feat(audio): implement MKL-accelerated speech-to-text for Mac

.github/workflows/benchmark.yml

@@ -144,4 +144,4 @@ jobs:
       #     alert-threshold: "200%"
       #     comment-on-alert: true
       #     fail-on-alert: true
-      #     alert-comment-cc-users: "@louis030195"
+      #     alert-comment-cc-users: "@louis030195"

.github/workflows/release-app.yml

@@ -40,9 +40,11 @@ jobs:
             args: "--target x86_64-apple-darwin --features metal"
             target: x86_64-apple-darwin
           - platform: "ubuntu-22.04" # Ubuntu x86_64
-            args: "" # TODO CUDA, mkl
+            args: "--features mkl"
+            target: x86_64-unknown-linux-gnu
           - platform: "windows-latest" # Windows x86_64
-            args: "--target x86_64-pc-windows-msvc" # TODO CUDA, mkl?  --features "openblas"
+            args: "--target x86_64-pc-windows-msvc --features mkl"
+            target: x86_64-pc-windows-msvc
             pre-build-args: "" # --openblas
             # windows arm: https://github.com/ahqsoftwares/tauri-ahq-store/blob/2fbc2103c222662b3c6ee0cd71fcde664824f0ef/.github/workflows/publish.yml#L136
 
@@ -150,6 +152,8 @@ jobs:
             export PKG_CONFIG_PATH="/usr/local/opt/ffmpeg/lib/pkgconfig:$PKG_CONFIG_PATH"
             export PKG_CONFIG_ALLOW_CROSS=1
             export RUSTFLAGS="-C link-arg=-Wl,-rpath,@executable_path/../Frameworks -C link-arg=-Wl,-rpath,@loader_path/../Frameworks -C link-arg=-Wl,-install_name,@rpath/libscreenpipe.dylib"
+          elif [[ "${{ matrix.platform }}" == "ubuntu-22.04" || "${{ matrix.platform }}" == "windows-latest" ]]; then
+            export RUSTFLAGS="-C target-cpu=native"
           fi
           cargo build --release ${{ matrix.args }}
           ls -R target

screenpipe-audio/Cargo.toml

@@ -80,10 +80,12 @@ criterion = { workspace = true }
 memory-stats = "1.0"
 
 [features]
+default = ["metal"]
 metal = ["candle/metal", "candle-nn/metal", "candle-transformers/metal"]
-cuda = ["candle/cuda", "candle-nn/cuda", "candle-transformers/cuda"]
 mkl = ["candle/mkl", "candle-nn/mkl", "candle-transformers/mkl"]
 
+
+
 [[bin]]
 name = "screenpipe-audio"
 path = "src/bin/screenpipe-audio.rs"

screenpipe-audio/benches/stt_benchmark.rs

@@ -69,4 +69,4 @@ fn criterion_benchmark(c: &mut Criterion) {
 }
 
 criterion_group!(benches, criterion_benchmark);
-criterion_main!(benches);
+criterion_main!(benches);

screenpipe-audio/src/stt.rs

@@ -40,8 +40,8 @@ pub struct WhisperModel {
 impl WhisperModel {
     pub fn new(engine: Arc<AudioTranscriptionEngine>) -> Result<Self> {
         debug!("Initializing WhisperModel");
-        let device = Device::new_metal(0).unwrap_or(Device::new_cuda(0).unwrap_or(Device::Cpu));
-        info!("device = {:?}", device);
+        let device = Self::get_optimal_device()?;
+        info!("Using device: {:?}", device);
 
         debug!("Fetching model files");
         let (config_filename, tokenizer_filename, weights_filename) = {
@@ -86,6 +86,17 @@ impl WhisperModel {
             device,
         })
     }
+
+    fn get_optimal_device() -> Result<Device> {
+        if let Ok(device) = Device::new_metal(0) {
+            info!("Using Metal GPU");
+            Ok(device)
+        } else {
+            info!("Metal not available, falling back to CPU");
+            Ok(Device::Cpu)
+        }
+    }
+
 }
 
 #[derive(Debug, Clone)]