feat/server behind feature (#404)

* feat: move server behind feature flag

* don't include vulkan runtime by default

* fix test

.github/workflows/linux_special.yml

@@ -17,7 +17,7 @@ jobs:
                 include:
                     # Ubuntu 22.04
                     - platform: "ubuntu-22.04"
-                      args: '--features "vulkan"'
+                      args: '--features "vulkan" --features "server"'
                       pre-build-args: "--vulkan"
                       rust-version: "stable"
                       name: "ubuntu-22.04"

.github/workflows/release.yml

@@ -12,14 +12,14 @@ jobs:
             matrix:
                 include:
                     - platform: "macos-latest" # for Arm based macs (M1 and above).
-                      args: "--target aarch64-apple-darwin"
+                      args: '--target aarch64-apple-darwin --features "server"'
                     - platform: "macos-latest" # for Intel based macs.
-                      args: "--target x86_64-apple-darwin"
+                      args: '--target x86_64-apple-darwin --features "server"'
                     - platform: "ubuntu-22.04" # Ubuntu 22.04 x86_64 (Works on 24.04 as well)
-                      args: '--features "vulkan"'
+                      args: '--features "vulkan,server"'
                       pre-build-args: "--vulkan"
                     - platform: "windows-latest" # Windows x86_64
-                      args: '--target x86_64-pc-windows-msvc --features "vulkan"'
+                      args: '--target x86_64-pc-windows-msvc --features "vulkan,server"'
                       pre-build-args: "--vulkan"
 
         runs-on: ${{ matrix.platform }}

.github/workflows/windows_special.yml

@@ -22,14 +22,14 @@ jobs:
                 include:
                     # Windows no AVX2
                     - platform: "windows-latest"
-                      args: ''
+                      args: '--features "server"'
                       pre-build-args: "--older-cpu"
                       name: "older-cpu"
                       rust-version: "stable"
 
                     # Windows no AVX2
                     - platform: "windows-latest"
-                      args: '--features="vulkan"'
+                      args: '--features="vulkan,server"'
                       pre-build-args: "--older-cpu --vulkan"
                       name: "older-cpu-vulkan"
                       rust-version: "stable"

core/Cargo.toml

@@ -16,7 +16,6 @@ serde = { version = "1.0.195", features = ["derive"] }
 num = "0.4.1"
 once_cell = "1.19.0"
 which = "6.0.1"
-utoipa = { version = "4.2.3", features = ["axum_extras"] }
 tracing = { version = "0.1.40", features = ["log"] }
 eyre = { workspace = true }
 hound = "3.5.1"
@@ -32,6 +31,9 @@ serde_json = { workspace = true }
 futures-util = "0.3.30"
 pyannote-rs = "0.2.7"
 
+# Server
+utoipa = { version = "4.2.3", features = ["axum_extras"], optional = true }
+
 [dev-dependencies]
 
 serial_test = "3.1.1"
@@ -49,3 +51,4 @@ openblas = ["whisper-rs/openblas"]
 metal = ["whisper-rs/metal"]
 rocm = ["whisper-rs/hipblas"]
 vulkan = ["whisper-rs/vulkan"]
+server = ["dep:utoipa"]

core/src/config.rs

@@ -1,8 +1,9 @@
 use core::fmt;
 use serde::{Deserialize, Serialize};
-use utoipa::ToSchema;
 
-#[derive(Deserialize, Serialize, ToSchema)]
+#[derive(Deserialize, Serialize)]
+#[cfg_attr(feature = "server", derive(utoipa::ToSchema))]
+
 pub struct TranscribeOptions {
     pub path: String,
     pub lang: Option<String>,

core/src/test.rs

@@ -27,7 +27,7 @@ fn test_transcribe() {
         word_timestamps: None,
     };
     let start = Instant::now();
-    let result = crate::transcribe::transcribe(&ctx, options, None, None, None, None);
+    let result = crate::transcribe::transcribe(&ctx, options, None, None, None, None, None);
     println!("{:?}", result);
     println!(
         "Elapsed time: {:.2} seconds",

core/src/transcript.rs

@@ -1,7 +1,6 @@
 use eyre::Result;
 use num::integer::div_floor;
 use serde::{Deserialize, Serialize};
-use utoipa::ToSchema;
 
 pub fn format_timestamp(seconds: i64, always_include_hours: bool, decimal_marker: &str) -> String {
     assert!(seconds >= 0, "non-negative timestamp expected");
@@ -25,13 +24,15 @@ pub fn format_timestamp(seconds: i64, always_include_hours: bool, decimal_marker
     format!("{hours_marker}{minutes:02}:{seconds:02}{decimal_marker}{milliseconds:03}")
 }
 
-#[derive(Debug, Serialize, Deserialize, ToSchema, Clone)]
+#[derive(Debug, Serialize, Deserialize, Clone)]
+#[cfg_attr(feature = "server", derive(utoipa::ToSchema))]
 pub struct Transcript {
     pub processing_time_sec: u64,
     pub segments: Vec<Segment>,
 }
 
-#[derive(Debug, Serialize, Deserialize, ToSchema, Clone)]
+#[derive(Debug, Serialize, Deserialize, Clone)]
+#[cfg_attr(feature = "server", derive(utoipa::ToSchema))]
 pub struct Segment {
     pub start: i64,
     pub stop: i64,

desktop/src-tauri/Cargo.toml

@@ -42,12 +42,17 @@ futures = "0.3.30"
 cpal = "0.15.3"
 hound = "3.5.1"
 rand = "0.8.5"
-axum = "0.7.5"
-utoipa = { version = "4.2.3", features = ["axum_extras"] }
-utoipa-swagger-ui = { version = "7.1.0", features = ["axum"] }
+
+# Server
+axum = { version = "0.7.5", optional = true }
+utoipa = { version = "4.2.3", features = ["axum_extras"], optional = true }
+utoipa-swagger-ui = { version = "7.1.0", features = ["axum"], optional = true }
+
+# Log
 tracing = { version = "0.1.40", features = ["log"] }
 tracing-log = "0.2.0"
 tracing-subscriber = { version = "0.3.18", features = ["env-filter", "json"] }
+
 chrono = "0.4.38"
 crash-handler = "0.6.2"
 urlencoding = "2.1.3"
@@ -84,9 +89,11 @@ objc_id = "0.1"
 core-graphics-helmer-fork = "0.24.0"
 
 [features]
+default = []
 cuda = ["vibe_core/cuda"]
 coreml = ["vibe_core/coreml"]
 metal = ["vibe_core/metal"]
 openblas = ["vibe_core/openblas"]
 rocm = ["vibe_core/rocm"]
 vulkan = ["vibe_core/vulkan", "dep:ash"]
+server = ["dep:utoipa", "dep:utoipa-swagger-ui", "dep:axum", "vibe_core/server"]

desktop/src-tauri/src/cli.rs

@@ -9,7 +9,6 @@ use vibe_core::config::TranscribeOptions;
 use vibe_core::transcribe;
 
 use crate::cmd::get_models_folder;
-use crate::server;
 
 /// Attach to console if cli detected in Windows
 #[cfg(all(windows, not(debug_assertions)))]
@@ -115,7 +114,7 @@ struct Args {
     diarize_speaker_id_model: Option<String>,
 
     /// Run http server
-    #[arg(long)]
+    #[cfg_attr(feature = "server", clap(long))]
     server: bool,
 
     #[arg(long, default_value = "0.0.0.0")]
@@ -180,8 +179,9 @@ pub async fn run(app_handle: &AppHandle) -> Result<()> {
         args.format = "json".into();
     }
 
+    #[cfg(feature = "server")]
     if args.server {
-        server::run(app_handle.clone(), args.host, args.port).await?;
+        crate::server::run(app_handle.clone(), args.host, args.port).await?;
     }
     let lang = language_name_to_whisper_lang(&args.language)?;
     let options = TranscribeOptions {

desktop/src-tauri/src/cmd/mod.rs

@@ -228,14 +228,14 @@ impl Default for FfmpegOptions {
 }
 
 impl FfmpegOptions {
-    pub fn to_vec(&mut self) -> Vec<String> {
+    pub fn to_vec(&self) -> Vec<String> {
         let mut cmd = Vec::<String>::new();
         if let Some(custom_cmd) = &self.custom_command {
             cmd.extend(custom_cmd.split_whitespace().map(|s| s.to_string()));
         } else if self.normalize_loudness {
             cmd.extend(["-af".to_string(), "loudnorm=I=-16:TP=-1.5:LRA=11".to_string()]);
         }
-        return cmd;
+        cmd
     }
 }
 
@@ -245,7 +245,7 @@ pub async fn transcribe(
     options: vibe_core::config::TranscribeOptions,
     model_context_state: State<'_, Mutex<Option<ModelContext>>>,
     diarize_options: DiarizeOptions,
-    mut ffmpeg_options: FfmpegOptions,
+    ffmpeg_options: FfmpegOptions,
 ) -> Result<Transcript> {
     let model_context = model_context_state.lock().await;
     if model_context.is_none() {

desktop/src-tauri/src/main.rs

@@ -5,7 +5,10 @@ mod cli;
 mod cmd;
 mod config;
 mod panic_hook;
+
+#[cfg(feature = "server")]
 mod server;
+
 mod setup;
 mod utils;
 use tauri::{Emitter, Manager};

desktop/src-tauri/src/server.rs

@@ -11,7 +11,7 @@ use serde::{Deserialize, Serialize};
 use serde_json::Value;
 use tauri::Manager;
 use tokio::sync::Mutex;
-use utoipa::{OpenApi, ToSchema};
+use utoipa::OpenApi;
 use utoipa_swagger_ui::SwaggerUi;
 use vibe_core::config::TranscribeOptions;
 use vibe_core::transcript::{Segment, Transcript};
@@ -39,7 +39,7 @@ pub async fn run(app_handle: tauri::AppHandle, host: String, port: u16) -> eyre:
     Ok(())
 }
 
-#[derive(Deserialize, Serialize, ToSchema)]
+#[derive(Deserialize, Serialize, utoipa::ToSchema)]
 struct LoadPayload {
     pub model_path: String,
     pub gpu_device: Option<i32>,

desktop/src-tauri/tauri.windows.conf.json

@@ -9,8 +9,7 @@
             "timestampUrl": "http://timestamp.digicert.com"
         },
         "resources": {
-            "ffmpeg\\bin\\x64\\*": "./",
-            "vulkan_runtime\\x64\\*.dll": "./"
+            "ffmpeg\\bin\\x64\\*": "./"
         }
     }
 }