Merge branch 'main' into update-formula-aarch64-apple-darwin-aebb9adb…

…d09d50e29f009a6b56f7e45011e3d0c9

.github/workflows/release-app.yml

@@ -20,7 +20,7 @@ jobs:
       - uses: actions/checkout@v4
 
       - name: create draft release
-        uses: crabnebula-dev/cloud-release@v0.2.0
+        uses: crabnebula-dev/cloud-release@v0
         with:
           command: release draft ${{ secrets.CN_APP_SLUG }} --framework tauri
           api-key: ${{ secrets.CN_API_KEY }}
@@ -180,7 +180,7 @@ jobs:
           tauriScript: bunx tauri -v
 
       - name: Upload Assets to CrabNebula Cloud
-        uses: crabnebula-dev/cloud-release@v0.2.2
+        uses: crabnebula-dev/cloud-release@v0
         with:
           command: release upload ${{ secrets.CN_APP_SLUG }} --framework tauri
           api-key: ${{ secrets.CN_API_KEY }}

Cargo.toml

@@ -15,7 +15,7 @@ resolver = "2"
 
 
 [workspace.package]
-version = "0.1.73"
+version = "0.1.74"
 authors = ["louis030195 <[email protected]>"]
 description = ""
 repository = "https://github.com/mediar-ai/screenpipe"

Formula/screenpipe.rb

@@ -10,12 +10,11 @@ class Screenpipe < Formula
       sha256 "6b6b540d9565a13bbf52f8c14224dc7876f40861044313aa8c992b866e015973" # arm64
     else
       url "https://github.com/mediar-ai/screenpipe/releases/download/v#{version}/screenpipe-#{version}-x86_64-apple-darwin.tar.gz"
-      sha256 "538fdeb9024a041d29b6647a47381ef701c0fe46841b5bced98038fffa816fd4" # x86_64
+      sha256 "8acabe46b935ac8ed6607a43a5d47e81b5647ee2c52193bcb3b646628a4ced9b" # x86_64
     end
   end
 
   depends_on "ffmpeg"
-  depends_on "tesseract"
 
   def install
     bin.install Dir["bin/*"]

examples/apps/screenpipe-app-tauri/src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "screenpipe-app"
-version = "0.1.81"
+version = "0.1.82"
 description = ""
 authors = ["you"]
 license = ""

screenpipe-audio/src/core.rs

@@ -187,6 +187,7 @@ async fn run_ffmpeg(
     duration: Duration,
 ) -> Result<()> {
     debug!("Starting FFmpeg process");
+
     let mut command = Command::new(find_ffmpeg_path().unwrap());
     command
         .args(&[
@@ -195,13 +196,17 @@ async fn run_ffmpeg(
             "-ar",
             &sample_rate.to_string(),
             "-ac",
-            &channels.to_string(),
+            &if channels > 2 { 2 } else { channels }.to_string(),
             "-i",
             "pipe:0",
             "-c:a",
             "aac",
             "-b:a",
-            "128k",
+            "64k", // Reduced bitrate for higher compression
+            "-profile:a",
+            "aac_low", // Use AAC-LC profile for better compatibility
+            "-movflags",
+            "+faststart", // Optimize for web streaming
             "-f",
             "mp4",
             output_path.to_str().unwrap(),
@@ -412,10 +417,11 @@ pub async fn record_and_transcribe(
         error!("Error joining audio thread: {:?}", e);
     }
 
-    tokio::fs::File::open(&output_path_clone_2.to_path_buf())
-        .await?
-        .sync_all()
-        .await?;
+    // Commented to chekc if its the "Access is denied on windows" error
+    // tokio::fs::File::open(&output_path_clone_2.to_path_buf())
+    //     .await?
+    //     .sync_all()
+    //     .await?;
 
     debug!("Sending audio to audio model");
     if let Err(e) = whisper_sender.send(AudioInput {

screenpipe-audio/src/pcm_decode.rs

@@ -21,6 +21,7 @@ pub fn pcm_decode<P: AsRef<std::path::Path>>(path: P) -> anyhow::Result<(Vec<f32
 
     // Create a probe hint using the file's extension. [Optional]
     let hint = symphonia::core::probe::Hint::new();
+    // hint.with_extension("mp4");
 
     // Use the default options for metadata and format readers.
     let meta_opts: symphonia::core::meta::MetadataOptions = Default::default();
@@ -31,6 +32,12 @@ pub fn pcm_decode<P: AsRef<std::path::Path>>(path: P) -> anyhow::Result<(Vec<f32
     // Get the instantiated format reader.
     let mut format = probed.format;
 
+    // Debug log the format information
+    debug!(
+        "Detected format: {:?}",
+        format.metadata().current().unwrap()
+    );
+
     // Find the first audio track with a known (decodeable) codec.
     let track = format
         .tracks()

screenpipe-server/src/cli.rs

@@ -69,7 +69,7 @@ pub struct Cli {
     pub fps: f64, // ! not crazy about this (unconsistent behaviour across platforms) see https://github.com/mediar-ai/screenpipe/issues/173
 
     /// Audio chunk duration in seconds
-    #[arg(short = 'd', long, default_value_t = 30)]
+    #[arg(short = 'd', long, default_value_t = 120)]
     pub audio_chunk_duration: u64,
 
     /// Port to run the server on
@@ -148,7 +148,7 @@ pub struct Cli {
     #[arg(long, default_value_t = false)]
     pub use_pii_removal: bool,
 
-    /// Restart recording process every X minutes (0 means no periodic restart)
+    /// Restart recording process every X minutes (0 means no periodic restart) - NOT RECOMMENDED
     #[arg(long, default_value_t = 0)]
     pub restart_interval: u64,
 }

screenpipe-server/src/core.rs

@@ -255,8 +255,9 @@ async fn record_audio(
                     let device_control_clone = device_control_clone.clone();
 
                     let new_file_name = Utc::now().format("%Y-%m-%d_%H-%M-%S").to_string();
+                    let sanitized_device_name = audio_device_clone.to_string().replace(['/', '\\'], "_");
                     let file_path = PathBuf::from(&*output_path_clone)
-                        .join(format!("{}_{}.mp4", audio_device_clone, new_file_name))
+                        .join(format!("{}_{}.mp4", sanitized_device_name, new_file_name))
                         .to_str()
                         .expect("Failed to create valid path")
                         .to_string();

screenpipe-vision/src/microsoft.rs

@@ -39,4 +39,4 @@ pub async fn perform_ocr_windows(image: &DynamicImage) -> (String, String, Optio
     .to_string();
 
     (text, json_output, Some(1.0))
-}
+}