Extract local whisper inference as tower service

Cargo.toml

@@ -63,6 +63,7 @@ hypr-template = { path = "crates/template", package = "template" }
 hypr-transcribe-aws = { path = "crates/transcribe-aws", package = "transcribe-aws" }
 hypr-transcribe-azure = { path = "crates/transcribe-azure", package = "transcribe-azure" }
 hypr-transcribe-gcp = { path = "crates/transcribe-gcp", package = "transcribe-gcp" }
+hypr-transcribe-whisper-local = { path = "crates/transcribe-whisper-local", package = "transcribe-whisper-local" }
 hypr-turso = { path = "crates/turso", package = "turso" }
 hypr-whisper = { path = "crates/whisper", package = "whisper" }
 hypr-whisper-cloud = { path = "crates/whisper-cloud", package = "whisper-cloud" }

crates/transcribe-aws/Cargo.toml

@@ -21,6 +21,8 @@ tracing = { workspace = true }
 aws-config = "1.8.3"
 aws-sdk-transcribe = "1.83.0"
 aws-sdk-transcribestreaming = "1.80.0"
+aws-smithy-runtime-api = "1.8.5"
+aws-smithy-types = "1.3.2"
 
 [dev-dependencies]
 hypr-data = { workspace = true }

crates/transcribe-aws/src/error.rs

@@ -1,7 +1,21 @@
 #[derive(Debug, thiserror::Error)]
 pub enum Error {
-    #[error("Invalid input: {0}")]
-    InvalidInput(String),
-    #[error("Service error: {0}")]
-    ServiceError(String),
+    #[error(transparent)]
+    GenericError(#[from] aws_sdk_transcribestreaming::Error),
+    #[error(transparent)]
+    TranscriptResultStreamError(
+        #[from]
+        aws_smithy_runtime_api::client::result::SdkError<
+            aws_sdk_transcribestreaming::types::error::TranscriptResultStreamError,
+            aws_smithy_types::event_stream::RawMessage,
+        >,
+    ),
+    #[error(transparent)]
+    StartStreamTranscriptionError(
+        #[from]
+        aws_smithy_runtime_api::client::result::SdkError<
+            aws_sdk_transcribestreaming::operation::start_stream_transcription::StartStreamTranscriptionError,
+            aws_smithy_runtime_api::client::orchestrator::HttpResponse,
+        >,
+    ),
 }

crates/transcribe-aws/src/lib.rs

@@ -1,3 +1,5 @@
+// AWS draft
+
 use bytes::Bytes;
 use serde::{Deserialize, Serialize};
 use std::sync::Arc;
@@ -28,7 +30,7 @@ use aws_sdk_transcribestreaming::primitives::Blob;
 use aws_sdk_transcribestreaming::types::{
     AudioEvent, AudioStream, LanguageCode, MediaEncoding, TranscriptResultStream,
 };
-use aws_sdk_transcribestreaming::{config::Region, Client, Error};
+use aws_sdk_transcribestreaming::{config::Region, Client};
 
 mod error;
 pub use error::*;
@@ -72,7 +74,7 @@ pub struct TranscribeService {
 }
 
 impl TranscribeService {
-    pub async fn new(config: TranscribeConfig) -> Result<Self, Error> {
+    pub async fn new(config: TranscribeConfig) -> Result<Self, crate::Error> {
         let region_provider =
             RegionProviderChain::first_try(config.region.clone().map(Region::new))
                 .or_default_provider()
@@ -143,7 +145,7 @@ impl TranscribeService {
         &self,
         mut audio_rx: mpsc::Receiver<Bytes>,
         result_tx: mpsc::Sender<WsMessage>,
-    ) -> Result<(), Error> {
+    ) -> Result<(), crate::Error> {
         // Create audio stream for AWS Transcribe
         let input_stream = stream! {
             while let Some(chunk) = audio_rx.recv().await {
@@ -166,7 +168,6 @@ impl TranscribeService {
             .send()
             .await?;
 
-        // Process transcription results
         while let Some(event) = output.transcript_result_stream.recv().await? {
             match event {
                 TranscriptResultStream::TranscriptEvent(transcript_event) => {

crates/transcribe-interface/Cargo.toml

@@ -0,0 +1,9 @@
+[package]
+name = "transcribe-interface"
+version = "0.1.0"
+edition = "2021"
+
+[dependencies]
+serde = { workspace = true, features = ["derive"] }
+serde_json = { workspace = true }
+thiserror = { workspace = true }

crates/transcribe-interface/src/lib.rs

@@ -0,0 +1 @@
+

crates/transcribe-whisper-local/Cargo.toml

@@ -0,0 +1,38 @@
+[package]
+name = "transcribe-whisper-local"
+version = "0.1.0"
+edition = "2021"
+
+[features]
+default = []
+coreml = ["hypr-whisper-local/coreml"]
+directml = ["hypr-pyannote-local/directml"]
+cuda = ["hypr-whisper-local/cuda"]
+hipblas = ["hypr-whisper-local/hipblas"]
+openblas = ["hypr-whisper-local/openblas"]
+metal = ["hypr-whisper-local/metal"]
+vulkan = ["hypr-whisper-local/vulkan"]
+openmp = ["hypr-whisper-local/openmp"]
+load-dynamic = ["hypr-pyannote-local/load-dynamic"]
+
+[dependencies]
+hypr-audio-utils = { workspace = true }
+hypr-chunker = { workspace = true }
+hypr-listener-interface = { workspace = true }
+hypr-pyannote-local = { workspace = true }
+hypr-whisper = { workspace = true }
+hypr-whisper-local = { workspace = true }
+hypr-ws-utils = { workspace = true }
+
+serde_json = { workspace = true }
+serde_qs = { workspace = true }
+thiserror = { workspace = true }
+
+axum = { workspace = true, features = ["ws"] }
+futures-util = { workspace = true }
+tokio = { workspace = true }
+tokio-util = { workspace = true }
+tower = { workspace = true }
+tracing = { workspace = true }
+
+rodio = { workspace = true }

crates/transcribe-whisper-local/src/error.rs

@@ -0,0 +1,2 @@
+#[derive(Debug, thiserror::Error)]
+pub enum Error {}

crates/transcribe-whisper-local/src/lib.rs

@@ -0,0 +1,6 @@
+mod error;
+mod manager;
+mod service;
+
+pub use error::*;
+pub use service::*;