crates/whisper-local/examples/direct_test.rs (1)

4-4: Use generic paths or CLI arguments instead of hardcoded user-specific path.

The hardcoded path /home/benediktb/... makes this example unusable for other developers.

The previous review suggested using dirs::data_dir() or accepting the path as a command-line argument. For reference, crates/transcribe-whisper-local/examples/show_data_dir.rs demonstrates the platform-appropriate data directory approach.

crates/whisper-local/src/model.rs (4)

68-76: Library code should return Result instead of panicking.

Lines 68 and 75 use unwrap() and panic!(), preventing callers from handling errors gracefully.

This issue was previously flagged. Library code should propagate errors via Result<Whisper, Error> rather than panicking.

---

106-107: Library code should return Result instead of panicking.

The unwrap_or_else with panic on line 106 prevents graceful error handling.

This was previously flagged. The build method should return Result<Whisper, Error> and propagate this error.

---

111-118: Remove "what" comment and avoid panic.

Line 111's comment violates the coding guideline, and line 118's panic was previously flagged.

Apply this diff to remove the comment:

-            // CPU only
             let mut cpu_params = WhisperContextParameters::default();

The panic issue was previously raised and remains unaddressed.

---

121-121: Library code should return Result instead of panicking.

The expect() call prevents callers from handling state creation failures.

This was previously flagged. With a Result-returning build() method, propagate this error using ? or map_err.

owhisper/owhisper-server/src/commands/run.rs (1)

17-19: Remove the duplicated function signature.

The second pub async fn handle_run(...) line is still present, so the file won’t compile (Rust rejects nested function definitions). Drop the inner signature so only one definition remains.

 /// Orchestrates audio transcription server with microphone input.
-pub async fn handle_run(args: RunArgs) -> anyhow::Result<()> {
-pub async fn handle_run(args: RunArgs) -> anyhow::Result<()> {
+pub async fn handle_run(args: RunArgs) -> anyhow::Result<()> {

crates/detect/src/browser/linux.rs (1)

45-74: Reset detection set each tick and tighten process matching.

detected_browsers only ever grows, so once (say) Firefox appears you never report it again after a restart. On top of that, stdout.contains(browser) fires on unrelated commands (e.g. “chromedriver”). Build the current set from the ps output each poll, compare against the previous set, and match the process name field instead of a raw substring.

-                        for browser in &BROWSER_NAMES {
-                            if stdout.contains(browser) {
-                                if let Ok(mut detected) = detected_browsers.lock() {
-                                    if !detected.contains(*browser) {
-                                        detected.insert(browser.to_string());
-                                        f(format!("{} running", browser));
-                                    }
-                                }
-                                // Silently continue if mutex is poisoned
-                            }
-                        }
+                        let mut current = std::collections::HashSet::new();
+                        for line in stdout.lines().skip(1) {
+                            if let Some(cmd) = line.split_whitespace().last() {
+                                if let Some(browser) = BROWSER_NAMES.iter().find(|name| *name == &cmd) {
+                                    current.insert(browser.to_string());
+                                }
+                            }
+                        }
+
+                        if let Ok(mut detected) = detected_browsers.lock() {
+                            for browser in current.difference(&*detected) {
+                                f(format!("{} running", browser));
+                            }
+                            *detected = current;
+                        }

crates/audio/src/speaker/linux.rs (1)

37-45: Mock speaker stream is ~480× too slow.

Line 39 only pushes a single sample every 10 ms, yet sample_rate() advertises 48 kHz. Consumers block for seconds waiting for enough data, exactly the issue raised in the earlier review.

Batch the silence so you emit ~480 samples per 10 ms (or adjust the advertised sample_rate). For example:

-            loop {
-                if sender.unbounded_send(0.0).is_err() {
-                    tracing::debug!("SpeakerStream channel closed, exiting thread");
-                    break;
-                }
-
-                thread::sleep(Duration::from_millis(10));
-            }
+            const SAMPLES_PER_BATCH: usize = 480; // 48 kHz / 100 batches per second
+            loop {
+                for _ in 0..SAMPLES_PER_BATCH {
+                    if sender.unbounded_send(0.0).is_err() {
+                        tracing::debug!("SpeakerStream channel closed, exiting thread");
+                        return;
+                    }
+                }
+
+                thread::sleep(Duration::from_millis(10));
+            }

crates/audio/src/lib.rs (2)

227-232: Propagate speaker initialization errors immediately.

If SpeakerInput::new() fails (line 227-230), from_speaker still returns Ok with speaker: None, deferring the error until stream() is called (line 257). This is inconsistent with from_mic which propagates errors immediately. A previous review raised a similar concern about this pattern.

Apply this diff to fail fast:

-    let speaker = match SpeakerInput::new() {
-        Ok(speaker) => { tracing::debug!("Successfully created SpeakerInput"); Some(speaker) }
-        Err(e) => { tracing::error!("Failed to create SpeakerInput: {}", e); None }
-    };
-
-    Ok(Self { source: AudioSource::RealtimeSpeaker, mic: None, speaker, speaker_device_name: None, data: None })
+    let speaker = SpeakerInput::new().map_err(|e| {
+        tracing::error!("Failed to create SpeakerInput: {}", e);
+        crate::Error::StreamInitFailed
+    })?;
+    tracing::debug!("Successfully created SpeakerInput");
+
+    Ok(Self { source: AudioSource::RealtimeSpeaker, mic: None, speaker: Some(speaker), speaker_device_name: None, data: None })

---

145-157: Wrap Linux-specific pactl command in platform guard.

The pactl command is Linux-specific (PulseAudio/PipeWire) and will fail on macOS/Windows. A previous review flagged this same issue. Wrap this block in #[cfg(target_os = "linux")] to prevent runtime failures on other platforms.

Apply this diff:

         .collect();

+    #[cfg(target_os = "linux")]
+    {
-    // Add virtual echo-cancel device if it exists
-    if std::process::Command::new("pactl")
-        .args(["list", "sources", "short"])
-        .output()
-        .map(|output| String::from_utf8_lossy(&output.stdout).contains("echo-cancel-source"))
-        .unwrap_or(false)
-    {
-        if !result.contains(&"echo-cancel-source".to_string()) {
-            tracing::debug!("Adding virtual echo-cancel-source device");
-            result.push("echo-cancel-source".to_string());
+        if std::process::Command::new("pactl")
+            .args(["list", "sources", "short"])
+            .output()
+            .map(|output| String::from_utf8_lossy(&output.stdout).contains("echo-cancel-source"))
+            .unwrap_or(false)
+        {
+            if !result.contains(&"echo-cancel-source".to_string()) {
+                tracing::debug!("Adding virtual echo-cancel-source device");
+                result.push("echo-cancel-source".to_string());
+            }
         }
     }

     tracing::debug!("Returning {} devices from list_mic_devices", result.len());

crates/whisper-local/src/model.rs (3)

70-70: Remove "what" comment.

This comment explains what the code does rather than why, violating the coding guideline.

As per coding guidelines

Apply this diff:

-        // Log the model path for debugging
         tracing::info!("Loading whisper model from: {}", model_path);

---

78-78: Remove "what" comment.

This comment explains the implementation rather than the rationale.

As per coding guidelines

Apply this diff:

-        // Try to initialize with GPU first, fall back to CPU if GPU fails
         let backends = crate::list_ggml_backends();

---

85-90: Remove "what" comment and extract parameter initialization.

Line 85's comment violates the coding guideline. The repeated parameter initialization across lines 86-90, 99-102, and 112-115 introduces duplication.

As per coding guidelines

Extract a helper function:

fn create_whisper_params(use_gpu: bool) -> WhisperContextParameters {
    let mut params = WhisperContextParameters::default();
    params.use_gpu = use_gpu;
    if use_gpu {
        params.gpu_device = 0;
    }
    params.flash_attn = false;
    params.dtw_parameters.mode = whisper_rs::DtwMode::None;
    params
}

Then simplify the initialization:

-            // Try GPU first
-            let mut gpu_params = WhisperContextParameters::default();
-            gpu_params.gpu_device = 0;
-            gpu_params.use_gpu = true;
-            gpu_params.flash_attn = false;
-            gpu_params.dtw_parameters.mode = whisper_rs::DtwMode::None;
+            let gpu_params = create_whisper_params(true);

crates/whisper-local/examples/list_backends.rs (1)

3-18: Remove redundant documentation.

The doc comment duplicates the implementation below and is unusual for an example binary. Doc comments are typically reserved for library APIs, not example executables.

Apply this diff:

-/// Lists available GGML backends and prints their kind, name, description, and memory stats.
-///
-/// # Examples
-///
-/// ```
-/// // Run the example binary to print detected backends to stdout.
-/// fn main() {
-///     let backends = whisper_local::list_ggml_backends();
-///     println!("Available backends:");
-///     for backend in backends {
-///         println!("  {}: {} - {} ({} MB free / {} MB total)",
-///                  backend.kind, backend.name, backend.description,
-///                  backend.free_memory_mb, backend.total_memory_mb);
-///     }
-/// }
-/// ```
 fn main() {

AGENTS.md (1)

60-62: Consider varying sentence structure.

The three consecutive sentences beginning with "Do not" could be reworded for better flow.

Apply this diff to vary the structure:

 ## Do Not
 - Do not commit large model binaries.
-- Do not add new formatters or global toolchains without discussion.
-- Do not enable heavy features for simple type export or CI smoke tests.
+- Avoid adding new formatters or global toolchains without discussion.
+- Keep heavy features disabled for simple type export or CI smoke tests.

crates/detect/src/mic/linux.rs (1)

48-68: Avoid blocking the async runtime with std::process::Command.

This background task runs on Tokio (note the async move and tokio::select!), but std::process::Command::output() is fully blocking. If pactl ever stalls—even briefly—you tie up a scheduler thread, delaying every other task on that worker. Please move the PulseAudio probe onto an async-friendly path (e.g. tokio::process::Command) or wrap it in spawn_blocking.

Apply this diff to switch to Tokio’s process API:

- use std::process::Command;
+ use tokio::process::Command;
@@
-                        if is_microphone_in_use() {
+                        if is_microphone_in_use().await {
@@
-fn is_microphone_in_use() -> bool {
-    // Check if any source-outputs exist (applications using microphone)
-    if let Ok(output) = Command::new("pactl")
-        .args(["list", "source-outputs", "short"])
-        .output()
-    {
+async fn is_microphone_in_use() -> bool {
+    // Check if any source-outputs exist (applications using microphone)
+    if let Ok(output) = Command::new("pactl")
+        .args(["list", "source-outputs", "short"])
+        .output()
+        .await
+    {

Also applies to: 99-108

crates/detect/src/browser/mod.rs (1)

39-41: Prefer structured logging over eprintln!.

This module otherwise uses structured logging, but eprintln! bypasses that and forces stderr noise on consumers embedding the detector. Please route the failure through tracing::warn! (or the crate’s chosen logger) so integrators can manage log levels properly.

crates/audio/src/speaker/mod.rs (1)

32-32: Remove "what" comments that duplicate obvious function signatures.

Per coding guidelines, comments should explain "why" not "what". These doc comments simply restate what the function signatures and cfg attributes already convey. Remove lines 32, 39, 47, 54, 71, 89, 94, and 100.

Apply this diff:

-    /// Initialize platform-specific speaker capture
     #[cfg(any(target_os = "macos", target_os = "windows", target_os = "linux"))]
     pub fn new() -> Result<Self> {
         let inner = PlatformSpeakerInput::new()?;
         Ok(Self { inner })
     }

-    /// Unsupported platforms return an error
     #[cfg(not(any(target_os = "macos", target_os = "windows", target_os = "linux")))]
     pub fn new() -> Result<Self> {
         Err(anyhow::anyhow!(
             "'SpeakerInput::new' is not supported on this platform"
         ))
     }

-    /// Consume input and open speaker stream
     #[cfg(any(target_os = "macos", target_os = "windows", target_os = "linux"))]
     pub fn stream(self) -> Result<SpeakerStream> {
         let inner = self.inner.stream();
         Ok(SpeakerStream { inner })
     }

-    /// Not supported on this platform
     #[cfg(not(any(target_os = "macos", target_os = "windows", target_os = "linux")))]
     pub fn stream(self) -> Result<SpeakerStream> {
         Err(anyhow::anyhow!(
             "'SpeakerInput::stream' is not supported on this platform"
         ))
     }

-    /// Poll for next audio sample
     fn poll_next(
         self: std::pin::Pin<&mut Self>,
         cx: &mut std::task::Context<'_>,
     ) -> std::task::Poll<Option<Self::Item>> {

-    /// Expose as async stream of f32 samples
     fn as_stream(&mut self) -> impl Stream<Item = f32> + '_ {
         self
     }

-    /// Sample rate in Hz
     #[cfg(any(target_os = "macos", target_os = "windows", target_os = "linux"))]
     fn sample_rate(&self) -> u32 {
         self.inner.sample_rate()
     }

-    /// Sample rate unavailable on unsupported targets
     #[cfg(not(any(target_os = "macos", target_os = "windows", target_os = "linux")))]
     fn sample_rate(&self) -> u32 {
         0
     }

Also applies to: 39-39, 47-47, 54-54, 71-71, 89-89, 94-94, 100-100

plugins/listener/src/fsm.rs (1)

205-228: Remove verbose "what" documentation that duplicates obvious signatures.

Per coding guidelines, comments should be minimal and explain "why" not "what". These extensive doc comment blocks (lines 205-228, 692-709, 749-778, 930-941) simply restate what the function signatures, parameter types, and return types already convey. The function names (setup_resources, update_session, common, on_transition) are self-explanatory, and the implementation shows the behavior.

Also applies to: 692-709, 749-778, 930-941

crates/audio/src/lib.rs (3)

86-86: Remove redundant "what" comments.

Per coding guidelines, comments should explain "why" not "what". Lines 86, 96, 106-109, and 171-174 simply restate what the function names already convey (get_default_mic_device_name, get_default_speaker_device_name, list_mic_devices, list_speaker_devices).

Also applies to: 96-96, 106-109, 171-174

---

113-124: Simplify error handling chain for device enumeration.

Lines 115-124 use confusing chaining: map for success (115-119), then map_err for error (120-123), then unwrap_or_else (124). The map already collects success into Vec<cpal::Device>, so the subsequent map_err doesn't transform the error in a useful way before unwrap_or_else swallows it. Simplify:

     let devices: Vec<cpal::Device> = host
         .input_devices()
-        .map(|devices| {
-            let device_vec: Vec<cpal::Device> = devices.collect();
-            tracing::debug!("Found {} input devices in list_mic_devices", device_vec.len());
-            device_vec
-        })
-        .map_err(|e| {
-            tracing::error!("Failed to enumerate input devices in list_mic_devices: {:?}", e);
-            e
-        })
-        .unwrap_or_else(|_| Vec::new());
+        .map(|devices| devices.collect::<Vec<_>>())
+        .unwrap_or_else(|e| {
+            tracing::error!("Failed to enumerate input devices in list_mic_devices: {:?}", e);
+            Vec::new()
+        });
+    tracing::debug!("Found {} input devices in list_mic_devices", devices.len());

---

212-218: Reduce logging verbosity in from_mic.

Two tracing calls (info + debug) for a single operation is excessive. Use debug level for both or consolidate into one call.

 pub fn from_mic(device_name: Option<String>) -> Result<Self, crate::Error> {
-    tracing::info!("Creating AudioInput from microphone with device name: {:?}", device_name);
+    tracing::debug!("Creating AudioInput from microphone: {:?}", device_name);
     let mic = MicInput::new(device_name)?;
-    tracing::debug!("Successfully created MicInput");

     Ok(Self { source: AudioSource::RealtimeMic, mic: Some(mic), speaker: None, speaker_device_name: None, data: None })
 }

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

crates/audio/src/bin/test_speaker_extended.rs (1)

12-12: Remove "what" comments; keep only "why" explanations.

Several comments describe what the code does rather than why, violating the coding guideline: "For comments, keep it minimal. It should be about 'Why', not 'What'."

As per coding guidelines.

Consider removing these comments:

• Line 12: "Create speaker audio input"
• Line 16: "Create stream"
• Line 20: "Test actual audio capture"
• Line 29: "Capture audio for a few seconds"
• Line 36: "Print progress every 0.5 seconds at 48kHz"
• Line 43: "Stop after ~3 seconds at 48kHz"

The code is self-explanatory without them.

Also applies to: 16-16, 20-20, 29-29, 36-36, 43-43

crates/audio/src/speaker/linux.rs (2)

40-40: Remove or rewrite "what" comments to explain "why".

Multiple comments describe what the code does rather than why, violating the coding guideline: "For comments, keep it minimal. It should be about 'Why', not 'What'."

As per coding guidelines.

Consider removing or rewriting these comments:

• Line 40: "Try PulseAudio first" → remove (obvious from control flow)
• Line 57: "Query PulseAudio for monitor sources" → remove (function name is clear)
• Line 72: "Find the first monitor source..." → remove (code is self-explanatory)
• Line 111: Keep the TODO as is
• Line 159: "Read audio data from PulseAudio" → remove
• Line 165: "Convert bytes to f32 samples and send to stream" → remove
• Line 168: "For stereo, we'll take the average..." → remove (code is clear)

Note: Line 201's comment is good because it explains why this approach avoids busy-looping.

Also applies to: 57-57, 72-72, 111-111, 159-159, 165-165, 168-168

---

156-156: Clarify buffer size comment for stereo frames.

The comment says "Buffer for 1024 f32 samples" but with 2-channel stereo audio, this is actually 512 stereo frames (each frame contains 2 samples). Consider clarifying to avoid confusion about frame vs. sample terminology.

-        let mut buffer = vec![0u8; 1024 * std::mem::size_of::<f32>()]; // Buffer for 1024 f32 samples
+        let mut buffer = vec![0u8; 1024 * std::mem::size_of::<f32>()]; // 1024 f32 samples = 512 stereo frames

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

.github/workflows/linux_packages_arch.yaml (1)

373-373: Upgrade deprecated GitHub Actions version — softprops/action-gh-release@v1.

Line 373 uses the deprecated softprops/action-gh-release@v1 action. Upgrade to v2 or later to align with the fixes needed in .github/workflows/linux_packages.yaml.

  - name: Upload package to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-pkg.outputs.pkg_path }}

.github/workflows/linux_packages_rpm.yaml (2)

116-120: Verify setup_protoc action has required metadata fields.

The past review flagged that ./.github/actions/setup_protoc/action.yaml is missing required top-level name and description fields. This is still reported by actionlint and will prevent the workflow from running. Ensure the action's action.yaml file includes:

name: "Setup Protocol Buffers Compiler"
description: "Sets up the Protocol Buffers compiler (protoc) with caching support"

---

294-300: Update softprops/action-gh-release to v2.

The past review flagged that softprops/action-gh-release@v1 is deprecated and too old to run on current GitHub Actions. This is still reported by actionlint. Upgrade to v2 and verify compatibility:

- uses: softprops/action-gh-release@v1
+ uses: softprops/action-gh-release@v2

Consult the v2 release documentation to confirm input/output compatibility, as the action's interface may have changed.

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

.github/workflows/linux_packages_rpm.yaml (1)

303-305: Upgrade softprops/action-gh-release from v1 to v2.

v1 has been deprecated for over a year and might not work on current GitHub Actions runners due to enforcing Node 20. The latest v2.3.2 is actively maintained and the files input remains compatible.

  - name: Upload RPM to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-rpm.outputs.rpm_path }}

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

.github/workflows/linux_packages.yaml (3)

310-310: Fix deprecated runner label at line 310 — ubuntu-20.04 was retired April 15, 2025.

The AppImage build matrix still references the deprecated ubuntu-20.04 runner, which will cause immediate workflow failure. Update to ubuntu-24.04 to match the build-linux-packages job at line 40.

          - target: "x86_64-unknown-linux-gnu"
            arch: "x86_64"
            features: "stt-openblas,llm-vulkan"
-           runner: "ubuntu-20.04"
+           runner: "ubuntu-24.04"

Also update the commented ARM64 section at line 315 so it's ready if ARM64 builds are re-enabled in the future.

---

292-292: Upgrade deprecated softprops/action-gh-release from v1 to v2.

Both release upload steps (lines 292 and 578) use the outdated softprops/action-gh-release@v1, which actionlint flags as incompatible with current GitHub Actions. Upgrade both to v2:

  - name: Upload .deb to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-deb.outputs.deb_path }}

  - name: Upload AppImage to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-appimage.outputs.appimage_path }}

Also applies to: 578-578

---

132-132: Add required metadata to custom action files.

Lines 132, 145, and 148 reference custom actions that are missing required name and description fields per GitHub Actions specifications. Actionlint flags all three:

• .github/actions/setup_protoc/action.yaml (line 132)
• .github/actions/pnpm_install/action.yaml (line 145)
• .github/actions/poetry_install/action.yaml (line 148)

Each action.yaml file must include top-level name and description fields. Example for setup_protoc:

name: Setup Protocol Buffers Compiler
description: Installs Protocol Buffers compiler (protoc) with caching support
inputs:
  repo-token:
    required: true
    type: string
runs:
  ...

Apply the same pattern to pnpm_install and poetry_install with appropriate names and descriptions.

Also applies to: 145-145, 148-148

.github/workflows/linux_packages_arch.yaml (1)

406-406: Upgrade deprecated softprops/action-gh-release from v1 to v2.

Line 406 uses the outdated softprops/action-gh-release@v1, which actionlint flags as incompatible with current GitHub Actions. Upgrade to v2:

  - name: Upload package to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-pkg.outputs.pkg_path }}

CLAUDE.md (1)

68-70: Reduce repetitive sentence starters in the "Do Not" section.

Three consecutive sentences begin with "Do not," which reduces readability. Vary the phrasing for better flow.

Apply this diff to improve readability:

-## Do Not
-- Do not commit large model binaries.
-- Do not add new formatters or global toolchains without discussion.
-- Do not enable heavy features for simple type export or CI smoke tests.
+## Do Not
+- Avoid committing large model binaries.
+- Do not add new formatters or global toolchains without discussion.
+- Heavy features should not be enabled for simple type export or CI smoke tests.

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

.github/workflows/linux_packages.yaml (2)

576-582: Upgrade deprecated softprops/action-gh-release@v1 to v2.

Line 578 uses the deprecated softprops/action-gh-release@v1, which is no longer supported on GitHub Actions. Upgrade to v2 (or later stable major version).

  - name: Upload AppImage to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-appimage.outputs.appimage_path }}

---

290-296: Upgrade deprecated softprops/action-gh-release@v1 to v2.

Line 292 uses the deprecated softprops/action-gh-release@v1, which is no longer supported on GitHub Actions. Upgrade to v2 (or later stable major version).

  - name: Upload .deb to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-deb.outputs.deb_path }}

.github/workflows/linux_packages.yaml (1)

310-315: Update commented ARM64 section to use supported runner.

While the ARM64 build is currently disabled, the commented section at line 315 references the deprecated ubuntu-20.04 runner (retired April 15, 2025). If ARM64 builds are re-enabled, this will cause immediate failure.

Update the commented section to use a supported runner for consistency and to prevent future issues:

  # ARM64 builds disabled temporarily due to cross-compilation dependency issues
  # - target: "aarch64-unknown-linux-gnu"
  #   arch: "aarch64"
  #   features: "stt-openblas,llm-vulkan"
-  #   runner: "ubuntu-20.04"
+  #   runner: "ubuntu-24.04"

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

.github/workflows/linux_packages.yaml (3)

131-149: Duplicate issue: Custom actions still missing required metadata.

Static analysis continues to report that three custom actions are missing required name and description metadata fields:

• Line 132: ./.github/actions/setup_protoc
• Line 145: ./.github/actions/pnpm_install
• Line 148: ./.github/actions/poetry_install

Each action.yaml file must include these top-level fields before the inputs or runs sections.

I can generate the required metadata for all three action files. Would you like me to create the updated action.yaml files with appropriate name and description fields for each action?

---

298-316: Update deprecated ubuntu-20.04 in commented-out ARM64 section to prevent future failures.

Line 315 in the commented-out ARM64 matrix references the deprecated ubuntu-20.04 runner, which was retired on April 15, 2025. When ARM64 builds are re-enabled, this will cause an immediate workflow failure. Update the commented section for consistency:

  # ARM64 builds disabled temporarily due to cross-compilation dependency issues
  # - target: "aarch64-unknown-linux-gnu"
  #   arch: "aarch64"
  #   features: "stt-openblas,llm-vulkan"
-  #   runner: "ubuntu-20.04"
+  #   runner: "ubuntu-24.04"

---

290-296: Upgrade deprecated softprops/action-gh-release@v1 to v2 at two locations.

Lines 292 and 581 use the deprecated v1 which will fail on GitHub Actions. Verification confirms v2 is compatible with the current files parameter format. Both instances need upgrading:

Line 292:

  - name: Upload .deb to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-deb.outputs.deb_path }}

Line 581:

  - name: Upload AppImage to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-appimage.outputs.appimage_path }}

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

.github/workflows/linux_packages.yaml (2)

577-583: Upgrade second occurrence of deprecated softprops/action-gh-release@v1 to v2.

Line 579 also uses the deprecated softprops/action-gh-release@v1. Apply the same upgrade for consistency:

  - name: Upload AppImage to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-appimage.outputs.appimage_path }}
    env:
      GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

---

290-296: Upgrade deprecated softprops/action-gh-release@v1 to v2.

Line 292 uses the deprecated softprops/action-gh-release@v1, which triggers Node 16 deprecation warnings on GitHub Actions. Upgrade to v2:

  - name: Upload .deb to release
    if: github.event_name == 'release'
-   uses: softprops/action-gh-release@v1
+   uses: softprops/action-gh-release@v2
    with:
      files: ${{ steps.find-deb.outputs.deb_path }}
    env:
      GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

Apply the same upgrade to the AppImage upload step at line 579.

.github/workflows/linux_packages.yaml (1)

311-315: Update deprecated runner label in commented ARM64 section.

Line 315 contains ubuntu-20.04 in the commented ARM64 build configuration. While this code is currently disabled, when ARM64 builds are re-enabled, this deprecated runner will cause failures. Update to ubuntu-24.04 now to prevent future issues:

          # ARM64 builds disabled temporarily due to cross-compilation dependency issues
          # - target: "aarch64-unknown-linux-gnu"
          #   arch: "aarch64"
          #   features: "stt-openblas,llm-vulkan"
-         #   runner: "ubuntu-20.04"
+         #   runner: "ubuntu-24.04"

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro