feat: Linux support + configurable push-to-talk key

crates/whimpr-asr/Cargo.toml

@@ -16,8 +16,8 @@ anyhow = { workspace = true }
 hound = "3"
 
 [target.'cfg(target_os = "macos")'.dependencies]
-whisper-rs = { version = "0.12", features = ["metal"] }
+whisper-rs = { version = "0.16", features = ["metal"] }
 
-# No GPU feature on Windows yet — CPU backend first, Vulkan/CUDA can follow.
+# No GPU feature on Windows/Linux yet — CPU backend first, Vulkan/CUDA can follow.
 [target.'cfg(not(target_os = "macos"))'.dependencies]
-whisper-rs = { version = "0.12" }
+whisper-rs = { version = "0.16" }

crates/whimpr-asr/src/lib.rs

@@ -59,13 +59,11 @@ impl AsrEngine for WhisperEngine {
             .full(params, pcm16k)
             .map_err(|e| anyhow::anyhow!("whisper full: {e}"))?;
 
-        let n = state
-            .full_n_segments()
-            .map_err(|e| anyhow::anyhow!("whisper n_segments: {e}"))?;
+        let n = state.full_n_segments();
         let mut text = String::new();
         for i in 0..n {
-            if let Ok(seg) = state.full_get_segment_text(i) {
-                text.push_str(&seg);
+            if let Some(seg) = state.get_segment(i) {
+                text.push_str(&seg.to_string());
             }
         }
 

crates/whimpr-core/src/settings.rs

@@ -36,6 +36,14 @@ pub struct Settings {
     pub anthropic_model: String,
     /// Play the record-start ping.
     pub sound_on_start: bool,
+    /// The push-to-talk key, stored as an rdev::Key variant string (e.g. "ControlRight").
+    /// Defaults to Right Ctrl.
+    #[serde(default = "default_push_to_talk_key")]
+    pub push_to_talk_key: String,
+}
+
+fn default_push_to_talk_key() -> String {
+    "ControlRight".to_string()
 }
 
 impl Default for Settings {
@@ -47,6 +55,7 @@ impl Default for Settings {
             openai_base_url: String::new(),
             anthropic_model: "claude-haiku-4-5".to_string(),
             sound_on_start: true,
+            push_to_talk_key: default_push_to_talk_key(),
         }
     }
 }

src-tauri/Cargo.toml

@@ -46,3 +46,10 @@ macos-accessibility-client = "0.0.1"
 objc2-app-kit = { version = "0.3", features = ["NSWorkspace", "NSRunningApplication"] }
 objc2-foundation = { version = "0.3", features = ["NSString"] }
 objc2-av-foundation = { version = "0.3", features = ["AVCaptureDevice", "AVMediaFormat"] }
+
+[target.'cfg(target_os = "linux")'.dependencies]
+# Global keyboard listener (X11 + Wayland). Needs /dev/input access — add user to
+# the 'input' group if events aren't received.
+rdev = "0.5"
+# Keyboard simulation is done via ydotool (Wayland) or xdotool (X11) as subprocess
+# calls, avoiding libxdo/libei C dependency issues.

src-tauri/src/appctx.rs

@@ -28,7 +28,24 @@ pub fn frontmost_bundle_id() -> Option<String> {
     }
 }
 
-#[cfg(not(target_os = "macos"))]
+#[cfg(target_os = "linux")]
+pub fn frontmost_bundle_id() -> Option<String> {
+    // Try xdotool first (most reliable on X11)
+    if let Ok(out) = std::process::Command::new("xdotool")
+        .args(["getactivewindow", "getwindowclassname"])
+        .output()
+    {
+        if out.status.success() {
+            let s = String::from_utf8_lossy(&out.stdout).trim().to_string();
+            if !s.is_empty() {
+                return Some(s);
+            }
+        }
+    }
+    None
+}
+
+#[cfg(target_os = "windows")]
 pub fn frontmost_bundle_id() -> Option<String> {
     None
 }

src-tauri/src/hotkey.rs

@@ -707,30 +707,9 @@ pub use crate::win::{
     history, install, rebuild_providers, stats_summary, update_settings,
 };
 
-// Other platforms (Linux, etc.): inert stubs so the crate still builds.
-#[cfg(not(any(target_os = "macos", target_os = "windows")))]
-mod other {
-    pub fn install(_app: tauri::AppHandle) {}
-    pub fn current_settings() -> whimpr_core::Settings {
-        whimpr_core::Settings::default()
-    }
-    pub fn update_settings(_new: whimpr_core::Settings) {}
-    pub fn rebuild_providers() {}
-    pub fn stats_summary(tz_offset_minutes: i32) -> whimpr_core::StatsSummary {
-        whimpr_core::StatsStore::default().summary(tz_offset_minutes, 0)
-    }
-    pub fn history(_limit: usize) -> Vec<whimpr_core::HistoryItem> {
-        Vec::new()
-    }
-    pub fn dictionary_entries() -> Vec<super::DictEntryDto> {
-        Vec::new()
-    }
-    pub fn dictionary_add(_correct: String, _mishears: Vec<String>) {}
-    pub fn dictionary_remove(_correct: &str) {}
-    pub fn dictionary_learn(_correct: String, _mishears: Vec<String>) {}
-}
-#[cfg(not(any(target_os = "macos", target_os = "windows")))]
-pub use other::{
+// Linux: real platform layer (rdev hotkey + enigo paste + full pipeline).
+#[cfg(target_os = "linux")]
+pub use crate::linux::{
     current_settings, dictionary_add, dictionary_entries, dictionary_learn, dictionary_remove,
     history, install, rebuild_providers, stats_summary, update_settings,
 };

src-tauri/src/lib.rs

@@ -13,6 +13,8 @@ mod local_llm;
 mod paste;
 #[cfg(target_os = "windows")]
 mod win;
+#[cfg(target_os = "linux")]
+mod linux;
 
 use serde::Serialize;
 use tauri::{
@@ -168,6 +170,16 @@ fn open_url(url: &str) {
     let _ = std::process::Command::new("open").arg(url).spawn();
 }
 
+#[cfg(target_os = "linux")]
+fn open_url(url: &str) {
+    let _ = std::process::Command::new("xdg-open").arg(url).spawn();
+}
+
+#[cfg(target_os = "windows")]
+fn open_url(url: &str) {
+    let _ = std::process::Command::new("cmd").args(["/c", "start", url]).spawn();
+}
+
 /// Request microphone access: trigger the native prompt (bundle has a usage string)
 /// by briefly opening the input device, and open the Microphone settings pane.
 #[tauri::command]
@@ -182,6 +194,18 @@ fn request_microphone() {
         });
         open_url("x-apple.systempreferences:com.apple.preference.security?Privacy_Microphone");
     }
+    #[cfg(target_os = "linux")]
+    {
+        // On Linux, microphone access is controlled by PipeWire/PulseAudio.
+        // Opening pavucontrol is the closest equivalent to macOS's permission pane.
+        std::thread::spawn(|| {
+            if let Ok(h) = whimpr_audio::start(|_: &[f32]| {}) {
+                std::thread::sleep(std::time::Duration::from_millis(400));
+                let _ = h.stop();
+            }
+        });
+        let _ = std::process::Command::new("pavucontrol").spawn();
+    }
 }
 
 /// Request Accessibility — the permission that makes the Fn key work in every app and
@@ -193,6 +217,13 @@ fn request_accessibility() {
         let _ = paste::prompt_accessibility();
         open_url("x-apple.systempreferences:com.apple.preference.security?Privacy_Accessibility");
     }
+    #[cfg(target_os = "linux")]
+    {
+        // On Linux, keyboard input access is via the 'input' group.
+        // Open the system settings or print instructions.
+        eprintln!("[whimpr] Linux: ensure your user is in the 'input' group: sudo usermod -aG input $USER");
+        let _ = open_url("https://wiki.archlinux.org/title/Input_device");
+    }
 }
 
 /// Request Input Monitoring (needed for the Fn key to be seen in every app, not
@@ -204,6 +235,10 @@ fn request_input_monitoring() {
         let _ = paste::request_input_monitoring();
         open_url("x-apple.systempreferences:com.apple.preference.security?Privacy_ListenEvent");
     }
+    #[cfg(target_os = "linux")]
+    {
+        eprintln!("[whimpr] Linux: input monitoring is handled via rdev (needs 'input' group membership)");
+    }
 }
 
 /// Save (or clear, when empty) an API key in the OS keychain, then rebuild providers