fix: all lint

Author: nekomeowww

Cargo.lock

@@ -22,6 +22,7 @@ members = [
     "apps/whisper-api",
     "apps/whisper-realtime",
 ]
+lints = { clippy = { nursery = {level = "warn", priority = -1}, pedantic = { level = "warn", priority = -1 }, missing_errors_doc = "allow", module_name_repetitions = "allow", unused_async = "allow", cast_precision_loss = "allow", cast_sign_loss = "allow", cast_possible_truncation = "allow", unused_self = "allow"}}
 
 [profile.dev]
 opt-level = 1

Cargo.toml

@@ -29,3 +29,6 @@ cuda = [
     "candle-nn/cuda",
     "candle-transformers/cuda"
 ]
+
+[lints]
+workspace = true

apps/orpheus-tts/Cargo.toml

@@ -21,7 +21,6 @@ crossbeam-channel = "0.5.15"
 tokenizers = "0.21.2"
 ndarray = "0.16.1"
 serde = { version = "1.0.219", features = ["derive"] }
-lazy_static = "1.5.0"
 rustfft = "6.4.0"
 ndarray-stats = "0.6.0"
 log = "0.4.27"
@@ -34,3 +33,6 @@ ort = { version = "2.0.0-rc.10", features = ["ndarray", "directml", "cuda"] }
 
 [target.'cfg(target_os = "linux")'.dependencies]
 ort = { version = "2.0.0-rc.10", features = ["ndarray", "cuda"] }
+
+[lints]
+workspace = true

apps/silero-vad-lite-whisper-realtime-ort/Cargo.toml

@@ -79,11 +79,11 @@ async fn process_audio_chunk(
   chunk: Vec<f32>,
   sample_rate: u32,
 ) -> Result<()> {
-  frame_buffer.extend_from_slice(&chunk);
-
   // Silero VAD works with specific frame sizes. 512 is a good choice for 16kHz audio.
   const VAD_FRAME_SIZE: usize = 512;
 
+  frame_buffer.extend_from_slice(&chunk);
+
   while frame_buffer.len() >= VAD_FRAME_SIZE {
     let frame: Vec<f32> = frame_buffer.drain(..VAD_FRAME_SIZE).collect();
 
@@ -92,14 +92,14 @@ async fn process_audio_chunk(
     let is_speech = vad.is_speech(speech_prob);
 
     // Add the frame to the audio buffer and check if a full utterance is ready
-    if let Some(complete_audio) = audio_buffer.add_chunk(&frame, is_speech) {
-      if !complete_audio.is_empty() {
-        // For debugging, you can save the audio chunk
-        // ... (your save_audio_chunk_to_file logic can go here)
-
-        // Transcribe the complete audio utterance
-        transcribe_audio(whisper, complete_audio, sample_rate).await?;
-      }
+    if let Some(complete_audio) = audio_buffer.add_chunk(&frame, is_speech)
+      && !complete_audio.is_empty()
+    {
+      // For debugging, you can save the audio chunk
+      // ... (your save_audio_chunk_to_file logic can go here)
+
+      // Transcribe the complete audio utterance
+      transcribe_audio(whisper, complete_audio, sample_rate).await?;
     }
   }
 

apps/silero-vad-lite-whisper-realtime-ort/src/main.rs

@@ -45,7 +45,7 @@ impl SileroVADModel {
       None => repo.download(model_path_sub_name)?,
     };
 
-    let session = Self::create_optimized_session(model_path.clone())?;
+    let session = Self::create_optimized_session(model_path)?;
 
     Ok(Self { session: Arc::new(Mutex::new(session)) })
   }
@@ -75,17 +75,18 @@ impl SileroVADModel {
 
   /// Stateless inference that matches JavaScript interface
   /// Returns both output and updated state like the JS version
+  #[allow(clippy::significant_drop_tightening)]
   pub fn inference(
     &self,
     input_data: SileroVADInput,
   ) -> Result<SileroVADResult> {
     // Validate input dimensions
-    if input_data.state.len() != 2 * 1 * 128 {
+    if input_data.state.len() != 2 * 128 {
       return Err(anyhow::anyhow!("State must have 256 elements (2*1*128), got {}", input_data.state.len()));
     }
 
     // Create input tensors for the ONNX model
-    let inputs = vec![("input", Tensor::from_array((vec![1, input_data.input.len()], input_data.input.clone()))?.into_dyn()), ("sr", Tensor::from_array(([1], vec![input_data.sr]))?.into_dyn()), ("state", Tensor::from_array((vec![2, 1, 128], input_data.state.clone()))?.into_dyn())];
+    let inputs = vec![("input", Tensor::from_array((vec![1, input_data.input.len()], input_data.input.clone()))?.into_dyn()), ("sr", Tensor::from_array(([1], vec![input_data.sr]))?.into_dyn()), ("state", Tensor::from_array((vec![2, 1, 128], input_data.state))?.into_dyn())];
 
     // Run inference and extract data while session is still locked
     let (state_data, speech_data) = {
@@ -127,7 +128,7 @@ impl SileroVADPipeline {
   ) -> Result<Self> {
     let model = SileroVADModel::new()?;
     // The initial state for the Silero VAD model is a tensor of shape (2, 1, 128) filled with zeros.
-    let state = vec![0.0; 2 * 1 * 128];
+    let state = vec![0.0; 2 * 128];
     Ok(Self { model, state, sample_rate: i64::from(sample_rate), threshold })
   }