config = Config::load()?;
+    let yaml = serde_yaml::to_string(&config)?;
+    println!("{yaml}");
+    Ok(())
+}
+
+pub fn reset() -> Result<()> {
+    let config = Config::default();
+    config.save()?;
+    println!("Config reset to defaults at {}", Config::path()?.display());
+    Ok(())
+}
+
+pub fn interactive() -> Result<()> {
+    let mut config = Config::load()?;
+
+    config.hotkey = Input::new()
+        .with_prompt("Hotkey")
+        .default(config.hotkey)
+        .interact_text()?;
+
+    let mode_idx = Select::new()
+        .with_prompt("Recording mode")
+        .items(&["push_to_talk", "toggle"])
+        .default(match config.mode {
+            RecordingMode::PushToTalk => 0,
+            RecordingMode::Toggle => 1,
+        })
+        .interact()?;
+    config.mode = match mode_idx {
+        0 => RecordingMode::PushToTalk,
+        _ => RecordingMode::Toggle,
+    };
+
+    config.cancel_key = Input::new()
+        .with_prompt("Cancel key")
+        .default(config.cancel_key)
+        .interact_text()?;
+
+    config.model = Input::new()
+        .with_prompt("Model")
+        .default(config.model)
+        .interact_text()?;
+
+    let accel_idx = Select::new()
+        .with_prompt("Accelerator")
+        .items(&["auto", "cpu", "cuda", "directml"])
+        .default(match config.accelerator {
+            Accelerator::Auto => 0,
+            Accelerator::Cpu => 1,
+            Accelerator::Cuda => 2,
+            Accelerator::DirectMl => 3,
+        })
+        .interact()?;
+    config.accelerator = match accel_idx {
+        0 => Accelerator::Auto,
+        1 => Accelerator::Cpu,
+        2 => Accelerator::Cuda,
+        _ => Accelerator::DirectMl,
+    };
+
+    config.gpu_device = Input::new()
+        .with_prompt("GPU device index")
+        .default(config.gpu_device)
+        .interact_text()?;
+
+    let paste_idx = Select::new()
+        .with_prompt("Paste method")
+        .items(&["ctrl_v", "shift_insert", "ctrl_shift_v", "clipboard_only"])
+        .default(match config.paste_method {
+            PasteMethod::CtrlV => 0,
+            PasteMethod::ShiftInsert => 1,
+            PasteMethod::CtrlShiftV => 2,
+            PasteMethod::ClipboardOnly => 3,
+        })
+        .interact()?;
+    config.paste_method = match paste_idx {
+        0 => PasteMethod::CtrlV,
+        1 => PasteMethod::ShiftInsert,
+        2 => PasteMethod::CtrlShiftV,
+        _ => PasteMethod::ClipboardOnly,
+    };
+
+    let overlay_idx = Select::new()
+        .with_prompt("Overlay position")
+        .items(&["top", "bottom", "none"])
+        .default(match config.overlay_position {
+            OverlayPosition::Top => 0,
+            OverlayPosition::Bottom => 1,
+            OverlayPosition::None => 2,
+        })
+        .interact()?;
+    config.overlay_position = match overlay_idx {
+        0 => OverlayPosition::Top,
+        1 => OverlayPosition::Bottom,
+        _ => OverlayPosition::None,
+    };
+
+    let feedback_idx = Select::new()
+        .with_prompt("Audio feedback")
+        .items(&["yes", "no"])
+        .default(if config.audio_feedback { 0 } else { 1 })
+        .interact()?;
+    config.audio_feedback = feedback_idx == 0;
+
+    let vad_idx = Select::new()
+        .with_prompt("VAD (voice activity detection)")
+        .items(&["enabled", "disabled"])
+        .default(if config.vad_enabled { 0 } else { 1 })
+        .interact()?;
+    config.vad_enabled = vad_idx == 0;
+
+    config.language = Input::new()
+        .with_prompt("Language")
+        .default(config.language)
+        .interact_text()?;
+
+    config.save()?;
+    println!("\nConfig saved to {}", Config::path()?.display());
+    Ok(())
+}